Erlang/OTP System Documentation

Erlang/OTP System Documentation
Copyright © 1997-2017 Ericsson AB. All Rights Reserved.
Erlang/OTP System Documentation 9.0
June 20, 2017
Copyright © 1997-2017 Ericsson AB. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance
with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless
required by applicable law or agreed to in writing, software distributed under the License is distributed on an
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See
the License for the specific language governing permissions and limitations under the License. Ericsson AB. All
Rights Reserved..
June 20, 2017
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 1
1.1 Installing the Binary Release
1 Installation Guide
This section describes how to install Erlang/OTP on UNIX and Windows.
1.1 Installing the Binary Release
1.1.1 Windows
The system is delivered as a Windows Installer executable. Get it from http://www.erlang.org/download.html
Installing
The installation procedure is automated. Double-click the .exe file icon and follow the instructions.
Verifying
•
Start Erlang/OTP by double-clicking on the Erlang shortcut icon on the desktop.
Expect a command-line window to pop up with an output looking something like this:
Erlang/OTP 17 [erts-6.0] [64-bit] [smp:2:2]
Eshell V6.0
1>
•
(abort with ^G)
Exit by entering the command halt().
2> halt().
This closes the Erlang/OTP shell.
1.2 Building and Installing Erlang/OTP
1.2.1 Introduction
This document describes how to build and install Erlang/OTP-20. Erlang/OTP should be possible to build from source
on any Unix/Linux system, including OS X. You are advised to read the whole document before attempting to build
and install Erlang/OTP.
The source code can be downloaded from the official site of Erlang/OTP or GitHub.
•
•
http://www.erlang.org
https://github.com/erlang/otp
1.2.2 Required Utilities
These are the tools you need in order to unpack and build Erlang/OTP.
2 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.2 Building and Installing Erlang/OTP
Warning:
Please have a look at the Known platform issues chapter before you start.
Unpacking
•
•
GNU unzip, or a modern uncompress.
A TAR program that understands the GNU TAR format for long filenames.
Building
•
•
•
•
•
•
GNU make
Compiler -- GNU C Compiler, gcc or the C compiler frontend for LLVM, clang.
Perl 5
GNU m4 -- If HiPE (native code) support is enabled. HiPE can be disabled using --disable-hipe
ncurses, termcap, or termlib -- The development headers and libraries are needed, often known as
ncurses-devel. Use --without-termcap to build without any of these libraries. Note that in this case
only the old shell (without any line editing) can be used.
sed -- Stream Editor for basic text transformation.
Building in Git
• GNU autoconf of at least version 2.59. Note that autoconf is not needed when building an unmodified
version of the released source.
Building on OS X
• Xcode -- Download and install via the Mac App Store. Read about Building on a Mac before proceeding.
Installing
•
An install program that can take multiple file names.
1.2.3 Optional Utilities
Some applications are automatically skipped if the dependencies aren't met. Here is a list of utilities needed for those
applications. You will also find the utilities needed for building the documentation.
Building
•
•
•
•
OpenSSL -- The opensource toolkit for Secure Socket Layer and Transport Layer Security. Required for building
the application crypto. Further, ssl and ssh require a working crypto application and will also be skipped
if OpenSSL is missing. The public_key application is available without crypto, but the functionality will
be very limited.
The development package of OpenSSL including the header files are needed as well as the binary command
program openssl. At least version 0.9.8 of OpenSSL is required. Read more and download from http://
www.openssl.org.
Oracle Java SE JDK -- The Java Development Kit (Standard Edition). Required for building the application
jinterface and parts of ic and orber. At least version 1.6.0 of the JDK is required.
Download from http://www.oracle.com/technetwork/java/javase/downloads. We have also tested with IBM's
JDK 1.6.0.
X Windows -- Development headers and libraries are needed to build the Erlang/OTP application gs on Unix/
Linux.
flex -- Headers and libraries are needed to build the flex scanner for the megaco application on Unix/Linux.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 3
1.2 Building and Installing Erlang/OTP
•
wxWidgets -- Toolkit for GUI applications. Required for building the wx application. At least version 3.0 of
wxWidgets is required.
Download from http://sourceforge.net/projects/wxwindows/files/3.0.0/ or get it from GitHub: https://
github.com/wxWidgets/wxWidgets
Further instructions on wxWidgets, read Building with wxErlang.
Building Documentation
•
•
xsltproc -- A command line XSLT processor.
A tool for applying XSLT stylesheets to XML documents. Download xsltproc from http://xmlsoft.org/XSLT/
xsltproc2.html.
fop -- Apache FOP print formatter (requires Java). Can be downloaded from http://xmlgraphics.apache.org/
fop.
1.2.4 How to Build and Install Erlang/OTP
The following instructions are for building the released source tar ball.
The variable $ERL_TOP will be mentioned a lot of times. It refers to the top directory in the source tree. More
information about $ERL_TOP can be found in the make and $ERL_TOP section below. If you are building in git you
probably want to take a look at the Building in Git section below before proceeding.
Unpacking
Start by unpacking the Erlang/OTP distribution file with your GNU compatible TAR program.
$ tar -zxf otp_src_20.0.tar.gz
# Assuming bash/sh
Now change directory into the base directory and set the $ERL_TOP variable.
$ cd otp_src_20.0
$ export ERL_TOP=`pwd`
# Assuming bash/sh
Configuring
Run the following commands to configure the build:
$ ./configure [ options ]
Note:
If you are building Erlang/OTP from git you will need to run ./otp_build autoconf to generate the
configure scripts.
By default, Erlang/OTP release will be installed in /usr/local/{bin,lib/erlang}. If you for instance don't
have the permission to install in the standard location, you can install Erlang/OTP somewhere else. For example, to
install in /opt/erlang/20.0/{bin,lib/erlang}, use the --prefix=/opt/erlang/20.0 option.
4 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.2 Building and Installing Erlang/OTP
On some platforms Perl may behave strangely if certain locales are set. If you get errors when building, try setting
the LANG variable:
$ export LANG=C
# Assuming bash/sh
Building
Build the Erlang/OTP release.
$ make
Testing
Before installation you should test whether your build is working properly by running our smoke test. The smoke test
is a subset of the complete Erlang/OTP test suites. First you will need to build and release the test suites.
$ make release_tests
This creates an additional folder in $ERL_TOP/release called tests. Now, it's time to start the smoke test.
$ cd release/tests/test_server
$ $ERL_TOP/bin/erl -s ts install -s ts smoke_test batch -s init stop
To verify that everything is ok you should open $ERL_TOP/release/tests/test_server/index.html
in your web browser and make sure that there are zero failed test cases.
Note:
On builds without crypto, ssl and ssh there is a failed test case for undefined functions. Verify that the failed
test case log only shows calls to skipped applications.
Installing
You are now ready to install the Erlang/OTP release! The following command will install the release on your system.
$ make install
Running
You should now have a working release of Erlang/OTP! Jump to System Principles for instructions on running Erlang/
OTP.
How to Build the Documentation
Make sure you're in the top directory in the source tree.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 5
1.2 Building and Installing Erlang/OTP
$ cd $ERL_TOP
If you have just built Erlang/OTP in the current source tree, you have already ran configure and do not need to
do this again; otherwise, run configure.
$ ./configure [Configure Args]
When building the documentation you need a full Erlang/OTP-20.0 system in the $PATH.
$ export PATH=$ERL_TOP/bin:$PATH
# Assuming bash/sh
For the FOP print formatter, two steps must be taken:
•
Adding the location of your installation of fop in $FOP_HOME.
$ export FOP_HOME=/path/to/fop/dir # Assuming bash/sh
•
Adding the fop script (in $FOP_HOME) to your $PATH, either by adding $FOP_HOME to $PATH, or by copying
the fop script to a directory already in your $PATH.
Build the documentation.
$ make docs
Build Issues
We have sometimes experienced problems with Oracle's java running out of memory when running fop. Increasing
the amount of memory available as follows has in our case solved the problem.
$ export FOP_OPTS="-Xmx<Installed amount of RAM in MB>m"
More information can be found at
•
http://xmlgraphics.apache.org/fop/0.95/running.html#memory.
How to Install the Documentation
The documentation can be installed either using the install-docs target, or using the release_docs target.
•
If you have installed Erlang/OTP using the install target, install the documentation using the installdocs target. Install locations determined by configure will be used. $DESTDIR can be used the same way
as when doing make install.
$ make install-docs
•
If you have installed Erlang/OTP using the release target, install the documentation using the release_docs
target. You typically want to use the same RELEASE_ROOT as when invoking make release.
6 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.2 Building and Installing Erlang/OTP
$ make release_docs RELEASE_ROOT=<release dir>
Accessing the Documentation
After installation you can access the documentation by
•
Reading man pages. Make sure that erl is referring to the installed version. For example /usr/local/bin/
erl. Try viewing at the man page for Mnesia
$ erl -man mnesia
•
Browsing the html pages by loading the page /usr/local/lib/erlang/doc/erlang/index.html or
<BaseDir>/lib/erlang/doc/erlang/index.html if the prefix option has been used.
How to Install the Pre-formatted Documentation
Pre-formatted html documentation and man pages can be downloaded from
•
http://www.erlang.org/download.html.
Extract the html archive in the installation directory.
$ cd <ReleaseDir>
$ tar -zxf otp_html_20.0.tar.gz
For erl -man <page> to work the Unix manual pages have to be installed in the same way, i.e.
$ cd <ReleaseDir>
$ tar -zxf otp_man_20.0.tar.gz
Where <ReleaseDir> is
•
•
•
<PrefixDir>/lib/erlang if you have installed Erlang/OTP using make install.
$DESTDIR<PrefixDir>/lib/erlang if you have installed Erlang/OTP using make install
DESTDIR=<TmpInstallDir>.
RELEASE_ROOT if you have installed using make release RELEASE_ROOT=<ReleaseDir>.
1.2.5 Advanced configuration and build of Erlang/OTP
If you want to tailor your Erlang/OTP build and installation, please read on for detailed information about the individual
steps.
make and $ERL_TOP
All the makefiles in the entire directory tree use the environment variable ERL_TOP to find the absolute path of the
installation. The configure script will figure this out and set it in the top level Makefile (which, when building, it
will pass on). However, when developing it is sometimes convenient to be able to run make in a subdirectory. To do
this you must set the ERL_TOP variable before you run make.
For example, assume your GNU make program is called make and you want to rebuild the application STDLIB, then
you could do:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 7
1.2 Building and Installing Erlang/OTP
$ cd lib/stdlib; env ERL_TOP=<Dir> make
where <Dir> would be what you find ERL_TOP is set to in the top level Makefile.
otp_build vs configure/make
Building Erlang/OTP can be done either by using the $ERL_TOP/otp_build script, or by invoking $ERL_TOP/
configure and make directly. Building using otp_build is easier since it involves fewer steps, but the
otp_build build procedure is not as flexible as the configure/make build procedure. The binary releases for
Windows that we deliver are built using otp_build.
Configuring
The configure script is created by the GNU autoconf utility, which checks for system specific features and then creates
a number of makefiles.
The configure script allows you to customize a number of parameters; type ./configure --help or ./
configure --help=recursive for details. ./configure --help=recursive will give help for all
configure scripts in all applications.
One of the things you can specify is where Erlang/OTP should be installed. By default Erlang/OTP will be installed in
/usr/local/{bin,lib/erlang}. To keep the same structure but install in a different place, <Dir> say, use
the --prefix argument like this: ./configure --prefix=<Dir>.
Some of the available configure options are:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
--prefix=PATH - Specify installation prefix.
--enable-plain-emulator - Build a threaded emulator that only uses one scheduler. This emulator type
is deprecated and will be removed in a future release.
--disable-threads - Build a non-threaded emulator. This emulator type is deprecated and will be
removed in a future release.
--{enable,disable}-kernel-poll - Kernel poll support (enabled by default if possible)
--{enable,disable}-hipe - HiPE support (enabled by default on supported platforms)
--{enable,disable}-fp-exceptions - Floating point exceptions (an optimization for floating point
operations). The default differs depending on operating system and hardware platform. Note that by enabling
this you might get a seemingly working system that sometimes fail on floating point operations.
--enable-m64-build - Build 64-bit binaries using the -m64 flag to (g)cc
--enable-m32-build - Build 32-bit binaries using the -m32 flag to (g)cc
--with-assumed-cache-line-size=SIZE - Set assumed cache-line size in bytes. Default is 64. Valid
values are powers of two between and including 16 and 8192. The runtime system use this value in order to
try to avoid false sharing. A too large value wastes memory. A to small value will increase the amount of false
sharing.
--{with,without}-termcap - termcap (without implies that only the old Erlang shell can be used)
--with-javac=JAVAC - Specify Java compiler to use
--{with,without}-javac - Java compiler (without implies that the jinterface application won't be
built)
--{enable,disable}-dynamic-ssl-lib - Dynamic OpenSSL libraries
--{enable,disable}-builtin-zlib - Use the built-in source for zlib.
--{with,without}-ssl - OpenSSL (without implies that the crypto, ssh, and ssl won't be built)
--with-ssl=PATH - Specify location of OpenSSL include and lib
--with-ssl-incl=PATH - Location of OpenSSL include directory, if different than specified by -with-ssl=PATH
8 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.2 Building and Installing Erlang/OTP
•
•
•
•
•
•
•
•
•
•
•
--with-ssl-rpath=yes|no|PATHS - Runtime library path for OpenSSL. Default is yes, which equates
to a number of standard locations. If no, then no runtime library paths will be used. Anything else should be a
comma separated list of paths.
--with-libatomic_ops=PATH - Use the libatomic_ops library for atomic memory accesses. If
configure should inform you about no native atomic implementation available, you typically want to try
using the libatomic_ops library. It can be downloaded from https://github.com/ivmai/libatomic_ops/.
--disable-smp-require-native-atomics - By default configure will fail if an SMP runtime
system is about to be built, and no implementation for native atomic memory accesses can be found. If
this happens, you are encouraged to find a native atomic implementation that can be used, e.g., using
libatomic_ops, but by passing --disable-smp-require-native-atomics you can build using a
fallback implementation based on mutexes or spinlocks. Performance of the SMP runtime system will however
suffer immensely without an implementation for native atomic memory accesses.
--enable-static-{nifs,drivers} - To allow usage of nifs and drivers on OSs that do not
support dynamic linking of libraries it is possible to statically link nifs and drivers with the main Erlang
VM binary. This is done by passing a comma separated list to the archives that you want to statically link.
e.g. --enable-static-nifs=/home/$USER/my_nif.a. The path has to be absolute and the
name of the archive has to be the same as the module, i.e. my_nif in the example above. This is also true
for drivers, but then it is the driver name that has to be the same as the filename. You also have to define
STATIC_ERLANG_{NIF,DRIVER} when compiling the .o files for the nif/driver. If your nif/driver depends
on some other dynamic library, you now have to link that to the Erlang VM binary. This is easily achieved by
passing LIBS=-llibname to configure.
--without-$app - By default all applications in Erlang/OTP will be included in a release. If this is not
wanted it is possible to specify that Erlang/OTP should be compiled without one or more applications, i.e. -without-wx. There is no automatic dependency handling between applications. If you disable an application
that another application depends on, you also have to disable the dependant application.
--enable-gettimeofday-as-os-system-time - Force usage of gettimeofday() for OS system
time.
--enable-prefer-elapsed-monotonic-time-during-suspend - Prefer an OS monotonic time
source with elapsed time during suspend.
--disable-prefer-elapsed-monotonic-time-during-suspend - Do not prefer an OS
monotonic time source with elapsed time during suspend.
--with-clock-resolution=high|low - Try to find clock sources for OS system time, and OS
monotonic time with higher or lower resolution than chosen by default. Note that both alternatives may have a
negative impact on the performance and scalability compared to the default clock sources chosen.
--disable-saved-compile-time - Disable saving of compile date and time in the emulator binary.
--enable-dirty-schedulers - Enable the experimental dirty schedulers functionality. Note that
the dirty schedulers functionality is experimental, and not supported. This functionality will be subject
to backward incompatible changes. Note that you should not enable the dirty scheduler functionality on
production systems. It is only provided for testing.
If you or your system has special requirements please read the Makefile for additional configuration information.
Atomic Memory Operations and the VM
The VM with SMP support makes quite a heavy use of atomic memory operations. An implementation providing
native atomic memory operations is therefore very important when building Erlang/OTP. By default the VM will
refuse to build if native atomic memory operations are not available.
Erlang/OTP itself provides implementations of native atomic memory operations that can be used when compiling
with a gcc compatible compiler for 32/64-bit x86, 32/64-bit SPARC V9, 32-bit PowerPC, or 32-bit Tile. When
compiling with a gcc compatible compiler for other architectures, the VM may be able to make use of native atomic
operations using the __atomic_* builtins (may be available when using a gcc of at least version 4.7) and/or using
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 9
1.2 Building and Installing Erlang/OTP
the __sync_* builtins (may be available when using a gcc of at least version 4.1). If only the gcc's __sync_*
builtins are available, the performance will suffer. Such a configuration should only be used as a last resort. When
compiling on Windows using a MicroSoft Visual C++ compiler native atomic memory operations are provided by
Windows APIs.
Native atomic implementation in the order preferred:
•
•
•
•
•
The implementation provided by Erlang/OTP.
The API provided by Windows.
The implementation based on the gcc __atomic_* builtins.
If none of the above are available for your architecture/compiler, you are recommended to build and install
libatomic_ops before building Erlang/OTP. The libatomic_ops library provides native atomic memory
operations for a variety of architectures and compilers. When building Erlang/OTP you need to inform the build
system of where the libatomic_ops library is installed using the --with-libatomic_ops=PATH
configure switch.
As a last resort, the implementation solely based on the gcc __sync_* builtins. This will however cause lots
of expensive and unnecessary memory barrier instructions to be issued. That is, performance will suffer. The
configure script will warn at the end of its execution if it cannot find any other alternative than this.
Building
Building Erlang/OTP on a relatively fast computer takes approximately 5 minutes. To speed it up, you can utilize
parallel make with the -j<num_jobs> option.
$ export MAKEFLAGS=-j8
$ make
# Assuming bash/sh
If you've upgraded the source with a patch you may need to clean up from previous builds before the new build. Make
sure to read the Pre-built Source Release section below before doing a make clean.
Within Git
When building in a Git working directory you also have to have a GNU autoconf of at least version 2.59 on your
system, because you need to generate the configure scripts before you can start building.
The configure scripts are generated by invoking ./otp_build autoconf in the $ERL_TOP directory. The
configure scripts also have to be regenerated when a configure.in or aclocal.m4 file has been modified.
Note that when checking out a branch a configure.in or aclocal.m4 file may change content, and you may
therefore have to regenerate the configure scripts when checking out a branch. Regenerated configure scripts
imply that you have to run configure and build again.
Note:
Running ./otp_build autoconf is not needed when building an unmodified version of the released source.
Other useful information can be found at our GitHub wiki:
•
http://wiki.github.com/erlang/otp
OS X (Darwin)
Make sure that the command hostname returns a valid fully qualified host name (this is configured in /etc/
hostconfig). Otherwise you might experience problems when running distributed systems.
10 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.2 Building and Installing Erlang/OTP
If you develop linked-in drivers (shared library) you need to link using gcc and the flags -bundle flat_namespace -undefined suppress. You also include -fno-common in CFLAGS when compiling.
Use .so as the library suffix.
If you have Xcode 4.3, or later, you will also need to download "Command Line Tools" via the Downloads preference
pane in Xcode.
Building with wxErlang
If you want to build the wx application, you will need to get wxWidgets-3.0 (wxWidgets-3.0.0.tar.bz2 from
http://sourceforge.net/projects/wxwindows/files/3.0.0/) or get it from github with bug fixes:
$ git clone --branch WX_3_0_BRANCH git@github.com:wxWidgets/wxWidgets.git
Be aware that the wxWidgets-3.0 is a new release of wxWidgets, it is not as mature as the old releases and the OS
X port still lags behind the other ports.
Configure and build wxWidgets (on Mavericks - 10.9):
$ ./configure --with-cocoa --prefix=/usr/local
or without support for old versions and with static libs
$ ./configure --with-cocoa --prefix=/usr/local --with-macosx-version-min=10.9 --disable-shared
$ make
$ sudo make install
$ export PATH=/usr/local/bin:$PATH
Check that you got the correct wx-config
$ which wx-config && wx-config --version-full
Build Erlang/OTP
$
$
$
$
$
export PATH=/usr/local/bin:$PATH
cd $ERL_TOP
./configure
make
sudo make install
Pre-built Source Release
The source release is delivered with a lot of platform independent build results already pre-built. If you want to remove
these pre-built files, invoke ./otp_build remove_prebuilt_files from the $ERL_TOP directory. After
you have done this, you can build exactly the same way as before, but the build process will take a much longer time.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 11
1.2 Building and Installing Erlang/OTP
Warning:
Doing make clean in an arbitrary directory of the source tree, may remove files needed for bootstrapping
the build.
Doing ./otp_build save_bootstrap from the $ERL_TOP directory before doing make clean will
ensure that it will be possible to build after doing make clean. ./otp_build save_bootstrap will
be invoked automatically when make is invoked from $ERL_TOP with either the clean target, or the default
target. It is also automatically invoked if ./otp_build remove_prebuilt_files is invoked.
If you need to verify the bootstrap beam files match the provided source files, use ./otp_build
update_primary to create a new commit that contains differences, if any exist.
How to Build a Debug Enabled Erlang RunTime System
After completing all the normal building steps described above a debug enabled runtime system can be built. To do
this you have to change directory to $ERL_TOP/erts/emulator.
In this directory execute:
$ make debug FLAVOR=$FLAVOR
where $FLAVOR is either plain or smp. The flavor options will produce a beam.debug and beam.smp.debug
executable respectively. The files are installed along side with the normal (opt) versions beam.smp and beam.
To start the debug enabled runtime system execute:
$ $ERL_TOP/bin/cerl -debug
The debug enabled runtime system features lock violation checking, assert checking and various sanity checks to help
a developer ensure correctness. Some of these features can be enabled on a normal beam using appropriate configure
options.
There are other types of runtime systems that can be built as well using the similar steps just described.
$ make $TYPE FLAVOR=$FLAVOR
where $TYPE is opt, gcov, gprof, debug, valgrind, or lcnt. These different beam types are useful for
debugging and profiling purposes.
Installing
•
Staged install using DESTDIR. You can perform the install phase in a temporary directory and later move the
installation into its correct location by use of the DESTDIR variable:
$ make DESTDIR=<tmp install dir> install
The installation will be created in a location prefixed by $DESTDIR. It can, however, not be run from there.
It needs to be moved into the correct location before it can be run. If DESTDIR have not been set but
INSTALL_PREFIX has been set, DESTDIR will be set to INSTALL_PREFIX. Note that INSTALL_PREFIX
12 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.2 Building and Installing Erlang/OTP
in pre R13B04 was buggy and behaved as EXTRA_PREFIX (see below). There are lots of areas of use for an
installation procedure using DESTDIR, e.g. when creating a package, cross compiling, etc. Here is an example
where the installation should be located under /opt/local:
$ ./configure --prefix=/opt/local
$ make
$ make DESTDIR=/tmp/erlang-build install
$ cd /tmp/erlang-build/opt/local
$
# gnu-tar is used in this example
$ tar -zcf /home/me/my-erlang-build.tgz *
$ su Password: *****
$ cd /opt/local
$ tar -zxf /home/me/my-erlang-build.tgz
•
Install using the release target. Instead of doing make install you can create the installation in whatever
directory you like using the release target and run the Install script yourself. RELEASE_ROOT is used
for specifying the directory where the installation should be created. This is what by default ends up under /
usr/local/lib/erlang if you do the install using make install. All installation paths provided in the
configure phase are ignored, as well as DESTDIR, and INSTALL_PREFIX. If you want links from a specific
bin directory to the installation you have to set those up yourself. An example where Erlang/OTP should be
located at /home/me/OTP:
$ ./configure
$ make
$ make RELEASE_ROOT=/home/me/OTP release
$ cd /home/me/OTP
$ ./Install -minimal /home/me/OTP
$ mkdir -p /home/me/bin
$ cd /home/me/bin
$ ln -s /home/me/OTP/bin/erl erl
$ ln -s /home/me/OTP/bin/erlc erlc
$ ln -s /home/me/OTP/bin/escript escript
...
The Install script should currently be invoked as follows in the directory where it resides (the top directory):
$ ./Install [-cross] [-minimal|-sasl] <ERL_ROOT>
where:
•
•
•
•
•
-minimal Creates an installation that starts up a minimal amount of applications, i.e., only kernel and
stdlib are started. The minimal system is normally enough, and is what make install uses.
-sasl Creates an installation that also starts up the sasl application.
-cross For cross compilation. Informs the install script that it is run on the build machine.
<ERL_ROOT> - The absolute path to the Erlang installation to use at run time. This is often the same as the
current working directory, but does not have to be. It can follow any other path through the file system to
the same directory.
If neither -minimal, nor -sasl is passed as argument you will be prompted.
Test install using EXTRA_PREFIX. The content of the EXTRA_PREFIX variable will prefix all installation paths
when doing make install. Note that EXTRA_PREFIX is similar to DESTDIR, but it does not have the same
effect as DESTDIR. The installation can and have to be run from the location specified by EXTRA_PREFIX.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 13
1.2 Building and Installing Erlang/OTP
That is, it can be useful if you want to try the system out, running test suites, etc, before doing the real install
without EXTRA_PREFIX.
Symbolic Links in --bindir
When doing make install and the default installation prefix is used, relative symbolic links will be created from
/usr/local/bin to all public Erlang/OTP executables in /usr/local/lib/erlang/bin. The installation
phase will try to create relative symbolic links as long as --bindir and the Erlang bin directory, located under -libdir, both have --exec-prefix as prefix. Where --exec-prefix defaults to --prefix. --prefix,
--exec-prefix, --bindir, and --libdir are all arguments that can be passed to configure. One can force
relative, or absolute links by passing BINDIR_SYMLINKS=relative|absolute as arguments to make during
the install phase. Note that such a request might cause a failure if the request cannot be satisfied.
Running
Using HiPE
HiPE supports the following system configurations:
•
x86: All 32-bit and 64-bit mode processors should work.
•
•
•
Linux: Fedora Core is supported. Both 32-bit and 64-bit modes are supported.
NPTL glibc is strongly preferred, or a LinuxThreads glibc configured for "floating stacks". Old non-floating
stacks glibcs have a fundamental problem that makes HiPE support and threads support mutually exclusive.
• Solaris: Solaris 10 (32-bit and 64-bit) and 9 (32-bit) are supported. The build requires a version of the GNU C
compiler (gcc) that has been configured to use the GNU assembler (gas). Sun's x86 assembler is emphatically
not supported.
• FreeBSD: FreeBSD 6.1 and 6.2 in 32-bit and 64-bit modes should work.
• OS X/Darwin: Darwin 9.8.0 in 32-bit mode should work.
PowerPC: All 32-bit 6xx/7xx(G3)/74xx(G4) processors should work. 32-bit mode on 970 (G5) and POWER5
processors should work.
• Linux (Yellow Dog) and OS X 10.4 are supported.
SPARC: All UltraSPARC processors running 32-bit user code should work.
•
•
Solaris 9 is supported. The build requires a gcc that has been configured to use Sun's assembler and linker.
Using the GNU assembler but Sun's linker has been known to cause problems.
• Linux (Aurora) is supported.
ARM: ARMv5TE (i.e. XScale) processors should work. Both big-endian and little-endian modes are supported.
•
Linux is supported.
HiPE is automatically enabled on the following systems:
•
•
•
•
•
x86 in 32-bit mode: Linux, Solaris, FreeBSD
x86 in 64-bit mode: Linux, Solaris, FreeBSD
PowerPC: Linux, Mac OSX
SPARC: Linux
ARM: Linux
On other supported systems, see Advanced Configure on how to enable HiPE.
If you are running on a platform supporting HiPE and if you have not disabled HiPE, you can compile a module into
native code like this from the Erlang shell:
1> c(Module, native).
14 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.3 Cross Compiling Erlang/OTP
or
1> c(Module, [native|OtherOptions]).
Using the erlc program, write like this
$ erlc +native Module.erl
The native code will be placed into the beam file and automatically loaded when the beam file is loaded.
To add hipe options, write like this from the Erlang shell:
1> c(Module, [native,{hipe,HipeOptions}|MoreOptions]).
Use hipe:help_options/0 to print out the available options.
1> hipe:help_options().
1.3 Cross Compiling Erlang/OTP
Table of Contents
•
•
•
•
•
Introduction
• otp_build Versus configure/make
• Cross Configuration
• What can be Cross Compiled?
• Compatibility
• Patches
Build and Install Procedure
• Building With configure/make Directly
• Building a Bootstrap System
• Cross Building the System
• Installing
• Installing Using Paths Determined by configure
• Installing Manually
• Building With the otp_build Script
Building and Installing the Documentation
Testing the cross compiled system
Currently Used Configuration Variables
• Variables for otp_build Only
• Cross Compiler and Other Tools
• Dynamic Erlang Driver Linking
• Large File Support
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 15
1.3 Cross Compiling Erlang/OTP
•
•
• Other Tools
Cross System Root Locations
Optional Feature, and Bug Tests
1.3.1 Introduction
This document describes how to cross compile Erlang/OTP-20. You are advised to read the whole document before
attempting to cross compile Erlang/OTP. However, before reading this document, you should read the $ERL_TOP/
HOWTO/INSTALL.md document which describes building and installing Erlang/OTP in general. $ERL_TOP is the
top directory in the source tree.
otp_build Versus configure/make
Building Erlang/OTP can be done either by using the $ERL_TOP/otp_build script, or by invoking $ERL_TOP/
configure and make directly. Building using otp_build is easier since it involves fewer steps, but the
otp_build build procedure is not as flexible as the configure/make build procedure. Note that otp_build
configure will produce a default configuration that differs from what configure will produce by default. For
example, currently --disable-dynamic-ssl-lib is added to the configure command line arguments unless
--enable-dynamic-ssl-lib has been explicitly passed. The binary releases that we deliver are built using
otp_build. The defaults used by otp_build configure may change at any time without prior notice.
Cross Configuration
The $ERL_TOP/xcomp/erl-xcomp.conf.template file contains all available cross configuration variables
and can be used as a template when creating a cross compilation configuration. All cross configuration
variables are also listed at the end of this document. For examples of working cross configurations see the
$ERL_TOP/xcomp/erl-xcomp-TileraMDE2.0-tilepro.conf file and the $ERL_TOP/xcomp/erlxcomp-x86_64-saf-linux-gnu.conf file. If the default behavior of a variable is satisfactory, the variable
does not need to be set. However, the configure script will issue a warning when a default value is used. When
a variable has been set, no warning will be issued.
A cross configuration file can be passed to otp_build configure using the --xcomp-conf command line
argument. Note that configure does not accept this command line argument. When using the configure script
directly, pass the configuration variables as arguments to configure using a <VARIABLE>=<VALUE> syntax.
Variables can also be passed as environment variables to configure. However, if you pass the configuration in the
environment, make sure to unset all of these environment variables before invoking make; otherwise, the environment
variables might set make variables in some applications, or parts of some applications, and you may end up with an
erroneously configured build.
What can be Cross Compiled?
All Erlang/OTP applications except the wx application can be cross compiled. The build of the wx driver will currently
be automatically disabled when cross compiling.
Compatibility
The build system, including cross compilation configuration variables used, may be subject to non backward
compatible changes without prior notice. Current cross build system has been tested when cross compiling some
Linux/GNU systems, but has only been partly tested for more esoteric platforms. The VxWorks example file is highly
dependent on our environment and is here more or less only for internal use.
Patches
Please submit any patches for cross compiling in a way consistent with this system. All input is welcome as we have
a very limited set of cross compiling environments to test with. If a new configuration variable is needed, add it to
16 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.3 Cross Compiling Erlang/OTP
$ERL_TOP/xcomp/erl-xcomp.conf.template, and use it in configure.in. Other files that might need
to be updated are:
•
•
•
•
•
$ERL_TOP/xcomp/erl-xcomp-vars.sh
$ERL_TOP/erl-build-tool-vars.sh
$ERL_TOP/erts/aclocal.m4
$ERL_TOP/xcomp/README.md
$ERL_TOP/xcomp/erl-xcomp-*.conf
Note that this might be an incomplete list of files that need to be updated.
General information on how to submit patches can be found at: http://wiki.github.com/erlang/otp/submittingpatches
1.3.2 Build and Install Procedure
If you are building in Git, you want to read the Building in Git section of $ERL_TOP/HOWTO/INSTALL.md before
proceeding.
We will first go through the configure/make build procedure which people probably are most familiar with.
Building With configure/make Directly
(1)
Change directory into the top directory of the Erlang/OTP source tree.
$ cd $ERL_TOP
In order to compile Erlang code, a small Erlang bootstrap system has to be built, or an Erlang/OTP system of the same
release as the one being built has to be provided in the $PATH. The Erlang/OTP for the target system will be built
using this Erlang system, together with the cross compilation tools provided.
If you want to build using a compatible Erlang/OTP system in the $PATH, jump to (3).
Building a Bootstrap System
(2)
$ ./configure --enable-bootstrap-only
$ make
The --enable-bootstrap-only argument to configure isn't strictly necessary, but will speed things up. It
will only run configure in applications necessary for the bootstrap, and will disable a lot of things not needed by
the bootstrap system. If you run configure without --enable-boostrap-only you also have to run make as
make bootstrap; otherwise, the whole system will be built.
Cross Building the System
(3)
$ ./configure --host=<HOST> --build=<BUILD> [Other Config Args]
$ make
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 17
1.3 Cross Compiling Erlang/OTP
<HOST> is the host/target system that you build for. It does not have to be a full CPU-VENDOR-OS triplet, but can
be. The full CPU-VENDOR-OS triplet will be created by executing $ERL_TOP/erts/autoconf/config.sub
<HOST>. If config.sub fails, you need to be more specific.
<BUILD> should equal the CPU-VENDOR-OS triplet of the system that you build on. If you execute $ERL_TOP/
erts/autoconf/config.guess, it will in most cases print the triplet you want to use for this.
Pass the cross compilation variables as command line arguments to configure using a <VARIABLE>=<VALUE>
syntax.
Note:
You can not pass a configuration file using the --xcomp-conf argument when you invoke configure
directly. The --xcomp-conf argument can only be passed to otp_build configure.
make will verify that the Erlang/OTP system used when building is of the same release as the system being
built, and will fail if this is not the case. It is possible, however not recommended, to force the cross
compilation even though the wrong Erlang/OTP system is used. This by invoking make like this: make
ERL_XCOMP_FORCE_DIFFERENT_OTP=yes.
Warning:
Invoking make ERL_XCOMP_FORCE_DIFFERENT_OTP=yes might fail, silently produce suboptimal code,
or silently produce erroneous code.
Installing
You can either install using the installation paths determined by configure (4), or install manually using (5).
Installing Using Paths Determined by configure
(4)
$ make install DESTDIR=<TEMPORARY_PREFIX>
make install will install at a location specified when doing configure. configure arguments specifying
where the installation should reside are for example: --prefix, --exec-prefix, --libdir, --bindir, etc.
By default it will install under /usr/local. You typically do not want to install your cross build under /usr/
local on your build machine. Using DESTDIR will cause the installation paths to be prefixed by $DESTDIR. This
makes it possible to install and package the installation on the build machine without having to place the installation
in the same directory on the build machine as it should be executed from on the target machine.
When make install has finished, change directory into $DESTDIR, package the system, move it to the target
machine, and unpack it. Note that the installation will only be working on the target machine at the location determined
by configure.
Installing Manually
(5)
18 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.3 Cross Compiling Erlang/OTP
$ make release RELEASE_ROOT=<RELEASE_DIR>
make release will copy what you have built for the target machine to <RELEASE_DIR>. The Install script
will not be run. The content of <RELEASE_DIR> is what by default ends up in /usr/local/lib/erlang.
The Install script used when installing Erlang/OTP requires common Unix tools such as sed to be present in your
$PATH. If your target system does not have such tools, you need to run the Install script on your build machine
before packaging Erlang/OTP. The Install script should currently be invoked as follows in the directory where
it resides (the top directory):
$ ./Install [-cross] [-minimal|-sasl] <ERL_ROOT>
where:
•
-minimal Creates an installation that starts up a minimal amount of applications, i.e., only kernel and
stdlib are started. The minimal system is normally enough, and is what make install uses.
-sasl Creates an installation that also starts up the sasl application.
-cross For cross compilation. Informs the install script that it is run on the build machine.
<ERL_ROOT> - The absolute path to the Erlang installation to use at run time. This is often the same as the
current working directory, but does not have to be. It can follow any other path through the file system to the
same directory.
•
•
•
If neither -minimal, nor -sasl is passed as argument you will be prompted.
You can now either do:
(6)
•
Decide where the installation should be located on the target machine, run the Install script on the build
machine, and package the installed installation. The installation just need to be unpacked at the right location on
the target machine:
$ cd <RELEASE_DIR>
$ ./Install -cross [-minimal|-sasl] <ABSOLUTE_INSTALL_DIR_ON_TARGET>
or:
(7)
•
Package the installation in <RELEASE_DIR>, place it wherever you want on your target machine, and run the
Install script on your target machine:
$ cd <ABSOLUTE_INSTALL_DIR_ON_TARGET>
$ ./Install [-minimal|-sasl] <ABSOLUTE_INSTALL_DIR_ON_TARGET>
Building With the otp_build Script
(8)
$ cd $ERL_TOP
(9)
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 19
1.3 Cross Compiling Erlang/OTP
$ ./otp_build configure --xcomp-conf=<FILE> [Other Config Args]
alternatively:
$ ./otp_build configure --host=<HOST> --build=<BUILD> [Other Config Args]
If you have your cross compilation configuration in a file, pass it using the --xcomp-conf=<FILE> command
line argument. If not, pass --host=<HOST>, --build=<BUILD>, and the configuration variables using a
<VARIABLE>=<VALUE> syntax on the command line (same as in (3)). Note that <HOST> and <BUILD> have to be
passed one way or the other; either by using erl_xcomp_host=<HOST> and erl_xcomp_build=<BUILD>
in the configuration file, or by using the --host=<HOST>, and --build=<BUILD> command line arguments.
otp_build configure will configure both for the boostrap system on the build machine and the cross host system.
(10)
$ ./otp_build boot -a
otp_build boot -a will first build a bootstrap system for the build machine and then do the cross build of the
system.
(11)
$ ./otp_build release -a <RELEASE_DIR>
otp_build release -a will do the same as (5), and you will after this have to do a manual install either by
doing (6), or (7).
1.3.3 Building and Installing the Documentation
After the system has been cross built you can build and install the documentation the same way as after a native build
of the system. See the How to Build the Documentation section in the $ERL_TOP/HOWTO/INSTALL.md document
for information on how to build the documentation.
1.3.4 Testing the cross compiled system
Some of the tests that come with erlang use native code to test. This means that when cross compiling erlang you
also have to cross compile test suites in order to run tests on the target host. To do this you first have to release the
tests as usual.
$ make release_tests
or
$ ./otp_build tests
20 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.3 Cross Compiling Erlang/OTP
The tests will be released into $ERL_TOP/release/tests. After releasing the tests you have to install the tests
on the build machine. You supply the same xcomp file as to ./otp_build in (9).
$ cd $ERL_TOP/release/tests/test_server/
$ $ERL_TOP/bootstrap/bin/erl -eval 'ts:install([{xcomp,"<FILE>"}])' -s ts compile_testcases -s init stop
You should get a lot of printouts as the testcases are compiled. Once done you should copy the entire $ERL_TOP/
release/tests folder to the cross host system.
Then go to the cross host system and setup the erlang installed in (4) or (5) to be in your $PATH. Then go to what
previously was $ERL_TOP/release/tests/test_server and issue the following command.
$ erl -s ts install -s ts run all_tests -s init stop
The configure should be skipped and all tests should hopefully pass. For more details about how to use ts run erl
-s ts help -s init stop
1.3.5 Currently Used Configuration Variables
Note that you cannot define arbitrary variables in a cross compilation configuration file. Only the ones listed below
will be guaranteed to be visible throughout the whole execution of all configure scripts. Other variables needs to
be defined as arguments to configure or exported in the environment.
Variables for otp_build Only
Variables in this section are only used, when configuring Erlang/OTP for cross compilation using $ERL_TOP/
otp_build configure.
Note:
These variables currently have no effect if you configure using the configure script directly.
•
•
•
erl_xcomp_build - The build system used. This value will be passed as --build=$erl_xcomp_build
argument to the configure script. It does not have to be a full CPU-VENDOR-OS triplet, but can
be. The full CPU-VENDOR-OS triplet will be created by $ERL_TOP/erts/autoconf/config.sub
$erl_xcomp_build. If set to guess, the build system will be guessed using $ERL_TOP/erts/
autoconf/config.guess.
erl_xcomp_host - Cross host/target system to build for. This value will be passed as --host=
$erl_xcomp_host argument to the configure script. It does not have to be a full CPU-VENDOR-OS
triplet, but can be. The full CPU-VENDOR-OS triplet will be created by $ERL_TOP/erts/autoconf/
config.sub $erl_xcomp_host.
erl_xcomp_configure_flags - Extra configure flags to pass to the configure script.
Cross Compiler and Other Tools
If the cross compilation tools are prefixed by <HOST>- you probably do not need to set these variables (where <HOST>
is what has been passed as --host=<HOST> argument to configure).
All variables in this section can also be used when native compiling.
•
CC - C compiler.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 21
1.3 Cross Compiling Erlang/OTP
•
•
•
•
•
•
•
•
•
•
CFLAGS - C compiler flags.
STATIC_CFLAGS - Static C compiler flags.
CFLAG_RUNTIME_LIBRARY_PATH - This flag should set runtime library search path for the shared libraries.
Note that this actually is a linker flag, but it needs to be passed via the compiler.
CPP - C pre-processor.
CPPFLAGS - C pre-processor flags.
CXX - C++ compiler.
CXXFLAGS - C++ compiler flags.
LD - Linker.
LDFLAGS - Linker flags.
LIBS - Libraries.
Dynamic Erlang Driver Linking
Note:
Either set all or none of the DED_LD* variables.
•
•
•
DED_LD - Linker for Dynamically loaded Erlang Drivers.
DED_LDFLAGS - Linker flags to use with DED_LD.
DED_LD_FLAG_RUNTIME_LIBRARY_PATH - This flag should set runtime library search path for shared
libraries when linking with DED_LD.
Large File Support
Note:
Either set all or none of the LFS_* variables.
•
•
•
LFS_CFLAGS - Large file support C compiler flags.
LFS_LDFLAGS - Large file support linker flags.
LFS_LIBS - Large file support libraries.
Other Tools
• RANLIB - ranlib archive index tool.
• AR - ar archiving tool.
• GETCONF - getconf system configuration inspection tool. getconf is currently used for finding out large file
support flags to use, and on Linux systems for finding out if we have an NPTL thread library or not.
Cross System Root Locations
•
•
erl_xcomp_sysroot - The absolute path to the system root of the cross compilation environment. Currently,
the crypto, odbc, ssh and ssl applications need the system root. These applications will be skipped if the
system root has not been set. The system root might be needed for other things too. If this is the case and the
system root has not been set, configure will fail and request you to set it.
erl_xcomp_isysroot - The absolute path to the system root for includes of the cross compilation
environment. If not set, this value defaults to $erl_xcomp_sysroot, i.e., only set this value if the include
system root path is not the same as the system root path.
22 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.3 Cross Compiling Erlang/OTP
Optional Feature, and Bug Tests
These tests cannot (always) be done automatically when cross compiling. You usually do not need to set these variables.
Warning:
Setting these variables wrong may cause hard to detect runtime errors. If you need to change these values, really
make sure that the values are correct.
Note:
Some of these values will override results of tests performed by configure, and some will not be used until
configure is sure that it cannot figure the result out.
The configure script will issue a warning when a default value is used. When a variable has been set, no warning
will be issued.
•
•
•
•
•
•
•
•
•
•
•
erl_xcomp_after_morecore_hook - yes|no. Defaults to no. If yes, the target system must have a
working __after_morecore_hook that can be used for tracking used malloc() implementations core
memory usage. This is currently only used by unsupported features.
erl_xcomp_bigendian - yes|no. No default. If yes, the target system must be big endian. If no, little
endian. This can often be automatically detected, but not always. If not automatically detected, configure will
fail unless this variable is set. Since no default value is used, configure will try to figure this out automatically.
erl_xcomp_double_middle - yes|no. Defaults to no. If yes, the target system must have doubles in
"middle-endian" format. If no, it has "regular" endianness.
erl_xcomp_clock_gettime_cpu_time - yes|no. Defaults to no. If yes, the target system must have
a working clock_gettime() implementation that can be used for retrieving process CPU time.
erl_xcomp_getaddrinfo - yes|no. Defaults to no. If yes, the target system must have a working
getaddrinfo() implementation that can handle both IPv4 and IPv6.
erl_xcomp_gethrvtime_procfs_ioctl - yes|no. Defaults to no. If yes, the target system must have
a working gethrvtime() implementation and is used with procfs ioctl().
erl_xcomp_dlsym_brk_wrappers - yes|no. Defaults to no. If yes, the target system must have a
working dlsym(RTLD_NEXT, <S>) implementation that can be used on brk and sbrk symbols used by the
malloc() implementation in use, and by this track the malloc() implementations core memory usage. This
is currently only used by unsupported features.
erl_xcomp_kqueue - yes|no. Defaults to no. If yes, the target system must have a working kqueue()
implementation that returns a file descriptor which can be used by poll() and/or select(). If no and the
target system has not got epoll() or /dev/poll, the kernel-poll feature will be disabled.
erl_xcomp_linux_clock_gettime_correction - yes|no. Defaults to yes on Linux; otherwise,
no. If yes, clock_gettime(CLOCK_MONOTONIC, _) on the target system must work. This variable is
recommended to be set to no on Linux systems with kernel versions less than 2.6.
erl_xcomp_linux_nptl - yes|no. Defaults to yes on Linux; otherwise, no. If yes, the target system
must have NPTL (Native POSIX Thread Library). Older Linux systems have LinuxThreads instead of NPTL
(Linux kernel versions typically less than 2.6).
erl_xcomp_linux_usable_sigaltstack - yes|no. Defaults to yes on Linux; otherwise, no. If yes,
sigaltstack() must be usable on the target system. sigaltstack() on Linux kernel versions less than
2.4 are broken.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 23
1.4 How to Build Erlang/OTP on Windows
•
•
•
•
•
erl_xcomp_linux_usable_sigusrx - yes|no. Defaults to yes. If yes, the SIGUSR1 and SIGUSR2
signals must be usable by the ERTS. Old LinuxThreads thread libraries (Linux kernel versions typically less than
2.2) used these signals and made them unusable by the ERTS.
erl_xcomp_poll - yes|no. Defaults to no on Darwin/MacOSX; otherwise, yes. If yes, the target system
must have a working poll() implementation that also can handle devices. If no, select() will be used
instead of poll().
erl_xcomp_putenv_copy - yes|no. Defaults to no. If yes, the target system must have a putenv()
implementation that stores a copy of the key/value pair.
erl_xcomp_reliable_fpe - yes|no. Defaults to no. If yes, the target system must have reliable floating
point exceptions.
erl_xcomp_posix_memalign - yes|no. Defaults to yes if posix_memalign system call exists;
otherwise no. If yes, the target system must have a posix_memalign implementation that accepts larger than
page size alignment.
1.4 How to Build Erlang/OTP on Windows
Table of Contents
•
•
•
•
•
•
•
•
Introduction
Short Version
Frequently Asked Questions
Tools you Need and Their Environment
The Shell Environment
Building and Installing
Development
Using GIT
1.4.1 Introduction
This section describes how to build the Erlang emulator and the OTP libraries on Windows. Note that the Windows
binary releases are still a preferred alternative if one does not have Microsoft’s development tools and/or don’t want
to install Cygwin, MSYS or MSYS2.
The instructions apply to versions of Windows supporting the Cygwin emulated gnuish environment or the MSYS or
MSYS2 ditto. We’ve built on the following platforms: Windows 2012, Windows 7, Windows 8 and Windows 10. It’s
probably possible to build on older platforms too, but you might not be able to install the appropriate Microsoft SDK,
Visual Studio or OpenSSL, in which case you will need to go back to earlier compilers etc.
The procedure described uses either Cygwin, MSYS or MSYS2 as a build environment. You run the bash shell in
Cygwin/MSYS/MSYS2 and use the gnu make/configure/autoconf etc to do the build. The emulator C-source code
is, however, mostly compiled with Microsoft Visual C++™, producing a native Windows binary. This is the same
procedure as we use to build the pre-built binaries. Why we use VC++ and not gcc is explained further in the FAQ
section.
If you are not familiar with Cygwin, MSYS, MSYS2 or a Unix environment, you’ll probably need to read up a bit on
how that works. There are plenty of documentation about this online.
These instructions apply for both 32-bit and 64-bit Windows. Note that even if you build a 64-bit version of Erlang,
most of the directories and files involved are still named win32. Some occurances of the name win64 are however
present. The installation file for a 64-bit Windows version of Erlang, for example, is otp_win64_20.exe.
If you feel comfortable with the environment and build system, and have all the necessary tools, you have a great
opportunity to make the Erlang/OTP distribution for Windows better. Please submit any suggestions to our JIRA
24 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.4 How to Build Erlang/OTP on Windows
and patches to our git project to let them find their way into the next version of Erlang. If making changes to
the build system (like makefiles etc) please bear in mind that the same makefiles are used on Unix/VxWorks, so
that your changes don't break other platforms. That of course goes for C-code too; system specific code resides
in the $ERL_TOP/erts/emulator/sys/win32 and $ERL_TOP/erts/etc/win32 directories mostly. The
$ERL_TOP/erts/emulator/beam directory is for common code.
We've used this build procedure for a couple of releases, and it has worked fine for us. Still, there might be all sorts of
troubles on different machines and with different setups. We'll try to give hints wherever we've encountered difficulties,
but please share your experiences by using the erlang-questions mailing list. We cannot, of course, help everyone
with all their issues, so please try to solve such issues and submit solutions/workarounds.
Lets go then! We’ll start with a short version of the setup procedure, followed by some FAQ, and then we’ll go into
more details of the setup.
1.4.2 Short Version
In the following sections, we've described as much as we could about the installation of the tools needed. Once the
tools are installed, building is quite easy. We have also tried to make these instructions understandable for people with
limited Unix experience. Cygwin/MSYS/MSYS2 is a whole new environment to some Windows users, why careful
explanation of environment variables etc seemed to be in place.
This is the short story though, for the experienced and impatient:
•
Get and install complete Cygwin (latest), complete MinGW with MSYS or complete MSYS2
•
•
•
•
•
•
•
•
•
Install Visual Studio 12.0 (2013)
Install Microsofts Windows SDK 8.1
Get and install Sun's JDK 1.6.0 or later
Get and install NSIS 2.01 or later (up to 2.46 tried and working)
Get, build and install OpenSSL 0.9.8r or later (up to 1.0.2d tried & working) with static libs.
Get the Erlang source distribution (from http://www.erlang.org/download.html) and unpack with Cygwin's/
MSYS's/MSYS2's tar.
Set ERL_TOP to where you unpacked the source distribution
$ cd $ERL_TOP
Modify PATH and other environment variables so that all these tools are runnable from a bash shell. Still
standing in $ERL_TOP, issue the following commands (for 32-bit Windows, remove the x64 from the first
row and change otp_win64_20 to otp_win32_20 on the last row):
$
$
$
$
$
$
$
eval `./otp_build env_win32 x64`
./otp_build autoconf
./otp_build configure
./otp_build boot -a
./otp_build release -a
./otp_build installer_win32
release/win32/otp_win64_20 /S
Voila! Start->Programs->Erlang OTP 20->Erlang starts the Erlang Windows shell.
1.4.3 Frequently Asked Questions
•
Q: So, now I can build Erlang using GCC on Windows?
A: No, unfortunately not. You'll need Microsoft's Visual C++ still. A Bourne-shell script (cc.sh) wraps the Visual
C++ compiler and runs it from within the Cygwin environment. All other tools needed to build Erlang are free-
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 25
1.4 How to Build Erlang/OTP on Windows
•
•
ware/open source, but not the C compiler. The Windows SDK is however enough to build Erlang, you do not
need to buy Visual C++, just download the SDK (SDK version 8.1 == Visual studio 2013).
Q: Why haven't you got rid of VC++ then, you ******?
A: Well, partly because it's a good compiler - really! Actually it's been possible in late R11-releases to build using
mingw instead of visual C++ (you might see the remnants of that in some scripts and directories). Unfortunately
the development of the SMP version for Windows broke the mingw build and we chose to focus on the VC++
build as the performance has been much better in the VC++ versions. The mingw build will possibly be back, but
as long as VC++ gives better performance, the commercial build will be a VC++ one.
Q: OK, you need VC++, but now you've started to demand a quite recent (and expensive) version of Visual Studio.
Why?
•
A: Well, it's not expensive, it's free (as in free beer). Just download and install the latest Windows SDK from
Microsoft and all the tools you need are there. The included debugger (WinDbg) is also quite usable. That's
what I used when porting Erlang to 64bit Windows. Another reason to use later Microsoft compilers is DLL
compatibility. DLL's using a new version of the standard library might not load if the VM is compiled with an old
VC++ version. So we should aim to use the latest freely available SDK and compiler.
Q: Can/will I build a Cygwin binary with the procedure you describe?
•
A: No, the result will be a pure Windows binary, and as far as I know, it's not possible to make a Cygwin binary
yet. That is of course something desirable, but there are still some problems with the dynamic linking (dynamic
Erlang driver loading) as well as the TCP/IP emulation in Cygwin, which, I'm sure of, will improve, but still has
some problems. Fixing those problems might be easy or might be hard. I suggest you try yourself and share your
experience. No one would be happier if a simple ./configure && make would produce a fully fledged
Cygwin binary.
Q: Hah, I saw you, you used GCC even though you said you didn't!
•
A: OK, I admit, one of the files is compiled using Cygwin's or MinGW's GCC and the resulting object
code is then converted to MS VC++ compatible coff using a small C hack. It's because that particular file,
beam_emu.c benefits immensely from being able to use the GCC labels-as-values extension, which boosts
emulator performance by up to 50%. That does unfortunately not (yet) mean that all of OTP could be compiled
using GCC. That particular source code does not do anything system specific and actually is adopted to the fact
that GCC is used to compile it on Windows.
Q: So now there's a MS VC++ project file somewhere and I can build OTP using the nifty VC++ GUI?
•
A: No, never. The hassle of keeping the project files up to date and do all the steps that constitute an OTP build
from within the VC++ GUI is simply not worth it, maybe even impossible. A VC++ project file for Erlang/OTP
will never happen.
Q: So how does it all work then?
A: Cygwin, MSYS or MSYS2 is the environment, which closely resembles the environment found on any Unix
machine. It's almost like you had a virtual Unix machine inside Windows. Configure, given certain parameters,
then creates makefiles that are used by the environment's gnu-make to built the system. Most of the actual
compilers etc are not, however, Cygwin/MSYS/MSYS2 tools, so we've written a couple of wrappers (Bourneshell scripts), which reside in $ERL_TOP/etc/win32/cygwin_tools and $ERL_TOP/etc/win32/
msys_tools. They all do conversion of parameters and switches common in the Unix environment to fit the
native Windows tools. Most notable is of course the paths, which in Cygwin/MSYS/MSYS2 are Unix-like paths
with "forward slashes" (/) and no drive letters. The Cygwin specific command cygpath is used for most of the
path conversions in a Cygwin environment. Other tools are used (when needed) in the corresponding MSYS and
MSYS2 environment. Luckily most compilers accept forward slashes instead of backslashes as path separators,
but one still have to get the drive letters etc right, though. The wrapper scripts are not general in the sense that, for
example, cc.sh would understand and translate every possible gcc option and pass correct options to cl.exe. The
principle is that the scripts are powerful enough to allow building of Erlang/OTP, no more, no less. They might
26 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.4 How to Build Erlang/OTP on Windows
need extensions to cope with changes during the development of Erlang, and that's one of the reasons we made
them into shell-scripts and not Perl-scripts. We believe they are easier to understand and change that way.
•
In $ERL_TOP, there is a script called otp_build. That script handles the hassle of giving all the right parameters
to configure/make and also helps you set up the correct environment variables to work with the Erlang source
under Cygwin/MSYS/MSYS2.
Q: You use and need Cygwin, but then you haven't taken the time to port Erlang to the Cygwin environment but
instead focus on your commercial release, is that really ethical?
•
A: No, not really, but see this as a step in the right direction.
Q: Can I build something that looks exactly as the commercial release?
•
A: Yes, we use the exact same build procedure.
Q: Which version of Cygwin/MSYS/MSYS2 and other tools do you use then?
•
A: For Cygwin, MSYS and MSYS2 alike, we try to use the latest releases available when building. What versions
you use shouldn't really matter. We try to include workarounds for the bugs we've found in different Cygwin/
MSYS/MSYS2 releases. Please help us add workarounds for new Cygwin/MSYS/MSYS2-related bugs as soon
as you encounter them. Also please do submit bug reports to the appropriate Cygwin, MSYS and/or MSYS2
developers. The GCC we used for 20 was version 4.8.1 (MinGW 32bit) and 4.8.5 (MSYS2 64bit). We used VC++
12.0 (i.e. Visual studio 2013), Sun's JDK 1.6.0_45 (32bit) and Sun's JDK 1.7.0_1 (64bit), NSIS 2.46, and Win32
OpenSSL 1.0.2d. Please read the next section for details on what you need.
Q: Can you help me setup X in Cygwin/MSYS/MSYS2?
A: No, unfortunately we haven't got time to help with Cygwin/MSYS/MSYS2 related user problems, please read
related websites, newsgroups and mailing lists.
1.4.4 Tools you Need and Their Environment
You need some tools to be able to build Erlang/OTP on Windows. Most notably you'll need Cygwin, MSYS or MSYS2,
Visual Studio and Microsofts Windows SDK, but you might also want a Java compiler, the NSIS install system and
OpenSSL. Well, here's some information about the different tools:
•
Cygwin, the very latest is usually best. Get all the development tools and of course all the basic ditto. Make sure
to get jar and also make sure not to install a Cygwin'ish Java, since the Cygwin jar command is used but Sun's
Java compiler and virtual machine.
If you are going to build a 64bit Windows version, you should make sure to get MinGW's 64bit gcc installed with
Cygwin. It's in one of the development packages.
URL: http://www.cygwin.com
Get the installer from the website and use it to install Cygwin. Be sure to have fair privileges. If you're on an NT
domain you should consider running mkpasswd -d and mkgroup -d after the installation to get the user
databases correct. See their respective manual pages.
When you start your first bash shell, you will get an awful prompt. You might also have a PATH environment
variable that contains backslashes and such. Edit $HOME/.profile and $HOME/.bashrc to set fair prompts
and a correct PATH. Also do an export SHELL in .profile. For some non-obvious reason the environment
variable $SHELL is not exported in bash. Also note that .profile is run at login time and .bashrc when
sub shells are created. You'll need to explicitly source .bashrc from .profile if you want the commands
there to be run at login time (like setting up aliases, shell functions and the like). You can for example do like
this at the end of .profile:
ENV=$HOME/.bashrc
export ENV
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 27
1.4 How to Build Erlang/OTP on Windows
. $ENV
You might also want to setup X-windows (XFree86). That might be as easy as running startx from the command
prompt and it might be much harder. Use Google to find help.
If you don't use X-windows, you might want to setup the Windows console window by selecting properties in
the console system menu (upper left corner of the window, the Cygwin icon in the title bar). Especially setting
a larger screen buffer size (lines) is useful as it gets you a scrollbar so you can see whatever error messages that
might appear.
•
There are a few other shells available, but in all examples below we assume that you use bash.
Alternatively you download MinGW and MSYS. You'll find the latest installer at:
URL: http://sourceforge.net/projects/mingw/files/Installer/mingw-get-inst/
Make sure to install the basic dev tools, but avoid the MinGW autoconf and install the msys one instead.
To be able to build the 64bit VM, you will also need the 64bit MinGW compiler from:
URL: http://sourceforge.net/projects/mingw-w64/files/latest/download?source=files
•
We've tried up to 1.0, but the latest version should do. Make sure you download the mingw-w64-bin_i686mingw_<something>.zip, not a linux version. You unzip the package on top of your MinGW installation
(c:\MinGW) and that's it.
A third alternative is to download and install MSYS2 from:
URL: https://msys2.github.io/
When you've followed the instructions there, you also need to install these packages: autoconf, make, perl, and
tar. You do so by running the following in the msys console:
pacman -S msys/autoconf msys/make msys/perl msys/tar
You also need a gcc. If you installed the 64 bit MSYS2 you run:
mingw64/mingw-w64-x86_64-gcc
And for 32 bit MSYS2:
pacman -S mingw32/mingw-w64-i686-gcc
pacman -S mingw-w64-i686-editrights
•
Visual Studio 2013 (Visual Studio 12.0). Download and run the web installer from:
https://www.visualstudio.com/
•
•
Microsofts Windows SDK version 8.1 (corresponding to VC++ 12.0 and Visual Studio 2013). You'll find it here:
URL: https://msdn.microsoft.com/en-us/windows/desktop/bg162891.aspx
To help setup the environment, there is a bat file, %PROGRAMFILES%\Mirosoft Visual Studio
12.0\VC\vcvarsall.bat, that set's the appropriate environment for a Windows command prompt.
This is not appropriate for bash, so you'll need to convert it to bash-style environments by editing your
.bash_profile. In my case, where the SDK is installed in the default directory and %PROGRAMFILES% is
28 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.4 How to Build Erlang/OTP on Windows
C:\Program Files, the commands for setting up a 32bit build environment (on a 64bit or 32bit machine)
look like this (in Cygwin):
# Some common paths
C_DRV=/cygdrive/c
PRG_FLS=$C_DRV/Program\ Files
# nsis
NSIS_BIN=$PRG_FLS/NSIS
# java
JAVA_BIN=$PROGRAMFILES/Java/jdk1.7.0_02/bin
##
## MS SDK
##
CYGWIN=nowinsymlinks
VISUAL_STUDIO_ROOT=$PRG_FLS/Microsoft\ Visual\ Studio\ 12.0
WIN_VISUAL_STUDIO_ROOT="C:\\Program Files\\Microsoft Visual Studio 12.0"
SDK=$PRG_FLS/Windows\ Kits/8.1
WIN_SDK="C:\\Program Files\\Windows Kits\\8.1"
PATH="$NSIS_BIN:\
$VISUAL_STUDIO_ROOT/VC/bin:\
$VISUAL_STUDIO_ROOT/VC/vcpackages:\
$VISUAL_STUDIO_ROOT/Common7/IDE:\
$VISUAL_STUDIO_ROOT/Common7/Tools:\
$SDK/bin/x86
/usr/local/bin:/usr/bin:/bin:\
/cygdrive/c/WINDOWS/system32:/cygdrive/c/WINDOWS:\
/cygdrive/c/WINDOWS/system32/Wbem:\
$JAVA_BIN"
LIBPATH="$WIN_VISUAL_STUDIO_ROOT\\VC\\lib"
LIB="$WIN_VISUAL_STUDIO_ROOT\\VC\\lib\\;$WIN_SDK\\lib\\winv6.3\\um\\x86"
INCLUDE="$WIN_VISUAL_STUDIO_ROOT\\VC\\include\\;$WIN_SDK\\include\\shared\\;\
$WIN_SDK\\include\\um;$WIN_SDK\\include\\winrt\\;$WIN_SDK\\include\\um\\gl"
export CYGWIN PATH LIBPATH LIB INCLUDE
If you're using MinGW's MSYS instead, you need to change the C_DRV setting, which would read:
C_DRV=/c
and you also need to change the PATH environment variable to:
MINGW_BIN=/c/MinGW/bin
PATH="$NSIS_BIN:\
$VISUAL_STUDIO_ROOT/VC/bin:\
$VISUAL_STUDIO_ROOT/VC/vcpackages:\
$VISUAL_STUDIO_ROOT/Common7/IDE:\
$VISUAL_STUDIO_ROOT/Common7/Tools:\
$SDK/bin/x86:/usr/local/bin:\
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 29
1.4 How to Build Erlang/OTP on Windows
$MINGW_BIN:\
/bin:/c/Windows/system32:/c/Windows:\
/c/Windows/System32/Wbem:\
$JAVA_BIN"
For MSYS2 you use the same C_DRV and PATH as for MSYS, only update the MINGW_BIN:
MINGW_BIN=/mingw32/bin
If you are building a 64 bit version of Erlang, you should set up PATHs etc a little differently. We have two
templates to make things work in both Cygwin and MSYS but needs editing to work with MSYS2 (see the
comments in the script). The following one is for 32 bits:
make_winpath()
{
P=$1
if [ "$IN_CYGWIN" = "true" ]; then
cygpath -d "$P"
else
(cd "$P" && /bin/cmd //C "for %i in (".") do @echo %~fsi")
fi
}
make_upath()
{
P=$1
if [ "$IN_CYGWIN" = "true" ]; then
cygpath "$P"
else
echo "$P" | /bin/sed 's,^\([a-zA-Z]\):\\,/\L\1/,;s,\\,/,g'
fi
}
# Some common paths
if [ -x /usr/bin/msys-?.0.dll ]; then
# Without this the path conversion won't work
COMSPEC='C:\Windows\System32\cmd.exe'
MSYSTEM=MINGW32 # Comment out this line if in MSYS2
export MSYSTEM COMSPEC
# For MSYS2: Change /mingw/bin to the msys bin dir on the line below
PATH=/usr/local/bin:/mingw/bin:/bin:/c/Windows/system32:\
/c/Windows:/c/Windows/System32/Wbem
C_DRV=/c
IN_CYGWIN=false
else
PATH=/ldisk/overrides:/usr/local/bin:/usr/bin:/bin:\
/usr/X11R6/bin:/cygdrive/c/windows/system32:\
/cygdrive/c/windows:/cygdrive/c/windows/system32/Wbem
C_DRV=/cygdrive/c
IN_CYGWIN=true
fi
obe_otp_gcc_vsn_map="
.*=>default
"
obe_otp_64_gcc_vsn_map="
.*=>default
"
# Program Files
30 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.4 How to Build Erlang/OTP on Windows
PRG_FLS=$C_DRV/Program\ Files
# Visual Studio
VISUAL_STUDIO_ROOT=$PRG_FLS/Microsoft\ Visual\ Studio\ 12.0
WIN_VISUAL_STUDIO_ROOT="C:\\Program Files\\Microsoft Visual Studio 12.0"
# SDK
SDK=$PRG_FLS/Windows\ Kits/8.1
WIN_SDK="C:\\Program Files\\Windows Kits\\8.1"
# NSIS
NSIS_BIN=$PROGRAMFILES/NSIS
# Java
JAVA_BIN=$PROGRAMFILES/Java/jdk1.7.0_02/bin
## The PATH variable should be Cygwin'ish
VCPATH=
$VISUAL_STUDIO_ROOT/VC/bin:\
$VISUAL_STUDIO_ROOT/VC/vcpackages:\
$VISUAL_STUDIO_ROOT/Common7/IDE:\
$VISUAL_STUDIO_ROOT/Common7/Tools:\
$SDK/bin/x86
## Microsoft SDK libs
LIBPATH=$WIN_VISUAL_STUDIO_ROOT\\VC\\lib
LIB=$WIN_VISUAL_STUDIO_ROOT\\VC\\lib\\;$WIN_KITS\\lib\\winv6.3\\um\\x86
INCLUDE=$WIN_VISUAL_STUDIO_ROOT\\VC\\include\\;\
$WIN_KITS\\include\\shared\\;$WIN_KITS\\include\\um;\
$WIN_KITS\\include\\winrt\\;$WIN_KITS\\include\\um\\gl
# Put nsis, c compiler and java in path
export PATH=$VCPATH:$PATH:$JAVA_BIN:$NSIS_BIN
# Make sure LIB and INCLUDE is available for others
export LIBPATH LIB INCLUDE
The first part of the 64 bit template is identical to the 32 bit one, but there are some environment variable
differences:
# Program Files
PRG_FLS64=$C_DRV/Program\ Files
PRG_FLS32=$C_DRV/Program\ Files\ \(x86\)
# Visual Studio
VISUAL_STUDIO_ROOT=$PRG_FLS32/Microsoft\ Visual\ Studio\ 12.0
WIN_VISUAL_STUDIO_ROOT="C:\\Program Files (x86)\\Microsoft Visual Studio 12.0"
# SDK
SDK=$PRG_FLS32/Windows\ Kits/8.1
WIN_SDK="C:\\Program Files (x86)\\Windows Kits\\8.1"
# NSIS
NSIS_BIN=$PROGRAMFILES/NSIS
# Java
JAVA_BIN=$PROGRAMFILES/Java/jdk1.7.0_02/bin
## The PATH variable should be Cygwin'ish
VCPATH=
$VISUAL_STUDIO_ROOT/VC/bin/amd64:\
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 31
1.4 How to Build Erlang/OTP on Windows
$VISUAL_STUDIO_ROOT/VC/vcpackages:\
$VISUAL_STUDIO_ROOT/Common7/IDE:\
$VISUAL_STUDIO_ROOT/Common7/Tools:\
$SDK/bin/x86
## Microsoft SDK libs
LIBPATH=$WIN_VISUAL_STUDIO_ROOT\\VC\\lib\\amd64
LIB=$WIN_VISUAL_STUDIO_ROOT\\VC\\lib\\amd64\\;\
$WIN_KITS\\lib\\winv6.3\\um\\x64
INCLUDE=$WIN_VISUAL_STUDIO_ROOT\\VC\\include\\;\
$WIN_KITS\\include\\shared\\;$WIN_KITS\\include\\um;\
$WIN_KITS\\include\\winrt\\;$WIN_KITS\\include\\um\\gl
# Put nsis, c compiler and java in path
export PATH=$VCPATH:$PATH:$JAVA_BIN:$NSIS_BIN
# Make sure LIB and INCLUDE is available for others
export LIBPATH LIB INCLUDE
Make sure to set the PATH so that NSIS and Microsoft SDK is found before the MSYS/Cygwin tools and that
Java is last in the PATH.
•
Make a simple hello world and try to compile it with the cl command from within bash. If that does not work,
your environment needs fixing. Remember, there should be no backslashes in your path environment variable
in Cygwin bash, but LIB and INCLUDE should contain Windows style paths with semicolon, drive letters and
backslashes.
Sun's Java JDK 1.6.0 or later. Our Java code (jinterface, ic) is written for JDK 1.6.0. Get it for Windows and install
it, the JRE is not enough. If you don't care about Java, you can skip this step. The result will be that jinterface
is not built.
URL: http://java.sun.com
Add javac LAST to your path environment in bash, in my case this means:
`PATH="$PATH:/cygdrive/c/Program Files/Java/jdk1.7.0_02/bin"`
•
No CLASSPATH or anything is needed. Type javac in the bash prompt and you should get a list of available Java
options. Make sure, e.g by typing type java, that you use the Java you installed. Note however that Cygwin's/
MinGW's/MSYS2's jar.exe is used. That's why the JDK bin-directory should be added last in the PATH.
Nullsoft NSIS installer system. You need this to build the self installing package. It's a free open source installer
that's much nicer to use than the commercial Wise and Install shield installers. This is the installer we use for
commercial releases as well.
URL: http://nsis.sourceforge.net/download
Install the lot, especially the modern user interface components, as it's definitely needed. Put makensis in your
path, in my case:
PATH=/cygdrive/c/Program\ Files/NSIS:$PATH
•
Type makensis at the bash prompt and you should get a list of options if everything is OK.
OpenSSL. This is if you want the SSL and crypto applications to compile (and run). There are prebuilt binaries,
which you can just download and install, available here:
32 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.4 How to Build Erlang/OTP on Windows
URL: http://openssl.org/community/binaries.html
•
We would recommend using 1.0.2d.
Building with wxWidgets. Download wxWidgets-3.0.3 or higher.
Install or unpack it to the pgm folder: Cygwin: DRIVE:/PATH/cygwin/opt/local/pgm MSYS:
DRIVE:/PATH/MinGW/msys/1.0/opt/local/pgm MSYS2: DRIVE:/PATH/msys<32/64>/opt/
local/pgm
If the wxUSE_POSTSCRIPT isn't enabled in <path\to\pgm>\wxMSW-3.0.3\include\wx\msw
\setup.h, enable it.
build: From a command prompt with the VC tools available (See the instructions for OpenSSL build above for
help on starting the proper command prompt in RELEASE mode):
C:\...\> cd <path\to\pgm>\wxMSW-3.0.3\build\msw
C:\...\> nmake BUILD=release SHARED=0 DIR_SUFFIX_CPU= -f makefile.vc
Or - if building a 64bit version:
C:\...\> cd <path\to\pgm>\wxMSW-3.0.3\build\msw
C:\...\> nmake TARGET_CPU=amd64 BUILD=release SHARED=0 DIR_SUFFIX_CPU= -f makefile.vc
•
Get the Erlang source distribution (from http://www.erlang.org/download.html). The same as for Unix
platforms. Preferably use tar from within Cygwin, MSYS or MSYS2 to unpack the source tar.gz (tar zxf
otp_src_20.tar.gz).
Set the environment ERL_TOP to point to the root directory of the source distribution. Let's say I stood in $HOME/
src and unpacked otp_src_20.tar.gz, I then add the following to .profile:
ERL_TOP=$HOME/src/otp_src_20
export $ERL_TOP
1.4.5 The Shell Environment
So, if you have followed the instructions above, when you start a bash shell, you should have an INCLUDE
environment with a Windows style path, a LIB environment variable also in Windows style, and finally a PATH that
let's you reach cl, makensis, javac etc from the command prompt (use which cl etc to verify from bash).
You should also have an ERL_TOP environment variable that is Cygwin style, and points to a directory containing,
among other files, the script otp_build.
A final massage of the environment is needed, and that is done by the script $ERL_TOP/otp_build. Start bash
and do the following, note the "back-ticks" (`), can be quite hard to get on some keyboards, but pressing the backtick key followed by the space bar might do it...
$ cd $ERL_TOP
$ eval `./otp_build env_win32`
If you're unable to produce back-ticks on your keyboard, you can use the ksh variant:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 33
1.4 How to Build Erlang/OTP on Windows
$ cd $ERL_TOP
$ eval $(./otp_build env_win32)
If you are building a 64 bit version, you supply otp_build with an architecture parameter:
$ cd $ERL_TOP
$ eval `./otp_build env_win32 x64`
This should do the final touch to the environment and building should be easy after this. You could run ./otp_build
env_win32 without eval just to see what it does, and to see that the environment it sets seems OK. The path is
cleaned of spaces if possible (using DOS style short names instead), the variables OVERRIDE_TARGET, CC, CXX, AR
and RANLIB are set to their respective wrappers and the directories $ERL_TOP/erts/etc/win32/<cygwin/
msys>_tools/vc and $ERL_TOP/erts/etc/win32/<cygwin/msys>_tool are added first in the PATH.
Now you can check which erlc you have by writing type erlc in your shell. It should reside in $ERL_TOP/erts/
etc/win32/cygwin_tools or $ERL_TOP/erts/etc/win32/msys_tools.
1.4.6 Building and Installing
Building is easiest using the otp_build script:
$
$
$
$
$
./otp_build
./otp_build
./otp_build
./otp_build
./otp_build
autoconf # Ignore the warning blob about versions of autoconf
configure <optional configure options>
boot -a
release -a <installation directory>
installer_win32 <installation directory> # optional
Now you will have a file called otp_win32_20.exe or otp_win64_20.exe in the <installation
directory>, i.e. $ERL_TOP/release/win32.
Lets get into more detail:
•
•
•
•
•
$ ./otp_build autoconf - This step rebuilds the configure scripts to work correctly in your environment.
In an ideal world, this would not be needed, but alas, we have encountered several incompatibilities between
our distributed configure scripts (generated on a Linux platform) and the Cygwin/MSYS/MSYS2 environment
over the years. Running autoconf in Cygwin/MSYS/MSYS2 ensures that the configure scripts are generated in a
compatible way and that they will work well in the next step.
$ ./otp_build configure - This runs the newly generated configure scripts with options making configure
behave nicely. The target machine type is plainly win32, so a lot of the configure-scripts recognize this awkward
target name and behave accordingly. The CC variable also makes the compiler be cc.sh, which wraps MSVC+
+, so all configure tests regarding the C compiler gets to run the right compiler. A lot of the tests are not needed
on Windows, but we thought it best to run the whole configure anyway.
$ ./otp_build boot -a - This uses the bootstrap directory (shipped with the source, $ERL_TOP/
bootstrap) to build a complete OTP system. When this is done you can run erl from within the source tree;
just type $ERL_TOP/bin/erl and you whould have the prompt.
$ ./otp_build release -a - Builds a commercial release tree from the source tree. The default is to put it
in $ERL_TOP/release/win32. You can give any directory as parameter (Cygwin style), but it doesn't really
matter if you're going to build a self extracting installer too.
$ ./otp_build installer_win32 - Creates the self extracting installer executable. The executable
otp_win32_20.exe or otp_win64_20.exe will be placed in the top directory of the release created in
the previous step. If no release directory is specified, the release is expected to have been built to $ERL_TOP/
34 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.4 How to Build Erlang/OTP on Windows
release/win32, which also will be the place where the installer executable will be placed. If you specified
some other directory for the release (i.e. ./otp_build release -a /tmp/erl_release), you're
expected to give the same parameter here, (i.e. ./otp_build installer_win32 /tmp/erl_release).
You need to have a full NSIS installation and makensis.exe in your path for this to work. Once you have
created the installer, you can run it to install Erlang/OTP in the regular way, just run the executable and follow
the steps in the installation wizard. To get all default settings in the installation without any questions asked, you
run the executable with the parameter /S (capital S) like in:
$ cd $ERL_TOP
$ release/win32/otp_win32_20 /S
...
or
$ cd $ERL_TOP
$ release/win32/otp_win64_20 /S
...
and after a while Erlang/OTP-20 will have been installed in C:\Program Files\erl9.0\, with shortcuts
in the menu etc.
1.4.7 Development
Once the system is built, you might want to change it. Having a test release in some nice directory might be useful,
but you can also run Erlang from within the source tree. The target local_setup, makes the program $ERL_TOP/
bin/erl.exe usable and it also uses all the OTP libraries in the source tree.
If you hack the emulator, you can build the emulator executable by standing in $ERL_TOP/erts/emulator and
do a simple
$ make opt
Note that you need to have run (cd $ERL_TOP && eval `./otp_build env_win32`) in the particular
shell before building anything on Windows. After doing a make opt you can test your result by running $ERL_TOP/
bin/erl. If you want to copy the result to a release directory (say /tmp/erl_release), you do this (still in
$ERL_TOP/erts/emulator)
$ make TESTROOT=/tmp/erl_release release
That will copy the emulator executables.
To make a debug build of the emulator, you need to recompile both beam.dll (the actual runtime system) and
erlexec.dll. Do like this
$
$
$
$
$
cd $ERL_TOP
rm bin/win32/erlexec.dll
cd erts/emulator
make debug
cd ../etc
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 35
1.4 How to Build Erlang/OTP on Windows
$ make debug
and sometimes
$ cd $ERL_TOP
$ make local_setup
So now when you run $ERL_TOP/erl.exe, you should have a debug compiled emulator, which you will see if
you do a:
1> erlang:system_info(system_version).
in the erlang shell. If the returned string contains [debug], you got a debug compiled emulator.
To hack the erlang libraries, you simply do a make opt in the specific "applications" directory, like:
$ cd $ERL_TOP/lib/stdlib
$ make opt
or even in the source directory...
$ cd $ERL_TOP/lib/stdlib/src
$ make opt
Note that you're expected to have a fresh Erlang in your path when doing this, preferably the plain 20 you have built
in the previous steps. You could also add $ERL_TOP/bootstrap/bin to your PATH before rebuilding specific
libraries. That would give you a good enough Erlang system to compile any OTP erlang code. Setting up the path
correctly is a little bit tricky. You still need to have $ERL_TOP/erts/etc/win32/cygwin_tools/vc and
$ERL_TOP/erts/etc/win32/cygwin_tools before the actual emulator in the path. A typical setting of the
path for using the bootstrap compiler would be:
$ export PATH=$ERL_TOP/erts/etc/win32/cygwin_tools/vc\
:$ERL_TOP/erts/etc/win32/cygwin_tools:$ERL_TOP/bootstrap/bin:$PATH
That should make it possible to rebuild any library without hassle...
If you want to copy a library (an application) newly built, to a release area, you do like with the emulator:
$ cd $ERL_TOP/lib/stdlib
$ make TESTROOT=/tmp/erlang_release release
Remember that:
•
•
Windows specific C-code goes in the $ERL_TOP/erts/emulator/sys/win32, $ERL_TOP/erts/
emulator/drivers/win32 or $ERL_TOP/erts/etc/win32.
Windows specific erlang code should be used conditionally and the host OS tested in runtime, the exactly same
beam files should be distributed for every platform! So write code like:
36 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.5 Patching OTP Applications
case os:type() of
{win32,_} ->
do_windows_specific();
Other ->
do_fallback_or_exit()
end,
That's basically all you need to get going.
1.4.8 Using GIT
You might want to check out versions of the source code from GitHUB. That is possible directly in Cygwin, but not
in MSYS. There is a project MsysGIT:
URL:http://code.google.com/p/msysgit/
that makes a nice Git port. The msys prompt you get from MsysGIT is however not compatible with the full version
from MinGW, so you will need to check out files using MsysGIT's command prompt and then switch to a common
MSYS command prompt for building. Also all test suites cannot be built as MsysGIT/MSYS does not handle symbolic
links.
1.5 Patching OTP Applications
1.5.1 Introduction
This document describes the process of patching an existing OTP installation with one or more Erlang/OTP
applications of newer versions than already installed. The tool otp_patch_apply is available for this specific
purpose. It resides in the top directory of the Erlang/OTP source tree.
The otp_patch_apply tool utilizes the runtime_dependencies tag in the application resource file. This information
is used to determine if the patch can be installed in the given Erlang/OTP installation directory.
Read more about the version handling introduced in Erlang/OTP release 17, which also describes how to determine
if an installation includes one or more patched applications.
If you want to apply patches of multiple OTP applications that resides in different OTP versions, you have to apply
these patches in multiple steps. It is only possible to apply multiple OTP applications from the same OTP version
at once.
1.5.2 Prerequisites
It's assumed that the reader is familiar with building and installing Erlang/OTP. To be able to patch an application,
the following must exist:
•
•
An Erlang/OTP installation.
An Erlang/OTP source tree containing the updated applications that you want to patch into the existing Erlang/
OTP installation.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 37
1.5 Patching OTP Applications
1.5.3 Using otp_patch_apply
Warning:
Patching applications is a one-way process. Create a backup of your OTP installation directory before proceeding.
First of all, build the OTP source tree at $ERL_TOP containing the updated applications.
Note:
Before applying a patch you need to do a full build of OTP in the source directory.
If you are building in git you first need to generate the configure scripts:
$ ./otp_build autoconf
Configure and build all applications in OTP:
$ configure
$ make
or
$ ./otp_build configure
$ ./otp_build boot -a
If you have installed documentation in the OTP installation, also build the documentation:
$ make docs
After the successful build it's time to patch. The source tree directory, the directory of the installation and the
applications to patch are given as arguments to otp_patch_apply. The dependencies of each application are
validated against the applications in the installation and the other applications given as arguments. If a dependency
error is detected, the script will be aborted.
The otp_patch_apply syntax:
$ otp_patch_apply -s <Dir> -i <Dir> [-l <Dir>] [-c] [-f] [-h] \
[-n] [-v] <App1> [... <AppN>]
-s <Dir>
-i <Dir>
-l <Dir>
-- OTP source directory that contains build results.
-- OTP installation directory to patch.
-- Alternative OTP source library directory path(s)
containing build results of OTP applications.
Multiple paths should be colon separated.
38 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
1.5 Patching OTP Applications
-c
-f
-h
-n
-v
<AppX>
-- Cleanup (remove) old versions of applications
patched in the installation.
-- Force patch of application(s) even though
dependencies are not fulfilled (should only be
considered in a test environment).
-- Print help then exit.
-- Do not install documentation.
-- Print version then exit.
-- Application to patch.
Environment Variable:
ERL_LIBS -- Alternative OTP source library directory path(s)
containing build results of OTP applications.
Multiple paths should be colon separated.
Note:
The complete build environment is required while running otp_patch_apply.
Note:
All source directories identified by -s and -l should contain build results of OTP applications.
For example, if the user wants to install patched versions of mnesia and ssl built in /home/me/git/otp into
the OTP installation located in /opt/erlang/my_otp type
$ otp_patch_apply -s /home/me/git/otp -i /opt/erlang/my_otp \
mnesia ssl
Note:
If the list of applications contains core applications, i.e erts, kernel, stdlib or sasl, the Install script
in the patched Erlang/OTP installation must be rerun.
The patched applications are appended to the list of installed applications. Take a look at <InstallDir>/
releases/OTP-REL/installed_application_versions.
1.5.4 Sanity check
The application dependencies can be checked using the Erlang shell. Application dependencies are verified among
installed applications by otp_patch_apply, but these are not necessarily those actually loaded. By calling
system_information:sanity_check() one can validate dependencies among applications actually loaded.
1> system_information:sanity_check().
ok
Please take a look at the reference of sanity_check() for more information.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 39
2.1 System Principles
2 System Principles
2.1 System Principles
2.1.1 Starting the System
An Erlang runtime system is started with command erl:
% erl
Erlang/OTP 17 [erts-6.0] [hipe] [smp:8:8]
Eshell V6.0
1>
(abort with ^G)
erl understands a number of command-line arguments, see the erl(1) manual page in ERTS. Some of them are also
described in this chapter.
Application programs can access the values of the command-line arguments by calling the function
init:get_argument(Key) or init:get_arguments(). See the init(3) manual page in ERTS.
2.1.2 Restarting and Stopping the System
The runtime system is halted by calling halt/0,1. For details, see the erlang(3) manual page in ERTS.
The module init contains functions for restarting, rebooting, and stopping the runtime system:
init:restart()
init:reboot()
init:stop()
For details, see the init(3) manual page in ERTS.
The runtime system terminates if the Erlang shell is terminated.
2.1.3 Boot Scripts
The runtime system is started using a boot script. The boot script contains instructions on which code to load and
which processes and applications to start.
A boot script file has the extension .script. The runtime system uses a binary version of the script. This binary
boot script file has the extension .boot.
Which boot script to use is specified by the command-line flag -boot. The extension .boot is to be omitted. For
example, using the boot script start_all.boot:
% erl -boot start_all
If no boot script is specified, it defaults to ROOT/bin/start, see Default Boot Scripts.
40 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.1 System Principles
The command-line flag -init_debug makes the init process write some debug information while interpreting
the boot script:
% erl -init_debug
{progress,preloaded}
{progress,kernel_load_completed}
{progress,modules_loaded}
{start,heart}
{start,error_logger}
...
For a detailed description of the syntax and contents of the boot script, see the script(4) manual page in SASL.
Default Boot Scripts
Erlang/OTP comes with these boot scripts:
•
•
•
start_clean.boot - Loads the code for and starts the applications Kernel and STDLIB.
start_sasl.boot - Loads the code for and starts the applications Kernel, STDLIB, and SASL).
no_dot_erlang.boot - Loads the code for and starts the applications Kernel and STDLIB. Skips
loading the file .erlang. Useful for scripts and other tools that are to behave the same irrespective of user
preferences.
Which of start_clean and start_sasl to use as default is decided by the user when installing Erlang/OTP
using Install. The user is asked "Do you want to use a minimal system startup instead of the SASL startup". If the
answer is yes, then start_clean is used, otherwise start_sasl is used. A copy of the selected boot script is
made, named start.boot and placed in directory ROOT/bin.
User-Defined Boot Scripts
It is sometimes useful or necessary to create a user-defined boot script. This is true especially when running Erlang
in embedded mode, see Code Loading Strategy.
A boot script can be written manually. However, it is recommended to create a boot script by generating it from a
release resource file Name.rel, using the function systools:make_script/1,2. This requires that the source
code is structured as applications according to the OTP design principles. (The program does not have to be started
in terms of OTP applications, but can be plain Erlang).
For more information about .rel files, see OTP Design Principles and the rel(4) manual page in SASL.
The binary boot script file Name.boot is generated from the boot script file Name.script, using the function
systools:script2boot(File).
2.1.4 Code Loading Strategy
The runtime system can be started in either embedded or interactive mode. Which one is decided by the commandline flag -mode.
% erl -mode embedded
Default mode is interactive.
The mode properties are as follows:
•
In embedded mode, all code is loaded during system startup according to the boot script. (Code can also be
loaded later by explicitly ordering the code server to do so.)
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 41
2.2 Error Logging
•
In interactive mode, the code is dynamically loaded when first referenced. When a call to a function in a module
is made, and the module is not loaded, the code server searches the code path and loads the module into the
system.
Initially, the code path consists of the current working directory and all object code directories under ROOT/lib,
where ROOT is the installation directory of Erlang/OTP. Directories can be named Name[-Vsn]. The code server,
by default, chooses the directory with the highest version number among those which have the same Name. The Vsn suffix is optional. If an ebin directory exists under the Name[-Vsn] directory, this directory is added to the
code path.
The code path can be extended by using the command-line flags -pa Directories and -pz Directories.
These add Directories to the head or the end of the code path, respectively. Example:
% erl -pa /home/arne/mycode
The code server module code contains a number of functions for modifying and checking the search path, see the
code(3) manual page in Kernel.
2.1.5 File Types
The following file types are defined in Erlang/OTP:
File Type
File Name/Extension
Documented in
Module
.erl
Erlang Reference Manual
Include file
.hrl
Erlang Reference Manual
Release resource file
.rel
rel(4) manual page in SASL
Application resource file
.app
app(4) manual page in Kernel
Boot script
.script
script(4) manual page in SASL
Binary boot script
.boot
-
Configuration file
.config
config(4) manual page in Kernel
Application upgrade file
.appup
appup(4) manual page in SASL
Release upgrade file
relup
relup(4) manual page in SASL
Table 1.1: File Types
2.2 Error Logging
2.2.1 Error Information From the Runtime System
Error information from the runtime system, that is, information about a process terminating because of an uncaught
error exception, is by default written to terminal (tty):
42 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.2 Error Logging
=ERROR REPORT==== 9-Dec-2003::13:25:02 ===
Error in process <0.27.0> with exit value: {{badmatch,[1,2,3]},[{m,f,1},{shell,eval_loop,2}]}
The error information is handled by the error logger, a system process registered as error_logger. This process
receives all error messages from the Erlang runtime system as well as from the standard behaviours and different
Erlang/OTP applications.
The exit reasons (such as badarg) used by the runtime system are described in Errors and Error Handling.
For information about the process error_logger and its user interface (with the same name), see the
error_logger(3) manual page in Kernel. The system can be configured so that error information is written to file or to
tty, or both. In addition, user-defined applications can send and format error information using error_logger.
2.2.2 SASL Error Logging
The standard behaviours (supervisor, gen_server, and so on) send progress and error information to
error_logger. If the SASL application is started, this information is written to tty as well. For more information,
see SASL Error Logging in the SASL User's Guide.
% erl -boot start_sasl
Erlang (BEAM) emulator version 5.4.13 [hipe] [threads:0] [kernel-poll]
=PROGRESS REPORT==== 31-Mar-2006::12:45:58 ===
supervisor: {local,sasl_safe_sup}
started: [{pid,<0.33.0>},
{name,alarm_handler},
{mfa,{alarm_handler,start_link,[]}},
{restart_type,permanent},
{shutdown,2000},
{child_type,worker}]
=PROGRESS REPORT==== 31-Mar-2006::12:45:58 ===
supervisor: {local,sasl_safe_sup}
started: [{pid,<0.34.0>},
{name,overload},
{mfa,{overload,start_link,[]}},
{restart_type,permanent},
{shutdown,2000},
{child_type,worker}]
=PROGRESS REPORT==== 31-Mar-2006::12:45:58 ===
supervisor: {local,sasl_sup}
started: [{pid,<0.32.0>},
{name,sasl_safe_sup},
{mfa,{supervisor,
start_link,
[{local,sasl_safe_sup},sasl,safe]}},
{restart_type,permanent},
{shutdown,infinity},
{child_type,supervisor}]
=PROGRESS REPORT==== 31-Mar-2006::12:45:58 ===
supervisor: {local,sasl_sup}
started: [{pid,<0.35.0>},
{name,release_handler},
{mfa,{release_handler,start_link,[]}},
{restart_type,permanent},
{shutdown,2000},
{child_type,worker}]
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 43
2.3 Creating and Upgrading a Target System
=PROGRESS REPORT==== 31-Mar-2006::12:45:58 ===
application: sasl
started_at: nonode@nohost
Eshell V5.4.13 (abort with ^G)
1>
2.3 Creating and Upgrading a Target System
When creating a system using Erlang/OTP, the simplest way is to install Erlang/OTP somewhere, install the
application-specific code somewhere else, and then start the Erlang runtime system, making sure the code path includes
the application-specific code.
It is often not desirable to use an Erlang/OTP system as is. A developer can create new Erlang/OTP-compliant
applications for a particular purpose, and several original Erlang/OTP applications can be irrelevant for the purpose
in question. Thus, there is a need to be able to create a new system based on a given Erlang/OTP system, where
dispensable applications are removed and new applications are included. Documentation and source code is irrelevant
and is therefore not included in the new system.
This chapter is about creating such a system, which is called a target system.
The following sections deal with target systems with different requirements of functionality:
•
•
•
A basic target system that can be started by calling the ordinary erl script.
A simple target system where also code replacement in runtime can be performed.
An embedded target system where there is also support for logging output from the system to file for later
inspection, and where the system can be started automatically at boot time.
Here is only considered the case when Erlang/OTP is running on a UNIX system.
The sasl application includes the example Erlang module target_system.erl, which contains functions for
creating and installing a target system. This module is used in the following examples. The source code of the module
is listed in Listing of target_system.erl
2.3.1 Creating a Target System
It is assumed that you have a working Erlang/OTP system structured according to the OTP design principles.
Step 1. Create a .rel file (see the rel(4) manual page in SASL), which specifies the ERTS version and lists all
applications that are to be included in the new basic target system. An example is the following mysystem.rel file:
%% mysystem.rel
{release,
{"MYSYSTEM", "FIRST"},
{erts, "5.10.4"},
[{kernel, "2.16.4"},
{stdlib, "1.19.4"},
{sasl, "2.3.4"},
{pea, "1.0"}]}.
The listed applications are not only original Erlang/OTP applications but possibly also new applications that you have
written (here exemplified by the application Pea (pea)).
Step 2. Start Erlang/OTP from the directory where the mysystem.rel file resides:
os> erl -pa /home/user/target_system/myapps/pea-1.0/ebin
44 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.3 Creating and Upgrading a Target System
Here also the path to the pea-1.0 ebin directory is provided.
Step 3. Create the target system:
1> target_system:create("mysystem").
The function target_system:create/1 performs the following:
•
•
•
Reads the file mysystem.rel and creates a new file plain.rel that is identical to the former, except that
it only lists the Kernel and STDLIB applications.
From the files mysystem.rel and plain.rel creates the files mysystem.script, mysystem.boot,
plain.script, and plain.boot through a call to systools:make_script/2.
Creates the file mysystem.tar.gz by a call to systools:make_tar/2. That file has the following
contents:
erts-5.10.4/bin/
releases/FIRST/start.boot
releases/FIRST/mysystem.rel
releases/mysystem.rel
lib/kernel-2.16.4/
lib/stdlib-1.19.4/
lib/sasl-2.3.4/
lib/pea-1.0/
The file releases/FIRST/start.boot is a copy of our mysystem.boot
•
•
•
•
•
•
•
•
The release resource file mysystem.rel is duplicated in the tar file. Originally, this file was only stored in
the releases directory to make it possible for the release_handler to extract this file separately. After
unpacking the tar file, release_handler would automatically copy the file to releases/FIRST. However,
sometimes the tar file is unpacked without involving the release_handler (for example, when unpacking
the first target system). The file is therefore now instead duplicated in the tar file so no manual copying is needed.
Creates the temporary directory tmp and extracts the tar file mysystem.tar.gz into that directory.
Deletes the files erl and start from tmp/erts-5.10.4/bin. These files are created again from source
when installing the release.
Creates the directory tmp/bin.
Copies the previously created file plain.boot to tmp/bin/start.boot.
Copies the files epmd, run_erl, and to_erl from the directory tmp/erts-5.10.4/bin to the directory
tmp/bin.
Creates the directory tmp/log, which is used if the system is started as embedded with the bin/start
script.
Creates the file tmp/releases/start_erl.data with the contents "5.10.4 FIRST". This file is to be
passed as data file to the start_erl script.
Recreates the file mysystem.tar.gz from the directories in the directory tmp and removes tmp.
2.3.2 Installing a Target System
Step 4. Install the created target system in a suitable directory.
2> target_system:install("mysystem", "/usr/local/erl-target").
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 45
2.3 Creating and Upgrading a Target System
The function target_system:install/2 performs the following:
•
•
•
•
Extracts the tar file mysystem.tar.gz into the target directory /usr/local/erl-target.
In the target directory reads the file releases/start_erl.data to find the Erlang runtime system
version ("5.10.4").
Substitutes %FINAL_ROOTDIR% and %EMU% for /usr/local/erl-target and beam, respectively, in
the files erl.src, start.src, and start_erl.src of the target erts-5.10.4/bin directory, and
puts the resulting files erl, start, and run_erl in the target bin directory.
Finally the target releases/RELEASES file is created from data in the file releases/mysystem.rel.
2.3.3 Starting a Target System
Now we have a target system that can be started in various ways. We start it as a basic target system by invoking:
os> /usr/local/erl-target/bin/erl
Here only the Kernel and STDLIB applications are started, that is, the system is started as an ordinary development
system. Only two files are needed for all this to work:
•
•
bin/erl (obtained from erts-5.10.4/bin/erl.src)
bin/start.boot (a copy of plain.boot)
We can also start a distributed system (requires bin/epmd).
To start all applications specified in the original mysystem.rel file, use flag -boot as follows:
os> /usr/local/erl-target/bin/erl -boot /usr/local/erl-target/releases/FIRST/start
We start a simple target system as above. The only difference is that also the file releases/RELEASES is present
for code replacement in runtime to work.
To start an embedded target system, the shell script bin/start is used. The script calls bin/run_erl, which
in turn calls bin/start_erl (roughly, start_erl is an embedded variant of erl).
The shell script start, which is generated from erts-5.10.4/bin/start.src during installation, is only an example. Edit
it to suite your needs. Typically it is executed when the UNIX system boots.
run_erl is a wrapper that provides logging of output from the runtime system to file. It also provides a simple
mechanism for attaching to the Erlang shell (to_erl).
start_erl requires:
•
•
•
The root directory ("/usr/local/erl-target")
The releases directory ("/usr/local/erl-target/releases"
The location of the file start_erl.data
It performs the following:
•
•
•
Reads the runtime system version ("5.10.4") and release version ("FIRST") from the file
start_erl.data.
Starts the runtime system of the version found.
Provides the flag -boot specifying the boot file of the release version found ("releases/FIRST/
start.boot").
46 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.3 Creating and Upgrading a Target System
start_erl also assumes that there is sys.config in the release version directory ("releases/FIRST/
sys.config"). That is the topic of the next section.
The start_erl shell script is normally not to be altered by the user.
2.3.4 System Configuration Parameters
As was mentioned in the previous section, start_erl requires a sys.config in the release version directory
("releases/FIRST/sys.config"). If there is no such file, the system start fails. Such a file must therefore
also be added.
If you have system configuration data that is neither file-location-dependent nor site-dependent, it can be convenient to
create sys.config early, so it becomes part of the target system tar file created by target_system:create/1.
In fact, if you in the current directory create not only the file mysystem.rel, but also file sys.config, the latter
file is tacitly put in the appropriate directory.
2.3.5 Differences From the Install Script
The previous install/2 procedure differs somewhat from that of the ordinary Install shell script. In fact,
create/1 makes the release package as complete as possible, and leave to the install/2 procedure to finish by
only considering location-dependent files.
2.3.6 Creating the Next Version
In this example the Pea application has been changed, and so are the applications ERTS, Kernel, STDLIB and SASL.
Step 1. Create the file .rel:
%% mysystem2.rel
{release,
{"MYSYSTEM", "SECOND"},
{erts, "6.0"},
[{kernel, "3.0"},
{stdlib, "2.0"},
{sasl, "2.4"},
{pea, "2.0"}]}.
Step 2. Create the application upgrade file (see the appup(4) manual page in SASL) for Pea, for example:
%% pea.appup
{"2.0",
[{"1.0",[{load_module,pea_lib}]}],
[{"1.0",[{load_module,pea_lib}]}]}.
Step 3. From the directory where the file mysystem2.rel resides, start the Erlang/OTP system, giving the path
to the new version of Pea:
os> erl -pa /home/user/target_system/myapps/pea-2.0/ebin
Step 4. Create the release upgrade file (see the relup(4) manual page in SASL):
1> systools:make_relup("mysystem2",["mysystem"],["mysystem"],
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 47
2.3 Creating and Upgrading a Target System
[{path,["/home/user/target_system/myapps/pea-1.0/ebin",
"/my/old/erlang/lib/*/ebin"]}]).
Here "mysystem" is the base release and "mysystem2" is the release to upgrade to.
The path option is used for pointing out the old version of all applications. (The new versions are already in the code
path - assuming of course that the Erlang node on which this is executed is running the correct version of Erlang/OTP.)
Step 5. Create the new release:
2> target_system:create("mysystem2").
Given that the file relup generated in Step 4 is now located in the current directory, it is automatically included in
the release package.
2.3.7 Upgrading the Target System
This part is done on the target node, and for this example we want the node to be running as an embedded system with
the -heart option, allowing automatic restart of the node. For more information, see Starting a Target System.
We add -heart to bin/start:
#!/bin/sh
ROOTDIR=/usr/local/erl-target/
if [ -z "$RELDIR" ]
then
RELDIR=$ROOTDIR/releases
fi
START_ERL_DATA=${1:-$RELDIR/start_erl.data}
$ROOTDIR/bin/run_erl -daemon /tmp/ $ROOTDIR/log "exec $ROOTDIR/bin/start_erl $ROOTDIR\
$RELDIR $START_ERL_DATA -heart
We use the simplest possible sys.config, which we store in releases/FIRST:
%% sys.config
[].
Finally, to prepare the upgrade, we must put the new release package in the releases directory of the first target
system:
os> cp mysystem2.tar.gz /usr/local/erl-target/releases
Assuming that the node has been started as follows:
os> /usr/local/erl-target/bin/start
It can be accessed as follows:
48 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.3 Creating and Upgrading a Target System
os> /usr/local/erl-target/bin/to_erl /tmp/erlang.pipe.1
Logs can be found in /usr/local/erl-target/log. This directory is specified as an argument to run_erlin
the start script listed above.
Step 1. Unpack the release:
1> {ok,Vsn} = release_handler:unpack_release("mysystem2").
Step 2. Install the release:
2> release_handler:install_release(Vsn).
{continue_after_restart,"FIRST",[]}
heart: Tue Apr 1 12:15:10 2014: Erlang has closed.
heart: Tue Apr 1 12:15:11 2014: Executed "/usr/local/erl-target/bin/start /usr/local/erl-target/releases/new
[End]
The above return value and output after the call to release_handler:install_release/1 means that the
release_handler has restarted the node by using heart. This is always done when the upgrade involves a
change of the applications ERTS, Kernel, STDLIB, or SASL. For more information, see Upgrade when Erlang/OTP
has Changed.
The node is accessible through a new pipe:
os> /usr/local/erl-target/bin/to_erl /tmp/erlang.pipe.2
Check which releases there are in the system:
1> release_handler:which_releases().
[{"MYSYSTEM","SECOND",
["kernel-3.0","stdlib-2.0","sasl-2.4","pea-2.0"],
current},
{"MYSYSTEM","FIRST",
["kernel-2.16.4","stdlib-1.19.4","sasl-2.3.4","pea-1.0"],
permanent}]
Our new release, "SECOND", is now the current release, but we can also see that our "FIRST" release is still permanent.
This means that if the node would be restarted now, it would come up running the "FIRST" release again.
Step 3. Make the new release permanent:
2> release_handler:make_permanent("SECOND").
Check the releases again:
3> release_handler:which_releases().
[{"MYSYSTEM","SECOND",
["kernel-3.0","stdlib-2.0","sasl-2.4","pea-2.0"],
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 49
2.3 Creating and Upgrading a Target System
permanent},
{"MYSYSTEM","FIRST",
["kernel-2.16.4","stdlib-1.19.4","sasl-2.3.4","pea-1.0"],
old}]
We see that the new release version is permanent, so it would be safe to restart the node.
2.3.8 Listing of target_system.erl
This module can also be found in the examples directory of the SASL application.
-module(target_system).
-export([create/1, create/2, install/2]).
%% Note: RelFileName below is the *stem* without trailing .rel,
%% .script etc.
%%
%% create(RelFileName)
%%
create(RelFileName) ->
create(RelFileName,[]).
create(RelFileName,SystoolsOpts) ->
RelFile = RelFileName ++ ".rel",
Dir = filename:dirname(RelFileName),
PlainRelFileName = filename:join(Dir,"plain"),
PlainRelFile = PlainRelFileName ++ ".rel",
io:fwrite("Reading file: ~tp ...~n", [RelFile]),
{ok, [RelSpec]} = file:consult(RelFile),
io:fwrite("Creating file: ~tp from ~tp ...~n",
[PlainRelFile, RelFile]),
{release,
{RelName, RelVsn},
{erts, ErtsVsn},
AppVsns} = RelSpec,
PlainRelSpec = {release,
{RelName, RelVsn},
{erts, ErtsVsn},
lists:filter(fun({kernel, _}) ->
true;
({stdlib, _}) ->
true;
(_) ->
false
end, AppVsns)
},
{ok, Fd} = file:open(PlainRelFile, [write]),
io:fwrite(Fd, "~p.~n", [PlainRelSpec]),
file:close(Fd),
io:fwrite("Making \"~ts.script\" and \"~ts.boot\" files ...~n",
[PlainRelFileName,PlainRelFileName]),
make_script(PlainRelFileName,SystoolsOpts),
io:fwrite("Making \"~ts.script\" and \"~ts.boot\" files ...~n",
[RelFileName, RelFileName]),
make_script(RelFileName,SystoolsOpts),
TarFileName = RelFileName ++ ".tar.gz",
io:fwrite("Creating tar file ~tp ...~n", [TarFileName]),
50 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.3 Creating and Upgrading a Target System
make_tar(RelFileName,SystoolsOpts),
TmpDir = filename:join(Dir,"tmp"),
io:fwrite("Creating directory ~tp ...~n",[TmpDir]),
file:make_dir(TmpDir),
io:fwrite("Extracting ~tp into directory ~tp ...~n", [TarFileName,TmpDir]),
extract_tar(TarFileName, TmpDir),
TmpBinDir = filename:join([TmpDir, "bin"]),
ErtsBinDir = filename:join([TmpDir, "erts-" ++ ErtsVsn, "bin"]),
io:fwrite("Deleting \"erl\" and \"start\" in directory ~tp ...~n",
[ErtsBinDir]),
file:delete(filename:join([ErtsBinDir, "erl"])),
file:delete(filename:join([ErtsBinDir, "start"])),
io:fwrite("Creating temporary directory ~tp ...~n", [TmpBinDir]),
file:make_dir(TmpBinDir),
io:fwrite("Copying file \"~ts.boot\" to ~tp ...~n",
[PlainRelFileName, filename:join([TmpBinDir, "start.boot"])]),
copy_file(PlainRelFileName++".boot",filename:join([TmpBinDir, "start.boot"])),
io:fwrite("Copying files \"epmd\", \"run_erl\" and \"to_erl\" from \n"
"~tp to ~tp ...~n",
[ErtsBinDir, TmpBinDir]),
copy_file(filename:join([ErtsBinDir, "epmd"]),
filename:join([TmpBinDir, "epmd"]), [preserve]),
copy_file(filename:join([ErtsBinDir, "run_erl"]),
filename:join([TmpBinDir, "run_erl"]), [preserve]),
copy_file(filename:join([ErtsBinDir, "to_erl"]),
filename:join([TmpBinDir, "to_erl"]), [preserve]),
%% This is needed if 'start' script created from 'start.src' shall
%% be used as it points out this directory as log dir for 'run_erl'
TmpLogDir = filename:join([TmpDir, "log"]),
io:fwrite("Creating temporary directory ~tp ...~n", [TmpLogDir]),
ok = file:make_dir(TmpLogDir),
StartErlDataFile = filename:join([TmpDir, "releases", "start_erl.data"]),
io:fwrite("Creating ~tp ...~n", [StartErlDataFile]),
StartErlData = io_lib:fwrite("~s ~s~n", [ErtsVsn, RelVsn]),
write_file(StartErlDataFile, StartErlData),
io:fwrite("Recreating tar file ~tp from contents in directory ~tp ...~n",
[TarFileName,TmpDir]),
{ok, Tar} = erl_tar:open(TarFileName, [write, compressed]),
%% {ok, Cwd} = file:get_cwd(),
%% file:set_cwd("tmp"),
ErtsDir = "erts-"++ErtsVsn,
erl_tar:add(Tar, filename:join(TmpDir,"bin"), "bin", []),
erl_tar:add(Tar, filename:join(TmpDir,ErtsDir), ErtsDir, []),
erl_tar:add(Tar, filename:join(TmpDir,"releases"), "releases", []),
erl_tar:add(Tar, filename:join(TmpDir,"lib"), "lib", []),
erl_tar:add(Tar, filename:join(TmpDir,"log"), "log", []),
erl_tar:close(Tar),
%% file:set_cwd(Cwd),
io:fwrite("Removing directory ~tp ...~n",[TmpDir]),
remove_dir_tree(TmpDir),
ok.
install(RelFileName, RootDir) ->
TarFile = RelFileName ++ ".tar.gz",
io:fwrite("Extracting ~tp ...~n", [TarFile]),
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 51
2.3 Creating and Upgrading a Target System
extract_tar(TarFile, RootDir),
StartErlDataFile = filename:join([RootDir, "releases", "start_erl.data"]),
{ok, StartErlData} = read_txt_file(StartErlDataFile),
[ErlVsn, _RelVsn| _] = string:tokens(StartErlData, " \n"),
ErtsBinDir = filename:join([RootDir, "erts-" ++ ErlVsn, "bin"]),
BinDir = filename:join([RootDir, "bin"]),
io:fwrite("Substituting in erl.src, start.src and start_erl.src to "
"form erl, start and start_erl ...\n"),
subst_src_scripts(["erl", "start", "start_erl"], ErtsBinDir, BinDir,
[{"FINAL_ROOTDIR", RootDir}, {"EMU", "beam"}],
[preserve]),
%%! Workaround for pre OTP 17.0: start.src and start_erl.src did
%%! not have correct permissions, so the above 'preserve' option did not help
ok = file:change_mode(filename:join(BinDir,"start"),8#0755),
ok = file:change_mode(filename:join(BinDir,"start_erl"),8#0755),
io:fwrite("Creating the RELEASES file ...\n"),
create_RELEASES(RootDir, filename:join([RootDir, "releases",
filename:basename(RelFileName)])).
%% LOCALS
%% make_script(RelFileName,Opts)
%%
make_script(RelFileName,Opts) ->
systools:make_script(RelFileName, [no_module_tests,
{outdir,filename:dirname(RelFileName)}
|Opts]).
%% make_tar(RelFileName,Opts)
%%
make_tar(RelFileName,Opts) ->
RootDir = code:root_dir(),
systools:make_tar(RelFileName, [{erts, RootDir},
{outdir,filename:dirname(RelFileName)}
|Opts]).
%% extract_tar(TarFile, DestDir)
%%
extract_tar(TarFile, DestDir) ->
erl_tar:extract(TarFile, [{cwd, DestDir}, compressed]).
create_RELEASES(DestDir, RelFileName) ->
release_handler:create_RELEASES(DestDir, RelFileName ++ ".rel").
subst_src_scripts(Scripts, SrcDir, DestDir, Vars, Opts) ->
lists:foreach(fun(Script) ->
subst_src_script(Script, SrcDir, DestDir,
Vars, Opts)
end, Scripts).
subst_src_script(Script, SrcDir, DestDir, Vars, Opts) ->
subst_file(filename:join([SrcDir, Script ++ ".src"]),
filename:join([DestDir, Script]),
Vars, Opts).
subst_file(Src, Dest, Vars, Opts) ->
{ok, Conts} = read_txt_file(Src),
NConts = subst(Conts, Vars),
write_file(Dest, NConts),
case lists:member(preserve, Opts) of
true ->
{ok, FileInfo} = file:read_file_info(Src),
file:write_file_info(Dest, FileInfo);
false ->
52 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.3 Creating and Upgrading a Target System
end.
ok
%% subst(Str, Vars)
%% Vars = [{Var, Val}]
%% Var = Val = string()
%% Substitute all occurrences of %Var% for Val in Str, using the list
%% of variables in Vars.
%%
subst(Str, Vars) ->
subst(Str, Vars, []).
subst([$%, C| Rest], Vars, Result) when $A =< C, C =< $Z ->
subst_var([C| Rest], Vars, Result, []);
subst([$%, C| Rest], Vars, Result) when $a =< C, C =< $z ->
subst_var([C| Rest], Vars, Result, []);
subst([$%, C| Rest], Vars, Result) when C == $_ ->
subst_var([C| Rest], Vars, Result, []);
subst([C| Rest], Vars, Result) ->
subst(Rest, Vars, [C| Result]);
subst([], _Vars, Result) ->
lists:reverse(Result).
subst_var([$%| Rest], Vars, Result, VarAcc) ->
Key = lists:reverse(VarAcc),
case lists:keysearch(Key, 1, Vars) of
{value, {Key, Value}} ->
subst(Rest, Vars, lists:reverse(Value, Result));
false ->
subst(Rest, Vars, [$%| VarAcc ++ [$%| Result]])
end;
subst_var([C| Rest], Vars, Result, VarAcc) ->
subst_var(Rest, Vars, Result, [C| VarAcc]);
subst_var([], Vars, Result, VarAcc) ->
subst([], Vars, [VarAcc ++ [$%| Result]]).
copy_file(Src, Dest) ->
copy_file(Src, Dest, []).
copy_file(Src, Dest, Opts) ->
{ok,_} = file:copy(Src, Dest),
case lists:member(preserve, Opts) of
true ->
{ok, FileInfo} = file:read_file_info(Src),
file:write_file_info(Dest, FileInfo);
false ->
ok
end.
write_file(FName, Conts) ->
Enc = file:native_name_encoding(),
{ok, Fd} = file:open(FName, [write]),
file:write(Fd, unicode:characters_to_binary(Conts,Enc,Enc)),
file:close(Fd).
read_txt_file(File) ->
{ok, Bin} = file:read_file(File),
{ok, binary_to_list(Bin)}.
remove_dir_tree(Dir) ->
remove_all_files(".", [Dir]).
remove_all_files(Dir, Files) ->
lists:foreach(fun(File) ->
FilePath = filename:join([Dir, File]),
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 53
2.4 Upgrade when Erlang/OTP has Changed
case filelib:is_dir(FilePath) of
true ->
{ok, DirFiles} = file:list_dir(FilePath),
remove_all_files(FilePath, DirFiles),
file:del_dir(FilePath);
_ ->
file:delete(FilePath)
end
end, Files).
2.4 Upgrade when Erlang/OTP has Changed
2.4.1 Introduction
As of Erlang/OTP 17, most applications deliver a valid application upgrade file (appup). In earlier releases,
a majority of the applications in Erlang/OTP did not support upgrade. Many of the applications use the
restart_application instruction. These are applications for which it is not crucial to support real soft upgrade,
for example, tools and library applications. The restart_application instruction ensures that all modules in
the application are reloaded and thereby running the new code.
2.4.2 Upgrade of Core Applications
The core applications ERTS, Kernel, STDLIB, and SASL never allow real soft upgrade, but require the Erlang emulator
to be restarted. This is indicated to the release_handler by the upgrade instruction restart_new_emulator.
This instruction is always the very first instruction executed, and it restarts the emulator with the new versions of the
above mentioned core applications and the old versions of all other applications. When the node is back up, all other
upgrade instructions are executed, making sure each application is finally running its new version.
It might seem strange to do a two-step upgrade instead of just restarting the emulator with the new version of all
applications. The reason for this design decision is to allow code_change functions to have side effects, for
example, changing data on disk. It also guarantees that the upgrade mechanism for non-core applications does not
differ depending on whether or not core applications are changed at the same time.
If, however, the more brutal variant is preferred, the the release upgrade file can be handwritten using only the single
upgrade instruction restart_emulator. This instruction, in contrast to restart_new_emulator, causes the
emulator to restart with the new versions of all applications.
Note: If other instructions are included before restart_emulator in the handwritten relup file, they are
executed in the old emulator. This is a big risk since there is no guarantee that new beam code can be loaded into
the old emulator. Adding instructions after restart_emulator has no effect as the release_handler will
not execute them.
For information about the release upgrade file, see the relup(4) manual page in SASL. For more information about
upgrade instructions, see the appup(4) manual page in SASL.
2.4.3 Applications that Still do Not Allow Code Upgrade
A few applications, such as HiPE, do not support upgrade. This is indicated by an application upgrade file containing
only {Vsn,[],[]}. Any attempt at creating a release upgrade file with such input fails. The only way to force an
upgrade involving applications like this is to handwrite the file relup, preferably as described above with only the
restart_emulator instruction.
54 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.5 Versions
2.5 Versions
2.5.1 OTP Version
As of OTP release 17, the OTP release number corresponds to the major part of the OTP version. The OTP version as
a concept was introduced in OTP 17. The version scheme used is described in detail in Version Scheme.
OTP of a specific version is a set of applications of specific versions. The application versions identified by an OTP
version corresponds to application versions that have been tested together by the Erlang/OTP team at Ericsson AB.
An OTP system can, however, be put together with applications from different OTP versions. Such a combination
of application versions has not been tested by the Erlang/OTP team. It is therefore always preferred to use OTP
applications from one single OTP version.
Release candidates have an -rc<N> suffix. The suffix -rc0 is used during development up to the first release
candidate.
Retrieving Current OTP Version
In an OTP source code tree, the OTP version can be read from the text file <OTP source root>/OTP_VERSION.
The absolute path to the file can be constructed by calling filename:join([code:root_dir(),
"OTP_VERSION"]).
In an installed OTP development system, the OTP version can be read from the text file <OTP
installation root>/releases/<OTP release number>/OTP_VERSION. The absolute path
to the file can by constructed by calling filename:join([code:root_dir(),
"releases",
erlang:system_info(otp_release), "OTP_VERSION"]).
If the version read from the OTP_VERSION file in a development system has a ** suffix, the system has been patched
using the otp_patch_apply tool. In this case, the system consists of application versions from multiple OTP
versions. The version preceding the ** suffix corresponds to the OTP version of the base system that has been patched.
Notice that if a development system is updated by other means than otp_patch_apply, the file OTP_VERSION
can identify an incorrect OTP version.
No OTP_VERSION file is placed in a target system created by OTP tools. This since one easily can create a target
system where it is hard to even determine the base OTP version. You can, however, place such a file there if you
know the OTP version.
OTP Versions Table
The text file <OTP source root>/otp_versions.table, which is part of the source code, contains
information about all OTP versions from OTP 17.0 up to the current OTP version. Each line contains information
about application versions that are part of a specific OTP version, and has the following format:
<OtpVersion> : <ChangedAppVersions> # <UnchangedAppVersions> :
<OtpVersion> has the format OTP-<VSN>, that is, the same as the git tag used to identify the source.
<ChangedAppVersions> and <UnchangedAppVersions> are space-separated lists of application versions
and has the format <application>-<vsn>.
•
•
<ChangedAppVersions> corresponds to changed applications with new version numbers in this OTP
version.
<UnchangedAppVersions> corresponds to unchanged application versions in this OTP version.
Both of them can be empty, but not at the same time. If <ChangedAppVersions> is empty, no changes have been
made that change the build result of any application. This could, for example, be a pure bug fix of the build system.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 55
2.5 Versions
The order of lines is undefined. All white-space characters in this file are either space (character 32) or line-break
(character 10).
By using ordinary UNIX tools like sed and grep one can easily find answers to various questions like:
•
Which OTP versions are kernel-3.0 part of?
•
$ grep ' kernel-3\.0 ' otp_versions.table
In which OTP version was kernel-3.0 introduced?
$ sed 's/#.*//;/ kernel-3\.0 /!d' otp_versions.table
The above commands give a bit more information than the exact answers, but adequate information when manually
searching for answers to these questions.
Warning:
The format of the otp_versions.table might be subject to changes during the OTP 17 release.
2.5.2 Application Version
As of OTP 17.0 application versions use the same version scheme as the OTP version. Application versions part of a
release candidate will however not have an -rc<N> suffix as the OTP version. Also note that a major increment in
an application version does not necessarily imply a major increment of the OTP version. This depends on whether the
major change in the application is considered as a major change for OTP as a whole or not.
2.5.3 Version Scheme
Note:
The version scheme was changed as of OTP 17.0. This implies that application versions used prior to OTP 17.0
do not adhere to this version scheme. A list of application versions used in OTP 17.0 is included at the end of
this section
In the normal case, a version is constructed as <Major>.<Minor>.<Patch>, where <Major> is the most
significant part.
However, more dot-separated parts than this can exist. The dot-separated parts consist of non-negative
integers. If all parts less significant than <Minor> equals 0, they are omitted. The three normal parts
<Major>.<Minor>.<Patch> are changed as follows:
•
•
•
<Major> - Increases when major changes, including incompatibilities, are made.
<Minor> - Increases when new functionality is added.
<Patch> - Increases when pure bug fixes are made.
When a part in the version number increases, all less significant parts are set to 0.
An application version or an OTP version identifies source code versions. That is, it implies nothing about how the
application or OTP has been built.
Order of Versions
Version numbers in general are only partially ordered. However, normal version numbers (with three parts) as of OTP
17.0 have a total or linear order. This applies both to normal OTP versions and normal application versions.
56 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
2.5 Versions
When comparing two version numbers that have an order, one compare each part as ordinary integers from the most
significant part to less significant parts. The order is defined by the first parts of the same significance that differ.
An OTP version with a larger version includes all changes that are part of a smaller OTP version. The same goes for
application versions.
In general, versions can have more than three parts. The versions are then only partially ordered. Such versions are
only used in exceptional cases. When an extra part (out of the normal three parts) is added to a version number, a new
branch of versions is made. The new branch has a linear order against the base version. However, versions on different
branches have no order, and therefore one can only conclude that they all include what is included in their closest
common ancestor. When branching multiple times from the same base version, 0 parts are added between the base
version and the least significant 1 part until a unique version is found. Versions that have an order can be compared
as described in the previous paragraph.
An example of branched versions: The version 6.0.2.1 is a branched version from the base version 6.0.2. Versions
on the form 6.0.2.<X> can be compared with normal versions smaller than or equal to 6.0.2, and other versions
on the form 6.0.2.<X>. The version 6.0.2.1 will include all changes in 6.0.2. However, 6.0.3 will most
likely not include all changes in 6.0.2.1 (note that these versions have no order). A second branched version from
the base version 6.0.2 will be version 6.0.2.0.1, and a third branched version will be 6.0.2.0.0.1.
2.5.4 OTP 17.0 Application Versions
The following list details the application versions that were part of OTP 17.0. If the normal part of an application
version number compares as smaller than the corresponding application version in the list, the version number does not
adhere to the version scheme introduced in OTP 17.0 and is to be considered as not having an order against versions
used as of OTP 17.0.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
asn1-3.0
common_test-1.8
compiler-5.0
cosEvent-2.1.15
cosEventDomain-1.1.14
cosFileTransfer-1.1.16
cosNotification-1.1.21
cosProperty-1.1.17
cosTime-1.1.14
cosTransactions-1.2.14
crypto-3.3
debugger-4.0
dialyzer-2.7
diameter-1.6
edoc-0.7.13
eldap-1.0.3
erl_docgen-0.3.5
erl_interface-3.7.16
erts-6.0
et-1.5
eunit-2.2.7
gs-1.5.16
hipe-3.10.3
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 57
2.5 Versions
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
ic-4.3.5
inets-5.10
jinterface-1.5.9
kernel-3.0
megaco-3.17.1
mnesia-4.12
observer-2.0
odbc-2.10.20
orber-3.6.27
os_mon-2.2.15
ose-1.0
otp_mibs-1.0.9
parsetools-2.0.11
percept-0.8.9
public_key-0.22
reltool-0.6.5
runtime_tools-1.8.14
sasl-2.4
snmp-4.25.1
ssh-3.0.1
ssl-5.3.4
stdlib-2.0
syntax_tools-1.6.14
test_server-3.7
tools-2.6.14
typer-0.9.6
webtool-0.8.10
wx-1.2
xmerl-1.3.7
58 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
3.1 Embedded Solaris
3 Embedded Systems User's Guide
This section describes the issues that are specific for running Erlang on an embedded system. It describes the
differences in installing and starting Erlang compared to how it is done for a non-embedded system.
Note:
This is a supplementary section. You also need to read Section 1 Installation Guide.
There is also target architecture-specific information in the top-level README file of the Erlang distribution.
3.1 Embedded Solaris
This section describes the operating system-specific parts of OTP that relate to Solaris.
3.1.1 Memory Use
Solaris takes about 17 MB of RAM on a system with 64 MB of total RAM. This leaves about 47 MB for the
applications. If the system uses swapping, these figures cannot be improved because unnecessary daemon processes
are swapped out. However, if swapping is disabled, or if the swap space is of limited resource in the system, it becomes
necessary to kill off unnecessary daemon processes.
3.1.2 Disk Space Use
The disk space required by Solaris can be minimized by using the Core User support installation. It requires about 80
MB of disk space. This installs only the minimum software required to boot and run Solaris. The disk space can be
further reduced by deleting unnecessary individual files. However, unless disk space is a critical resource the effort
required and the risks involved cannot be justified.
3.1.3 Installing an Embedded System
This section is about installing an embedded system. The following topics are considered:
•
•
•
•
•
•
•
•
Creating user and installation directory
Installing an embedded system
Configuring automatic start at boot
Making a hardware watchdog available
Changing permission for reboot
Setting TERM environment variable
Adding patches
Installing module os_sup in application os_mon
Several of the procedures in this section require expert knowledge of the Solaris operating system. For most of them
super user privilege is needed.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 59
3.1 Embedded Solaris
Creating User and Installation Directory
It is recommended that the embedded environment is run by an ordinary user, that is, a user who does not have super
user privileges.
In this section, it is assumed that the username is otpuser and that the home directory of that user is:
/export/home/otpuser
It is also assumed that in the home directory of otpuser, there is a directory named otp, the full path of which is:
/export/home/otpuser/otp
This directory is the installation directory of the embedded environment.
Installing an Embedded System
The procedure for installing an embedded system is the same as for an ordinary system (see Installation Guide), except
for the following:
•
•
The (compressed) tape archive file is to be extracted in the installation directory defined above.
It is not needed to link the start script to a standard directory like /usr/local/bin.
Configuring Automatic Start at Boot
A true embedded system must start when the system boots. This section accounts for the necessary configurations
needed to achieve that.
The embedded system and all the applications start automatically if the script file shown below is added to directory
/etc/rc3.d. The file must be owned and readable by root. Its name cannot be arbitrarily assigned; the following
name is recommended:
S75otp.system
For more details on initialization (and termination) scripts, and naming thereof, see the Solaris documentation.
#!/bin/sh
#
# File name:
# Purpose:
#
# Author:
# Resides in:
#
S75otp.system
Automatically starts Erlang and applications when the
system starts
janne@erlang.ericsson.se
/etc/rc3.d
if [ ! -d /usr/bin ]
then
exit
fi
# /usr not mounted
killproc() {
# kill the named process(es)
pid=`/usr/bin/ps -e |
/usr/bin/grep -w $1 |
/usr/bin/sed -e 's/^ *//' -e 's/ .*//'`
60 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
3.1 Embedded Solaris
}
[ "$pid" != "" ] && kill $pid
# Start/stop processes required for Erlang
case "$1" in
'start')
# Start the Erlang emulator
#
su - otpuser -c "/export/home/otpuser/otp/bin/start" &
;;
'stop')
killproc beam
;;
*)
echo "Usage: $0 { start | stop }"
;;
esac
File /export/home/otpuser/otp/bin/start referred to in the above script is precisely the start script
described in Starting Erlang. The script variable OTP_ROOT in that start script corresponds to the following
example path used in this section:
/export/home/otpuser/otp
The start script is to be edited accordingly.
Use of the killproc procedure in the above script can be combined with a call to erl_call, for example:
$SOME_PATH/erl_call -n Node init stop
To take Erlang down gracefully, see the erl_call(1) manual page in erl_interface for details on the use of
erl_call. However, that requires that Erlang runs as a distributed node, which is not always the case.
The killproc procedure is not to be removed. The purpose is here to move from run level 3 (multi-user mode with
networking resources) to run level 2 (multi-user mode without such resources), in which Erlang is not to run.
Making Hardware Watchdog Available
For Solaris running on VME boards from Force Computers, the onboard hardware watchdog can be activated, provided
a VME bus driver is added to the operating system (see also Installation Problems).
See also the heart(3) manual page in Kernel.
Changing Permissions for Reboot
If the HEART_COMMAND environment variable is to be set in the start script in Starting Erlang, and if the value
is to be set to the path of the Solaris reboot command, that is:
HEART_COMMAND=/usr/sbin/reboot
then the ownership and file permissions for /usr/sbin/reboot must be changed as follows:
chown 0 /usr/sbin/reboot
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 61
3.1 Embedded Solaris
chmod 4755 /usr/sbin/reboot
See also the heart(3) manual page in Kernel.
Setting TERM Environment Variable
When the Erlang runtime system is automatically started from the S75otp.system script, the TERM environment
variable must be set. The following is a minimal setting:
TERM=sun
This is to be added to the start script.
Adding Patches
For proper functioning of flushing file system data to disk on Solaris 2.5.1, the version-specific patch with number
103640-02 must be added to the operating system. Other patches might be needed, see the release README file
<ERL_INSTALL_DIR>/README.
Installing Module os_sup in Application os_mon
The following four installation procedures require super user privilege:
Installation
• Make a copy of the Solaris standard configuration file for syslogd:
• Make a copy of the Solaris standard configuration file for syslogd. This file is usually named
syslog.conf and found in directory /etc.
• The filename of the copy must be syslog.conf.ORIG. The directory location is optional; usually it is /
etc. A simple way to do this is to issue the following command:
cp /etc/syslog.conf /etc/syslog.conf.ORIG
•
Make an Erlang-specific configuration file for syslogd:
• Make an edited copy of the backup copy previously made.
• The filename must be syslog.conf.OTP. The path must be the same as the backup copy.
• The format of the configuration file is found in the syslog.conf(5) manual page, by issuing the
command man syslog.conf.
• Usually a line is added that is to state:
• Which types of information that is to be supervised by Erlang
• The name of the file (actually a named pipe) that is to receive the information
• If, for example, only information originating from the UNIX kernel is to be supervised, the line is to begin
with kern.LEVEL. For the possible values of LEVEL, see syslog.conf(5).
• After at least one tab-character, the line added is to contain the full name of the named pipe where
syslogd writes its information. The path must be the same as for the files syslog.conf.ORIG and
syslog.conf.OTP. The filename must be syslog.otp.
• If the directory for the files syslog.conf.ORIG and syslog.conf.OTP is /etc, the line in
syslog.conf.OTP is as follows:
kern.LEVEL
/etc/syslog.otp
62 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
3.1 Embedded Solaris
•
Check the file privileges of the configuration files:
• The configuration files is to have rw-r--r-- file privileges and be owned by root.
• A simple way to do this is to issue these commands:
chmod 644 /etc/syslog.conf
chmod 644 /etc/syslog.conf.ORIG
chmod 644 /etc/syslog.conf.OTP
•
•
Notice that if the files syslog.conf.ORIG and syslog.conf.OTP are not in directory /etc, the
file path in the second and third command must be modified.
Modify file privileges and ownership of the mod_syslog utility:
• The file privileges and ownership of the mod_syslog utility must be modified.
• The full name of the binary executable file is derived from the position of application os_mon in the file
system by adding /priv/bin/mod_syslog. The generic full name of the binary executable file is thus:
<OTP_ROOT>/lib/os_mon-<REV>/priv/bin/mod_syslog
•
•
Example: If the path to otp-root is /usr/otp, then the path to the os_mon application is /usr/otp/
lib/os_mon-1.0 (assuming revision 1.0) and the full name of the binary executable file is /usr/otp/
lib/os_mon-1.0/priv/bin/mod_syslog.
The binary executable file must be owned by root, have rwsr-xr-x file privileges, in particular the
setuid bit of the user must be set.
A simple way to do this is to issue the following commands:
cd <OTP_ROOT>/lib/os_mon-<REV>/priv/bin/mod_syslog
chmod 4755 mod_syslog
chown root mod_syslog
Testing the Application Configuration File
The following procedure does not require root privilege:
•
•
Ensure that the configuration parameters for the os_sup module in the os_mon application are correct.
Browse the application configuration file (do not edit it). The full name of the application configuration file is
derived from the position of the os_mon application in the file system by adding /ebin/os_mon.app.
The generic full name of the file is thus:
<OTP_ROOT>/lib/os_mon-<REV>/ebin/os_mon.app.
•
Example: If the path to otp-root is /usr/otp, then the path to the os_mon application is /usr/otp/
lib/os_mon-1.0 (assuming revision 1.0) and the full name of the binary executable file is /usr/otp/
lib/os_mon-1.0/ebin/os_mon.app.
Ensure that the following configuration parameters have correct values:
Parameter
Function
Standard value
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 63
3.1 Embedded Solaris
true for the first instance on the
hardware; false for the other
instances
start_os_sup
Specifies if os_sup is to be started
or not.
os_sup_own
The directory for (1) back-up copy
and (2) Erlang-specific configuration "/etc"
file for syslogd
os_sup_syslogconf
The full name for the Solaris
standard configuration file for
syslogd
"/etc/syslog.conf"
error_tag
The tag for the messages that are
sent to the error logger in the Erlang
runtime system
std_error
Table 1.1: Configuration Parameters
If the values listed in os_mon.app do not suit your needs, do not edit that file. Instead override the values in a
system configuration file, the full pathname of which is given on the command line to erl.
Example: Contents of an application configuration file:
[{os_mon, [{start_os_sup, true}, {os_sup_own, "/etc"},
{os_sup_syslogconf, "/etc/syslog.conf"}, {os_sup_errortag, std_error}]}].
Related Documents
See the os_mon(3) application, the application(3) manual page in Kernel, and the erl(1) manual page in
ERTS.
Installation Problems
The hardware watchdog timer, which is controlled by the heart port program, requires package FORCEvme, which
contains the VME bus driver, to be installed. However, this driver can clash with the Sun mcp driver and cause the
system to refuse to boot. To cure this problem, the following lines are to be added to /etc/system:
•
•
•
exclude: drv/mcp
exclude: drv/mcpzsa
exclude: drv/mcpp
Warning:
It is recommended to add these lines to avoid a clash. The clash can make it impossible to boot the system.
3.1.4 Starting Erlang
This section describes how an embedded system is started. Four programs are involved and they normally reside in the
directory <ERL_INSTALL_DIR>/bin. The only exception is the start program, which can be located anywhere,
and is also the only program that must be modified by the user.
64 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
3.1 Embedded Solaris
In an embedded system, there is usually no interactive shell. However, an operator can attach to the Erlang system by
command to_erl. The operator is then connected to the Erlang shell and can give ordinary Erlang commands. All
interaction with the system through this shell is logged in a special directory.
Basically, the procedure is as follows:
•
•
•
The start program is called when the machine is started.
It calls run_erl, which sets up things so the operator can attach to the system.
It calls start_erl, which calls the correct version of erlexec (which is located in
<ERL_INSTALL_DIR>/erts-EVsn/bin) with the correct boot and config files.
3.1.5 Programs
start
This program is called when the machine is started. It can be modified or rewritten to suit a special system. By default,
it must be called start and reside in <ERL_INSTALL_DIR>/bin. Another start program can be used, by using
configuration parameter start_prg in application SASL.
The start program must call run_erl as shown below. It must also take an optional parameter, which defaults to
<ERL_INSTALL_DIR>/releases/start_erl.data.
This program is to set static parameters and environment variables such as -sname Name and HEART_COMMAND
to reboot the machine.
The <RELDIR> directory is where new release packets are installed, and where the release handler keeps information
about releases. For more information, see the release_handler(3) manual page in SASL.
The following script illustrates the default behaviour of the program:
#!/bin/sh
# Usage: start [DataFile]
#
ROOTDIR=/usr/local/otp
if [ -z "$RELDIR" ]
then
RELDIR=$ROOTDIR/releases
fi
START_ERL_DATA=${1:-$RELDIR/start_erl.data}
$ROOTDIR/bin/run_erl /tmp/ $ROOTDIR/log "exec $ROOTDIR/bin/start_erl \
$ROOTDIR $RELDIR $START_ERL_DATA" > /dev/null 2>&1 &
The following script illustrates a modification where the node is given the name cp1, and where the environment
variables HEART_COMMAND and TERM have been added to the previous script:
#!/bin/sh
# Usage: start [DataFile]
#
HEART_COMMAND=/usr/sbin/reboot
TERM=sun
export HEART_COMMAND TERM
ROOTDIR=/usr/local/otp
if [ -z "$RELDIR" ]
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 65
3.1 Embedded Solaris
then
RELDIR=$ROOTDIR/releases
fi
START_ERL_DATA=${1:-$RELDIR/start_erl.data}
$ROOTDIR/bin/run_erl /tmp/ $ROOTDIR/log "exec $ROOTDIR/bin/start_erl \
$ROOTDIR $RELDIR $START_ERL_DATA -heart -sname cp1" > /dev/null 2>&1 &
If a diskless and/or read-only client node is about to start, file start_erl.data is located in the client directory
at the master node. Thus, the START_ERL_DATA line is to look like:
CLIENTDIR=$ROOTDIR/clients/clientname
START_ERL_DATA=${1:-$CLIENTDIR/bin/start_erl.data}
run_erl
This program is used to start the emulator, but you will not be connected to the shell. to_erl is used to connect
to the Erlang shell.
Usage: run_erl pipe_dir/ log_dir "exec command [parameters ...]"
Here:
•
•
•
•
pipe_dir/ is to be /tmp/ (to_erl uses this name by default).
log_dir is where the log files are written.
command [parameters] is executed.
Everything written to stdin and stdout is logged in log_dir.
Log files are written in log_dir. Each log file has a name of the form erlang.log.N, where N is a generation
number, ranging from 1 to 5. Each log file holds up to 100 kB text. As time goes by, the following log files are found
in the log file directory:
erlang.log.1
erlang.log.1,
erlang.log.1,
erlang.log.1,
erlang.log.2,
erlang.log.3,
...
erlang.log.2
erlang.log.2,
erlang.log.2,
erlang.log.3,
erlang.log.4,
erlang.log.3
erlang.log.3, erlang.log.4
erlang.log.4, erlang.log.5
erlang.log.5, erlang.log.1
The most recent log file is the rightmost in each row. That is, the most recent file is the one with the highest number,
or if there are already four files, the one before the skip.
When a log file is opened (for appending or created), a time stamp is written to the file. If nothing has been written to
the log files for 15 minutes, a record is inserted that says that we are still alive.
to_erl
This program is used to attach to a running Erlang runtime system, started with run_erl.
Usage: to_erl [pipe_name | pipe_dir]
66 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
3.2 Windows NT
Here pipe_name defaults to /tmp/erlang.pipe.N.
To disconnect from the shell without exiting the Erlang system, type Ctrl-D.
start_erl
This program starts the Erlang emulator with parameters -boot and -config set. It reads data about where these
files are located from a file named start_erl.data, which is located in <RELDIR>. Each new release introduces
a new data file. This file is automatically generated by the release handler in Erlang.
The following script illustrates the behaviour of the program:
#!/bin/sh
#
# This program is called by run_erl. It starts
# the Erlang emulator and sets -boot and -config parameters.
# It should only be used at an embedded target system.
#
# Usage: start_erl RootDir RelDir DataFile [ErlFlags ...]
#
ROOTDIR=$1
shift
RELDIR=$1
shift
DataFile=$1
shift
ERTS_VSN=`awk '{print $1}' $DataFile`
VSN=`awk '{print $2}' $DataFile`
BINDIR=$ROOTDIR/erts-$ERTS_VSN/bin
EMU=beam
PROGNAME=`echo $0 | sed 's/.*\///'`
export EMU
export ROOTDIR
export BINDIR
export PROGNAME
export RELDIR
exec $BINDIR/erlexec -boot $RELDIR/$VSN/start -config $RELDIR/$VSN/sys $*
If a diskless and/or read-only client node with the SASL configuration parameter static_emulator set to true
is about to start, the -boot and -config flags must be changed.
As such a client cannot read a new start_erl.data file (the file cannot be changed dynamically). The boot and
config files are always fetched from the same place (but with new contents if a new release has been installed).
The release_handler copies these files to the bin directory in the client directory at the master nodes whenever
a new release is made permanent.
Assuming the same CLIENTDIR as above, the last line is to look like:
exec $BINDIR/erlexec -boot $CLIENTDIR/bin/start \
-config $CLIENTDIR/bin/sys $*
3.2 Windows NT
This section describes the operating system-specific parts of OTP that relate to Windows NT.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 67
3.2 Windows NT
A normal installation of Windows NT 4.0, with Service Pack 4 or later, is required for an embedded Windows NT
running OTP.
3.2.1 Memory Use
RAM memory of 96 MB is recommended to run OTP on Windows NT. A system with less than 64 MB of RAM is
not recommended.
3.2.2 Disk Space Use
A minimum Windows NT installation with networking needs 250 MB, and an extra 130 MB for the swap file.
3.2.3 Installing an Embedded System
Normal Windows NT installation is performed. No additional application programs are needed, such as Internet
Explorer or web server. Networking with TCP/IP is required.
Service Pack 4 or later must be installed.
Hardware Watchdog
For Windows NT running on standard PCs with ISA and/or PCI bus, an extension card with a hardware watchdog
can be installed.
For more information, see the heart(3) manual page in Kernel.
3.2.4 Starting Erlang
On an embedded system, the erlsrv module is to be used to install the Erlang process as a Windows system service.
This service can start after Windows NT has booted.
For more information, see the erlsrv manual page in ERTS.
68 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.1 Introduction
4 Getting Started With Erlang
4.1 Introduction
This section is a quick start tutorial to get you started with Erlang. Everything in this section is true, but only part of
the truth. For example, only the simplest form of the syntax is shown, not all esoteric forms. Also, parts that are greatly
simplified are indicated with *manual*. This means that a lot more information on the subject is to be found in the
Erlang book or in Erlang Reference Manual.
4.1.1 Prerequisites
The reader of this section is assumed to be familiar with the following:
•
•
Computers in general
Basics on how computers are programmed
4.1.2 Omitted Topics
The following topics are not treated in this section:
•
•
•
•
•
•
•
•
•
•
References.
Local error handling (catch/throw).
Single direction links (monitor).
Handling of binary data (binaries / bit syntax).
List comprehensions.
How to communicate with the outside world and software written in other languages (ports); this is described in
Interoperability Tutorial.
Erlang libraries (for example, file handling).
OTP and (in consequence) the Mnesia database.
Hash tables for Erlang terms (ETS).
Changing code in running systems.
4.2 Sequential Programming
4.2.1 The Erlang Shell
Most operating systems have a command interpreter or shell, UNIX and Linux have many, Windows has the command
prompt. Erlang has its own shell where bits of Erlang code can be written directly, and be evaluated to see what happens
(see the shell(3) manual page in STDLIB).
Start the Erlang shell (in Linux or UNIX) by starting a shell or command interpreter in your operating system and
typing erl. You will see something like this.
% erl
Erlang R15B (erts-5.9.1) [source] [smp:8:8] [rq:8] [async-threads:0] [hipe] [kernel-poll:false]
Eshell V5.9.1
(abort with ^G)
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 69
4.2 Sequential Programming
1>
Type "2 + 5." in the shell and then press Enter (carriage return). Notice that you tell the shell you are done entering
code by finishing with a full stop "." and a carriage return.
1> 2 + 5.
7
2>
As shown, the Erlang shell numbers the lines that can be entered, (as 1> 2>) and that it correctly says that 2 + 5 is 7.
If you make writing mistakes in the shell, you can delete with the backspace key, as in most shells. There are many
more editing commands in the shell (see tty - A command line interface in ERTS User's Guide).
(Notice that many line numbers given by the shell in the following examples are out of sequence. This is because this
tutorial was written and code-tested in separate sessions).
Here is a bit more complex calculation:
2> (42 + 77) * 66 / 3.
2618.0
Notice the use of brackets, the multiplication operator "*", and the division operator "/", as in normal arithmetic (see
Expressions).
Press Control-C to shut down the Erlang system and the Erlang shell.
The following output is shown:
BREAK: (a)bort (c)ontinue (p)roc info (i)nfo (l)oaded
(v)ersion (k)ill (D)b-tables (d)istribution
a
%
Type "a" to leave the Erlang system.
Another way to shut down the Erlang system is by entering halt():
3> halt().
%
4.2.2 Modules and Functions
A programming language is not much use if you only can run code from the shell. So here is a small Erlang program.
Enter it into a file named tut.erl using a suitable text editor. The file name tut.erl is important, and also that
it is in the same directory as the one where you started erl). If you are lucky your editor has an Erlang mode that
makes it easier for you to enter and format your code nicely (see The Erlang mode for Emacs in Tools User's Guide),
but you can manage perfectly well without. Here is the code to enter:
-module(tut).
-export([double/1]).
70 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
double(X) ->
2 * X.
It is not hard to guess that this program doubles the value of numbers. The first two lines of the code are described
later. Let us compile the program. This can be done in an Erlang shell as follows, where c means compile:
3> c(tut).
{ok,tut}
The {ok,tut} means that the compilation is OK. If it says "error" it means that there is some mistake in the text
that you entered. Additional error messages gives an idea to what is wrong so you can modify the text and then try
to compile the program again.
Now run the program:
4> tut:double(10).
20
As expected, double of 10 is 20.
Now let us get back to the first two lines of the code. Erlang programs are written in files. Each file contains an Erlang
module. The first line of code in the module is the module name (see Modules):
-module(tut).
Thus, the module is called tut. Notice the full stop "." at the end of the line. The files which are used to store the
module must have the same name as the module but with the extension ".erl". In this case the file name is tut.erl.
When using a function in another module, the syntax module_name:function_name(arguments) is used.
So the following means call function double in module tut with argument "10".
4> tut:double(10).
The second line says that the module tut contains a function called double, which takes one argument (X in our
example):
-export([double/1]).
The second line also says that this function can be called from outside the module tut. More about this later. Again,
notice the "." at the end of the line.
Now for a more complicated example, the factorial of a number. For example, the factorial of 4 is 4 * 3 * 2 * 1,
which equals 24.
Enter the following code in a file named tut1.erl:
-module(tut1).
-export([fac/1]).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 71
4.2 Sequential Programming
fac(1) ->
1;
fac(N) ->
N * fac(N - 1).
So this is a module, called tut1 that contains a function called fac>, which takes one argument, N.
The first part says that the factorial of 1 is 1.:
fac(1) ->
1;
Notice that this part ends with a semicolon ";" that indicates that there is more of the function fac> to come.
The second part says that the factorial of N is N multiplied by the factorial of N - 1:
fac(N) ->
N * fac(N - 1).
Notice that this part ends with a "." saying that there are no more parts of this function.
Compile the file:
5> c(tut1).
{ok,tut1}
And now calculate the factorial of 4.
6> tut1:fac(4).
24
Here the function fac> in module tut1 is called with argument 4.
A function can have many arguments. Let us expand the module tut1 with the function to multiply two numbers:
-module(tut1).
-export([fac/1, mult/2]).
fac(1) ->
1;
fac(N) ->
N * fac(N - 1).
mult(X, Y) ->
X * Y.
Notice that it is also required to expand the -export line with the information that there is another function mult
with two arguments.
Compile:
72 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
7> c(tut1).
{ok,tut1}
Try out the new function mult:
8> tut1:mult(3,4).
12
In this example the numbers are integers and the arguments in the functions in the code N, X, and Y are called variables.
Variables must start with a capital letter (see Variables). Examples of variables are Number, ShoeSize, and Age.
4.2.3 Atoms
Atom is another data type in Erlang. Atoms start with a small letter (see Atom), for example, charles, centimeter,
and inch. Atoms are simply names, nothing else. They are not like variables, which can have a value.
Enter the next program in a file named tut2.erl). It can be useful for converting from inches to centimeters and
conversely:
-module(tut2).
-export([convert/2]).
convert(M, inch) ->
M / 2.54;
convert(N, centimeter) ->
N * 2.54.
Compile:
9> c(tut2).
{ok,tut2}
Test:
10> tut2:convert(3, inch).
1.1811023622047243
11> tut2:convert(7, centimeter).
17.78
Notice the introduction of decimals (floating point numbers) without any explanation. Hopefully you can cope with
that.
Let us see what happens if something other than centimeter or inch is entered in the convert function:
12> tut2:convert(3, miles).
** exception error: no function clause matching tut2:convert(3,miles) (tut2.erl, line 4)
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 73
4.2 Sequential Programming
The two parts of the convert function are called its clauses. As shown, miles is not part of either of the clauses.
The Erlang system cannot match either of the clauses so an error message function_clause is returned. The
shell formats the error message nicely, but the error tuple is saved in the shell's history list and can be output by the
shell command v/1:
13> v(12).
{'EXIT',{function_clause,[{tut2,convert,
[3,miles],
[{file,"tut2.erl"},{line,4}]},
{erl_eval,do_apply,5,[{file,"erl_eval.erl"},{line,482}]},
{shell,exprs,7,[{file,"shell.erl"},{line,666}]},
{shell,eval_exprs,7,[{file,"shell.erl"},{line,621}]},
{shell,eval_loop,3,[{file,"shell.erl"},{line,606}]}]}}
4.2.4 Tuples
Now the tut2 program is hardly good programming style. Consider:
tut2:convert(3, inch).
Does this mean that 3 is in inches? Or does it mean that 3 is in centimeters and is to be converted to inches? Erlang
has a way to group things together to make things more understandable. These are called tuples and are surrounded
by curly brackets, "{" and "}".
So, {inch,3} denotes 3 inches and {centimeter,5} denotes 5 centimeters. Now let us write a new program
that converts centimeters to inches and conversely. Enter the following code in a file called tut3.erl):
-module(tut3).
-export([convert_length/1]).
convert_length({centimeter, X}) ->
{inch, X / 2.54};
convert_length({inch, Y}) ->
{centimeter, Y * 2.54}.
Compile and test:
14> c(tut3).
{ok,tut3}
15> tut3:convert_length({inch, 5}).
{centimeter,12.7}
16> tut3:convert_length(tut3:convert_length({inch, 5})).
{inch,5.0}
Notice on line 16 that 5 inches is converted to centimeters and back again and reassuringly get back to the
original value. That is, the argument to a function can be the result of another function. Consider how line 16
(above) works. The argument given to the function {inch,5} is first matched against the first head clause of
convert_length, that is, convert_length({centimeter,X}). It can be seen that {centimeter,X}
does not match {inch,5} (the head is the bit before the "->"). This having failed, let us try the head of the next
clause that is, convert_length({inch,Y}). This matches, and Y gets the value 5.
74 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
Tuples can have more than two parts, in fact as many parts as you want, and contain any valid Erlang term. For
example, to represent the temperature of various cities of the world:
{moscow, {c, -10}}
{cape_town, {f, 70}}
{paris, {f, 28}}
Tuples have a fixed number of items in them. Each item in a tuple is called an element. In the tuple {moscow,
{c,-10}}, element 1 is moscow and element 2 is {c,-10}. Here c represents Celsius and f Fahrenheit.
4.2.5 Lists
Whereas tuples group things together, it is also needed to represent lists of things. Lists in Erlang are surrounded by
square brackets, "[" and "]". For example, a list of the temperatures of various cities in the world can be:
[{moscow, {c, -10}}, {cape_town, {f, 70}}, {stockholm, {c, -4}},
{paris, {f, 28}}, {london, {f, 36}}]
Notice that this list was so long that it did not fit on one line. This does not matter, Erlang allows line breaks at all
"sensible places" but not, for example, in the middle of atoms, integers, and others.
A useful way of looking at parts of lists, is by using "|". This is best explained by an example using the shell:
17> [First |TheRest] = [1,2,3,4,5].
[1,2,3,4,5]
18> First.
1
19> TheRest.
[2,3,4,5]
To separate the first elements of the list from the rest of the list, | is used. First has got value 1 and TheRest
has got the value [2,3,4,5].
Another example:
20> [E1, E2 | R] = [1,2,3,4,5,6,7].
[1,2,3,4,5,6,7]
21> E1.
1
22> E2.
2
23> R.
[3,4,5,6,7]
Here you see the use of | to get the first two elements from the list. If you try to get more elements from the list than
there are elements in the list, an error is returned. Notice also the special case of the list with no elements, []:
24> [A, B | C] = [1, 2].
[1,2]
25> A.
1
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 75
4.2 Sequential Programming
26> B.
2
27> C.
[]
In the previous examples, new variable names are used, instead of reusing the old ones: First, TheRest, E1, E2,
R, A, B, and C. The reason for this is that a variable can only be given a value once in its context (scope). More about
this later.
The following example shows how to find the length of a list. Enter the following code in a file named tut4.erl):
-module(tut4).
-export([list_length/1]).
list_length([]) ->
0;
list_length([First | Rest]) ->
1 + list_length(Rest).
Compile and test:
28> c(tut4).
{ok,tut4}
29> tut4:list_length([1,2,3,4,5,6,7]).
7
Explanation:
list_length([]) ->
0;
The length of an empty list is obviously 0.
list_length([First | Rest]) ->
1 + list_length(Rest).
The length of a list with the first element First and the remaining elements Rest is 1 + the length of Rest.
(Advanced readers only: This is not tail recursive, there is a better way to write this function.)
In general, tuples are used where "records" or "structs" are used in other languages. Also, lists are used when
representing things with varying sizes, that is, where linked lists are used in other languages.
Erlang does not have a string data type. Instead, strings can be represented by lists of Unicode characters. This implies
for example that the list [97,98,99] is equivalent to "abc". The Erlang shell is "clever" and guesses what list you
mean and outputs it in what it thinks is the most appropriate form, for example:
30> [97,98,99].
"abc"
76 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
4.2.6 Maps
Maps are a set of key to value associations. These associations are encapsulated with "#{" and "}". To create an
association from "key" to value 42:
> #{ "key" => 42 }.
#{"key" => 42}
Let us jump straight into the deep end with an example using some interesting features.
The following example shows how to calculate alpha blending using maps to reference color and alpha channels. Enter
the code in a file named color.erl):
-module(color).
-export([new/4, blend/2]).
-define(is_channel(V), (is_float(V) andalso V >= 0.0 andalso V =< 1.0)).
new(R,G,B,A) when ?is_channel(R), ?is_channel(G),
?is_channel(B), ?is_channel(A) ->
#{red => R, green => G, blue => B, alpha => A}.
blend(Src,Dst) ->
blend(Src,Dst,alpha(Src,Dst)).
blend(Src,Dst,Alpha) when Alpha > 0.0 ->
Dst#{
red
:= red(Src,Dst) / Alpha,
green := green(Src,Dst) / Alpha,
blue := blue(Src,Dst) / Alpha,
alpha := Alpha
};
blend(_,Dst,_) ->
Dst#{
red
:= 0.0,
green := 0.0,
blue := 0.0,
alpha := 0.0
}.
alpha(#{alpha := SA}, #{alpha := DA}) ->
SA + DA*(1.0 - SA).
red(#{red := SV, alpha := SA}, #{red := DV, alpha := DA}) ->
SV*SA + DV*DA*(1.0 - SA).
green(#{green := SV, alpha := SA}, #{green := DV, alpha := DA}) ->
SV*SA + DV*DA*(1.0 - SA).
blue(#{blue := SV, alpha := SA}, #{blue := DV, alpha := DA}) ->
SV*SA + DV*DA*(1.0 - SA).
Compile and test:
> c(color).
{ok,color}
> C1 = color:new(0.3,0.4,0.5,1.0).
#{alpha => 1.0,blue => 0.5,green => 0.4,red => 0.3}
> C2 = color:new(1.0,0.8,0.1,0.3).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 77
4.2 Sequential Programming
#{alpha => 0.3,blue => 0.1,green => 0.8,red => 1.0}
> color:blend(C1,C2).
#{alpha => 1.0,blue => 0.5,green => 0.4,red => 0.3}
> color:blend(C2,C1).
#{alpha => 1.0,blue => 0.38,green => 0.52,red => 0.51}
This example warrants some explanation:
-define(is_channel(V), (is_float(V) andalso V >= 0.0 andalso V =< 1.0)).
First a macro is_channel is defined to help with the guard tests. This is only here for convenience and to reduce
syntax cluttering. For more information about macros, see The Preprocessor.
new(R,G,B,A) when ?is_channel(R), ?is_channel(G),
?is_channel(B), ?is_channel(A) ->
#{red => R, green => G, blue => B, alpha => A}.
The function new/4 creates a new map term and lets the keys red, green, blue, and alpha be associated with
an initial value. In this case, only float values between and including 0.0 and 1.0 are allowed, as ensured by the ?
is_channel/1 macro for each argument. Only the => operator is allowed when creating a new map.
By calling blend/2 on any color term created by new/4, the resulting color can be calculated as determined by
the two map terms.
The first thing blend/2 does is to calculate the resulting alpha channel:
alpha(#{alpha := SA}, #{alpha := DA}) ->
SA + DA*(1.0 - SA).
The value associated with key alpha is fetched for both arguments using the := operator. The other keys in the map
are ignored, only the key alpha is required and checked for.
This is also the case for functions red/2, blue/2, and green/2.
red(#{red := SV, alpha := SA}, #{red := DV, alpha := DA}) ->
SV*SA + DV*DA*(1.0 - SA).
The difference here is that a check is made for two keys in each map argument. The other keys are ignored.
Finally, let us return the resulting color in blend/3:
blend(Src,Dst,Alpha) when Alpha > 0.0 ->
Dst#{
red
:= red(Src,Dst) / Alpha,
green := green(Src,Dst) / Alpha,
blue := blue(Src,Dst) / Alpha,
alpha := Alpha
};
The Dst map is updated with new channel values. The syntax for updating an existing key with a new value is with
the := operator.
78 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
4.2.7 Standard Modules and Manual Pages
Erlang has many standard modules to help you do things. For example, the module io contains many functions that
help in doing formatted input/output. To look up information about standard modules, the command erl -man can
be used at the operating shell or command prompt (the same place as you started erl). Try the operating system
shell command:
% erl -man io
ERLANG MODULE DEFINITION
io(3)
MODULE
io - Standard I/O Server Interface Functions
DESCRIPTION
This module provides an interface to standard Erlang IO
servers. The output functions all return ok if they are suc...
If this does not work on your system, the documentation is included as HTML in the Erlang/OTP release. You can also
read the documentation as HTML or download it as PDF from either of the sites www.erlang.se (commercial Erlang)
or www.erlang.org (open source). For example, for Erlang/OTP release R9B:
http://www.erlang.org/doc/r9b/doc/index.html
4.2.8 Writing Output to a Terminal
It is nice to be able to do formatted output in examples, so the next example shows a simple way to use the io:format
function. Like all other exported functions, you can test the io:format function in the shell:
31> io:format("hello world~n", []).
hello world
ok
32> io:format("this outputs one Erlang term: ~w~n", [hello]).
this outputs one Erlang term: hello
ok
33> io:format("this outputs two Erlang terms: ~w~w~n", [hello, world]).
this outputs two Erlang terms: helloworld
ok
34> io:format("this outputs two Erlang terms: ~w ~w~n", [hello, world]).
this outputs two Erlang terms: hello world
ok
The function format/2 (that is, format with two arguments) takes two lists. The first one is nearly always a list
written between " ". This list is printed out as it is, except that each ~w is replaced by a term taken in order from the
second list. Each ~n is replaced by a new line. The io:format/2 function itself returns the atom ok if everything
goes as planned. Like other functions in Erlang, it crashes if an error occurs. This is not a fault in Erlang, it is a
deliberate policy. Erlang has sophisticated mechanisms to handle errors which are shown later. As an exercise, try to
make io:format crash, it should not be difficult. But notice that although io:format crashes, the Erlang shell
itself does not crash.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 79
4.2 Sequential Programming
4.2.9 A Larger Example
Now for a larger example to consolidate what you have learnt so far. Assume that you have a list of temperature
readings from a number of cities in the world. Some of them are in Celsius and some in Fahrenheit (as in the previous
list). First let us convert them all to Celsius, then let us print the data neatly.
%% This module is in file tut5.erl
-module(tut5).
-export([format_temps/1]).
%% Only this function is exported
format_temps([])->
% No output for an empty list
ok;
format_temps([City | Rest]) ->
print_temp(convert_to_celsius(City)),
format_temps(Rest).
convert_to_celsius({Name, {c, Temp}}) ->
{Name, {c, Temp}};
convert_to_celsius({Name, {f, Temp}}) ->
{Name, {c, (Temp - 32) * 5 / 9}}.
% No conversion needed
% Do the conversion
print_temp({Name, {c, Temp}}) ->
io:format("~-15w ~w c~n", [Name, Temp]).
35> c(tut5).
{ok,tut5}
36> tut5:format_temps([{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
moscow
-10 c
cape_town
21.11111111111111 c
stockholm
-4 c
paris
-2.2222222222222223 c
london
2.2222222222222223 c
ok
Before looking at how this program works, notice that a few comments are added to the code. A comment starts with
a %-character and goes on to the end of the line. Notice also that the -export([format_temps/1]). line only
includes the function format_temps/1. The other functions are local functions, that is, they are not visible from
outside the module tut5.
Notice also that when testing the program from the shell, the input is spread over two lines as the line was too long.
When format_temps is called the first time, City gets the value {moscow,{c,-10}} and Rest is the rest of
the list. So the function print_temp(convert_to_celsius({moscow,{c,-10}})) is called.
Here is a function call as convert_to_celsius({moscow,{c,-10}}) as the argument to the function
print_temp. When function calls are nested like this, they execute (evaluate) from the inside out. That is, first
convert_to_celsius({moscow,{c,-10}}) is evaluated, which gives the value {moscow,{c,-10}}
as the temperature is already in Celsius. Then print_temp({moscow,{c,-10}}) is evaluated. The function
convert_to_celsius works in a similar way to the convert_length function in the previous example.
print_temp simply calls io:format in a similar way to what has been described above. Notice that ~-15w says
to print the "term" with a field length (width) of 15 and left justify it. (see the io(3)) manual page in STDLIB.
Now format_temps(Rest) is called with the rest of the list as an argument. This way of doing things is
similar to the loop constructs in other languages. (Yes, this is recursion, but do not let that worry you.) So the same
80 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
format_temps function is called again, this time City gets the value {cape_town,{f,70}} and the same
procedure is repeated as before. This is done until the list becomes empty, that is [], which causes the first clause
format_temps([]) to match. This simply returns (results in) the atom ok, so the program ends.
4.2.10 Matching, Guards, and Scope of Variables
It can be useful to find the maximum and minimum temperature in lists like this. Before extending the program to do
this, let us look at functions for finding the maximum value of the elements in a list:
-module(tut6).
-export([list_max/1]).
list_max([Head|Rest]) ->
list_max(Rest, Head).
list_max([], Res) ->
Res;
list_max([Head|Rest], Result_so_far) when Head > Result_so_far ->
list_max(Rest, Head);
list_max([Head|Rest], Result_so_far) ->
list_max(Rest, Result_so_far).
37> c(tut6).
{ok,tut6}
38> tut6:list_max([1,2,3,4,5,7,4,3,2,1]).
7
First notice that two functions have the same name, list_max. However, each of these takes a different number of
arguments (parameters). In Erlang these are regarded as completely different functions. Where you need to distinguish
between these functions, you write Name/Arity, where Name is the function name and Arity is the number of
arguments, in this case list_max/1 and list_max/2.
In this example you walk through a list "carrying" a value, in this case Result_so_far. list_max/1 simply
assumes that the max value of the list is the head of the list and calls list_max/2 with the rest of the list and the
value of the head of the list. In the above this would be list_max([2,3,4,5,7,4,3,2,1],1). If you tried
to use list_max/1 with an empty list or tried to use it with something that is not a list at all, you would cause an
error. Notice that the Erlang philosophy is not to handle errors of this type in the function they occur, but to do so
elsewhere. More about this later.
In list_max/2, you walk down the list and use Head instead of Result_so_far when Head >
Result_so_far. when is a special word used before the -> in the function to say that you only use this part of
the function if the test that follows is true. A test of this type is called guard. If the guard is false (that is, the guard
fails), the next part of the function is tried. In this case, if Head is not greater than Result_so_far, then it must
be smaller or equal to it. This means that a guard on the next part of the function is not needed.
Some useful operators in guards are:
•
•
•
•
•
•
< less than
> greater than
== equal
>= greater or equal
=< less or equal
/= not equal
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 81
4.2 Sequential Programming
(see Guard Sequences).
To change the above program to one that works out the minimum value of the element in a list, you only need to write
< instead of >. (But it would be wise to change the name of the function to list_min.)
Earlier it was mentioned that a variable can only be given a value once in its scope. In the above you see that
Result_so_far is given several values. This is OK since every time you call list_max/2 you create a new
scope and one can regard Result_so_far as a different variable in each scope.
Another way of creating and giving a variable a value is by using the match operator = . So if you write M = 5, a
variable called M is created with the value 5. If, in the same scope, you then write M = 6, an error is returned. Try
this out in the shell:
39> M = 5.
5
40> M = 6.
** exception error: no match of right hand side value 6
41> M = M + 1.
** exception error: no match of right hand side value 6
42> N = M + 1.
6
The use of the match operator is particularly useful for pulling apart Erlang terms and creating new ones.
43> {X, Y} = {paris, {f, 28}}.
{paris,{f,28}}
44> X.
paris
45> Y.
{f,28}
Here X gets the value paris and Y{f,28}.
If you try to do the same again with another city, an error is returned:
46> {X, Y} = {london, {f, 36}}.
** exception error: no match of right hand side value {london,{f,36}}
Variables can also be used to improve the readability of programs. For example, in function list_max/2 above,
you can write:
list_max([Head|Rest], Result_so_far) when Head > Result_so_far ->
New_result_far = Head,
list_max(Rest, New_result_far);
This is possibly a little clearer.
4.2.11 More About Lists
Remember that the | operator can be used to get the head of a list:
47> [M1|T1] = [paris, london, rome].
82 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
[paris,london,rome]
48> M1.
paris
49> T1.
[london,rome]
The | operator can also be used to add a head to a list:
50> L1 = [madrid | T1].
[madrid,london,rome]
51> L1.
[madrid,london,rome]
Now an example of this when working with lists - reversing the order of a list:
-module(tut8).
-export([reverse/1]).
reverse(List) ->
reverse(List, []).
reverse([Head | Rest], Reversed_List) ->
reverse(Rest, [Head | Reversed_List]);
reverse([], Reversed_List) ->
Reversed_List.
52> c(tut8).
{ok,tut8}
53> tut8:reverse([1,2,3]).
[3,2,1]
Consider how Reversed_List is built. It starts as [], then successively the heads are taken off of the list to be
reversed and added to the the Reversed_List, as shown in the following:
reverse([1|2,3], []) =>
reverse([2,3], [1|[]])
reverse([2|3], [1]) =>
reverse([3], [2|[1])
reverse([3|[]], [2,1]) =>
reverse([], [3|[2,1]])
reverse([], [3,2,1]) =>
[3,2,1]
The module lists contains many functions for manipulating lists, for example, for reversing them. So before writing
a list-manipulating function it is a good idea to check if one not already is written for you (see the lists(3) manual
page in STDLIB).
Now let us get back to the cities and temperatures, but take a more structured approach this time. First let us convert
the whole list to Celsius as follows:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 83
4.2 Sequential Programming
-module(tut7).
-export([format_temps/1]).
format_temps(List_of_cities) ->
convert_list_to_c(List_of_cities).
convert_list_to_c([{Name, {f, F}} | Rest]) ->
Converted_City = {Name, {c, (F -32)* 5 / 9}},
[Converted_City | convert_list_to_c(Rest)];
convert_list_to_c([City | Rest]) ->
[City | convert_list_to_c(Rest)];
convert_list_to_c([]) ->
[].
Test the function:
54> c(tut7).
{ok, tut7}.
55> tut7:format_temps([{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
[{moscow,{c,-10}},
{cape_town,{c,21.11111111111111}},
{stockholm,{c,-4}},
{paris,{c,-2.2222222222222223}},
{london,{c,2.2222222222222223}}]
Explanation:
format_temps(List_of_cities) ->
convert_list_to_c(List_of_cities).
Here format_temps/1 calls convert_list_to_c/1. convert_list_to_c/1 takes off the head of the
List_of_cities, converts it to Celsius if needed. The | operator is used to add the (maybe) converted to the
converted rest of the list:
[Converted_City | convert_list_to_c(Rest)];
or:
[City | convert_list_to_c(Rest)];
This is done until the end of the list is reached, that is, the list is empty:
convert_list_to_c([]) ->
[].
Now when the list is converted, a function to print it is added:
84 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
-module(tut7).
-export([format_temps/1]).
format_temps(List_of_cities) ->
Converted_List = convert_list_to_c(List_of_cities),
print_temp(Converted_List).
convert_list_to_c([{Name, {f, F}} | Rest]) ->
Converted_City = {Name, {c, (F -32)* 5 / 9}},
[Converted_City | convert_list_to_c(Rest)];
convert_list_to_c([City | Rest]) ->
[City | convert_list_to_c(Rest)];
convert_list_to_c([]) ->
[].
print_temp([{Name, {c, Temp}} | Rest]) ->
io:format("~-15w ~w c~n", [Name, Temp]),
print_temp(Rest);
print_temp([]) ->
ok.
56> c(tut7).
{ok,tut7}
57> tut7:format_temps([{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
moscow
-10 c
cape_town
21.11111111111111 c
stockholm
-4 c
paris
-2.2222222222222223 c
london
2.2222222222222223 c
ok
Now a function has to be added to find the cities with the maximum and minimum temperatures. The following program
is not the most efficient way of doing this as you walk through the list of cities four times. But it is better to first strive
for clarity and correctness and to make programs efficient only if needed.
-module(tut7).
-export([format_temps/1]).
format_temps(List_of_cities) ->
Converted_List = convert_list_to_c(List_of_cities),
print_temp(Converted_List),
{Max_city, Min_city} = find_max_and_min(Converted_List),
print_max_and_min(Max_city, Min_city).
convert_list_to_c([{Name, {f, Temp}} | Rest]) ->
Converted_City = {Name, {c, (Temp -32)* 5 / 9}},
[Converted_City | convert_list_to_c(Rest)];
convert_list_to_c([City | Rest]) ->
[City | convert_list_to_c(Rest)];
convert_list_to_c([]) ->
[].
print_temp([{Name, {c, Temp}} | Rest]) ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 85
4.2 Sequential Programming
io:format("~-15w ~w c~n", [Name, Temp]),
print_temp(Rest);
print_temp([]) ->
ok.
find_max_and_min([City | Rest]) ->
find_max_and_min(Rest, City, City).
find_max_and_min([{Name, {c, Temp}} | Rest],
{Max_Name, {c, Max_Temp}},
{Min_Name, {c, Min_Temp}}) ->
if
Temp > Max_Temp ->
Max_City = {Name, {c, Temp}};
% Change
true ->
Max_City = {Max_Name, {c, Max_Temp}} % Unchanged
end,
if
Temp < Min_Temp ->
Min_City = {Name, {c, Temp}};
% Change
true ->
Min_City = {Min_Name, {c, Min_Temp}} % Unchanged
end,
find_max_and_min(Rest, Max_City, Min_City);
find_max_and_min([], Max_City, Min_City) ->
{Max_City, Min_City}.
print_max_and_min({Max_name, {c, Max_temp}}, {Min_name, {c, Min_temp}}) ->
io:format("Max temperature was ~w c in ~w~n", [Max_temp, Max_name]),
io:format("Min temperature was ~w c in ~w~n", [Min_temp, Min_name]).
58> c(tut7).
{ok, tut7}
59> tut7:format_temps([{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
moscow
-10 c
cape_town
21.11111111111111 c
stockholm
-4 c
paris
-2.2222222222222223 c
london
2.2222222222222223 c
Max temperature was 21.11111111111111 c in cape_town
Min temperature was -10 c in moscow
ok
4.2.12 If and Case
The function find_max_and_min works out the maximum and minimum temperature. A new construct, if, is
introduced here. If works as follows:
if
Condition 1 ->
Action 1;
Condition 2 ->
Action 2;
Condition 3 ->
Action 3;
Condition 4 ->
Action 4
86 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
end
Notice that there is no ";" before end. Conditions do the same as guards, that is, tests that succeed or fail. Erlang
starts at the top and tests until it finds a condition that succeeds. Then it evaluates (performs) the action following
the condition and ignores all other conditions and actions before the end. If no condition matches, a run-time failure
occurs. A condition that always succeeds is the atom true. This is often used last in an if, meaning, do the action
following the true if all other conditions have failed.
The following is a short program to show the workings of if.
-module(tut9).
-export([test_if/2]).
test_if(A, B) ->
if
A == 5 ->
io:format("A == 5~n", []),
a_equals_5;
B == 6 ->
io:format("B == 6~n", []),
b_equals_6;
A == 2, B == 3 ->
%That is A equals 2 and B equals 3
io:format("A == 2, B == 3~n", []),
a_equals_2_b_equals_3;
A == 1 ; B == 7 ->
%That is A equals 1 or B equals 7
io:format("A == 1 ; B == 7~n", []),
a_equals_1_or_b_equals_7
end.
Testing this program gives:
60> c(tut9).
{ok,tut9}
61> tut9:test_if(5,33).
A == 5
a_equals_5
62> tut9:test_if(33,6).
B == 6
b_equals_6
63> tut9:test_if(2, 3).
A == 2, B == 3
a_equals_2_b_equals_3
64> tut9:test_if(1, 33).
A == 1 ; B == 7
a_equals_1_or_b_equals_7
65> tut9:test_if(33, 7).
A == 1 ; B == 7
a_equals_1_or_b_equals_7
66> tut9:test_if(33, 33).
** exception error: no true branch found when evaluating an if expression
in function tut9:test_if/2 (tut9.erl, line 5)
Notice that tut9:test_if(33,33) does not cause any condition to succeed. This leads to the run time error
if_clause, here nicely formatted by the shell. See Guard Sequences for details of the many guard tests available.
case is another construct in Erlang. Recall that the convert_length function was written as:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 87
4.2 Sequential Programming
convert_length({centimeter, X}) ->
{inch, X / 2.54};
convert_length({inch, Y}) ->
{centimeter, Y * 2.54}.
The same program can also be written as:
-module(tut10).
-export([convert_length/1]).
convert_length(Length) ->
case Length of
{centimeter, X} ->
{inch, X / 2.54};
{inch, Y} ->
{centimeter, Y * 2.54}
end.
67> c(tut10).
{ok,tut10}
68> tut10:convert_length({inch, 6}).
{centimeter,15.24}
69> tut10:convert_length({centimeter, 2.5}).
{inch,0.984251968503937}
Both case and if have return values, that is, in the above example case returned either {inch,X/2.54} or
{centimeter,Y*2.54}. The behaviour of case can also be modified by using guards. The following example
clarifies this. It tells us the length of a month, given the year. The year must be known, since February has 29 days
in a leap year.
-module(tut11).
-export([month_length/2]).
month_length(Year, Month) ->
%% All years divisible by 400 are leap
%% Years divisible by 100 are not leap (except the 400 rule above)
%% Years divisible by 4 are leap (except the 100 rule above)
Leap = if
trunc(Year / 400) * 400 == Year ->
leap;
trunc(Year / 100) * 100 == Year ->
not_leap;
trunc(Year / 4) * 4 == Year ->
leap;
true ->
not_leap
end,
case Month of
sep -> 30;
apr -> 30;
jun -> 30;
nov -> 30;
feb when Leap == leap -> 29;
feb -> 28;
jan -> 31;
mar -> 31;
may -> 31;
88 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
jul
aug
oct
dec
end.
->
->
->
->
31;
31;
31;
31
70> c(tut11).
{ok,tut11}
71> tut11:month_length(2004, feb).
29
72> tut11:month_length(2003, feb).
28
73> tut11:month_length(1947, aug).
31
4.2.13 Built-In Functions (BIFs)
BIFs are functions that for some reason are built-in to the Erlang virtual machine. BIFs often implement functionality
that is impossible or is too inefficient to implement in Erlang. Some BIFs can be called using the function name only
but they are by default belonging to the erlang module. For example, the call to the BIF trunc below is equivalent
to a call to erlang:trunc.
As shown, first it is checked if a year is leap. If a year is divisible by 400, it is a leap year. To determine this, first
divide the year by 400 and use the BIF trunc (more about this later) to cut off any decimals. Then multiply by 400
again and see if the same value is returned again. For example, year 2004:
2004 / 400 = 5.01
trunc(5.01) = 5
5 * 400 = 2000
2000 is not the same as 2004, so 2004 is not divisible by 400. Year 2000:
2000 / 400 = 5.0
trunc(5.0) = 5
5 * 400 = 2000
That is, a leap year. The next two trunc-tests evaluate if the year is divisible by 100 or 4 in the same way. The first
if returns leap or not_leap, which lands up in the variable Leap. This variable is used in the guard for feb in
the following case that tells us how long the month is.
This example showed the use of trunc. It is easier to use the Erlang operator rem that gives the remainder after
division, for example:
74> 2004 rem 400.
4
So instead of writing:
trunc(Year / 400) * 400 == Year ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 89
4.2 Sequential Programming
leap;
it can be written:
Year rem 400 == 0 ->
leap;
There are many other BIFs such as trunc. Only a few BIFs can be used in guards, and you cannot use functions you
have defined yourself in guards. (see Guard Sequences) (For advanced readers: This is to ensure that guards do not
have side effects.) Let us play with a few of these functions in the shell:
75> trunc(5.6).
5
76> round(5.6).
6
77> length([a,b,c,d]).
4
78> float(5).
5.0
79> is_atom(hello).
true
80> is_atom("hello").
false
81> is_tuple({paris, {c, 30}}).
true
82> is_tuple([paris, {c, 30}]).
false
All of these can be used in guards. Now for some BIFs that cannot be used in guards:
83> atom_to_list(hello).
"hello"
84> list_to_atom("goodbye").
goodbye
85> integer_to_list(22).
"22"
These three BIFs do conversions that would be difficult (or impossible) to do in Erlang.
4.2.14 Higher-Order Functions (Funs)
Erlang, like most modern functional programming languages, has higher-order functions. Here is an example using
the shell:
86> Xf = fun(X) -> X * 2 end.
#Fun<erl_eval.5.123085357>
87> Xf(5).
10
Here is defined a function that doubles the value of a number and assigned this function to a variable. Thus Xf(5)
returns value 10. Two useful functions when working with lists are foreach and map, which are defined as follows:
90 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.2 Sequential Programming
foreach(Fun, [First|Rest]) ->
Fun(First),
foreach(Fun, Rest);
foreach(Fun, []) ->
ok.
map(Fun, [First|Rest]) ->
[Fun(First)|map(Fun,Rest)];
map(Fun, []) ->
[].
These two functions are provided in the standard module lists. foreach takes a list and applies a fun to every
element in the list. map creates a new list by applying a fun to every element in a list. Going back to the shell, map
is used and a fun to add 3 to every element of a list:
88> Add_3 = fun(X) -> X + 3 end.
#Fun<erl_eval.5.123085357>
89> lists:map(Add_3, [1,2,3]).
[4,5,6]
Let us (again) print the temperatures in a list of cities:
90> Print_City = fun({City, {X, Temp}}) -> io:format("~-15w ~w ~w~n",
[City, X, Temp]) end.
#Fun<erl_eval.5.123085357>
91> lists:foreach(Print_City, [{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
moscow
c -10
cape_town
f 70
stockholm
c -4
paris
f 28
london
f 36
ok
Let us now define a fun that can be used to go through a list of cities and temperatures and transform them all to Celsius.
-module(tut13).
-export([convert_list_to_c/1]).
convert_to_c({Name, {f, Temp}}) ->
{Name, {c, trunc((Temp - 32) * 5 / 9)}};
convert_to_c({Name, {c, Temp}}) ->
{Name, {c, Temp}}.
convert_list_to_c(List) ->
lists:map(fun convert_to_c/1, List).
92> tut13:convert_list_to_c([{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
[{moscow,{c,-10}},
{cape_town,{c,21}},
{stockholm,{c,-4}},
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 91
4.2 Sequential Programming
{paris,{c,-2}},
{london,{c,2}}]
The convert_to_c function is the same as before, but here it is used as a fun:
lists:map(fun convert_to_c/1, List)
When a function defined elsewhere is used as a fun, it can be referred to as Function/Arity (remember that
Arity = number of arguments). So in the map-call lists:map(fun convert_to_c/1, List) is written.
As shown, convert_list_to_c becomes much shorter and easier to understand.
The standard module lists also contains a function sort(Fun, List) where Fun is a fun with two arguments.
This fun returns true if the first argument is less than the second argument, or else false. Sorting is added to the
convert_list_to_c:
-module(tut13).
-export([convert_list_to_c/1]).
convert_to_c({Name, {f, Temp}}) ->
{Name, {c, trunc((Temp - 32) * 5 / 9)}};
convert_to_c({Name, {c, Temp}}) ->
{Name, {c, Temp}}.
convert_list_to_c(List) ->
New_list = lists:map(fun convert_to_c/1, List),
lists:sort(fun({_, {c, Temp1}}, {_, {c, Temp2}}) ->
Temp1 < Temp2 end, New_list).
93> c(tut13).
{ok,tut13}
94> tut13:convert_list_to_c([{moscow, {c, -10}}, {cape_town, {f, 70}},
{stockholm, {c, -4}}, {paris, {f, 28}}, {london, {f, 36}}]).
[{moscow,{c,-10}},
{stockholm,{c,-4}},
{paris,{c,-2}},
{london,{c,2}},
{cape_town,{c,21}}]
In sort the fun is used:
fun({_, {c, Temp1}}, {_, {c, Temp2}}) -> Temp1 < Temp2 end,
Here the concept of an anonymous variable "_" is introduced. This is simply shorthand for a variable that gets a
value, but the value is ignored. This can be used anywhere suitable, not just in funs. Temp1 < Temp2 returns true
if Temp1 is less than Temp2.
92 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
4.3 Concurrent Programming
4.3.1 Processes
One of the main reasons for using Erlang instead of other functional languages is Erlang's ability to handle concurrency
and distributed programming. By concurrency is meant programs that can handle several threads of execution at the
same time. For example, modern operating systems allow you to use a word processor, a spreadsheet, a mail client,
and a print job all running at the same time. Each processor (CPU) in the system is probably only handling one thread
(or job) at a time, but it swaps between the jobs at such a rate that it gives the illusion of running them all at the same
time. It is easy to create parallel threads of execution in an Erlang program and to allow these threads to communicate
with each other. In Erlang, each thread of execution is called a process.
(Aside: the term "process" is usually used when the threads of execution share no data with each other and the term
"thread" when they share data in some way. Threads of execution in Erlang share no data, that is why they are called
processes).
The Erlang BIF spawn is used to create a new process: spawn(Module, Exported_Function, List of
Arguments). Consider the following module:
-module(tut14).
-export([start/0, say_something/2]).
say_something(What, 0) ->
done;
say_something(What, Times) ->
io:format("~p~n", [What]),
say_something(What, Times - 1).
start() ->
spawn(tut14, say_something, [hello, 3]),
spawn(tut14, say_something, [goodbye, 3]).
5> c(tut14).
{ok,tut14}
6> tut14:say_something(hello, 3).
hello
hello
hello
done
As shown, the function say_something writes its first argument the number of times specified by second argument.
The function start starts two Erlang processes, one that writes "hello" three times and one that writes "goodbye"
three times. Both processes use the function say_something. Notice that a function used in this way by spawn,
to start a process, must be exported from the module (that is, in the -export at the start of the module).
9> tut14:start().
hello
goodbye
<0.63.0>
hello
goodbye
hello
goodbye
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 93
4.3 Concurrent Programming
Notice that it did not write "hello" three times and then "goodbye" three times. Instead, the first process wrote a "hello",
the second a "goodbye", the first another "hello" and so forth. But where did the <0.63.0> come from? The return
value of a function is the return value of the last "thing" in the function. The last thing in the function start is
spawn(tut14, say_something, [goodbye, 3]).
spawn returns a process identifier, or pid, which uniquely identifies the process. So <0.63.0> is the pid of the spawn
function call above. The next example shows how to use pids.
Notice also that ~p is used instead of ~w in io:format. To quote the manual: "~p Writes the data with standard
syntax in the same way as ~w, but breaks terms whose printed representation is longer than one line into many lines
and indents each line sensibly. It also tries to detect lists of printable characters and to output these as strings".
4.3.2 Message Passing
In the following example two processes are created and they send messages to each other a number of times.
-module(tut15).
-export([start/0, ping/2, pong/0]).
ping(0, Pong_PID) ->
Pong_PID ! finished,
io:format("ping finished~n", []);
ping(N, Pong_PID) ->
Pong_PID ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping(N - 1, Pong_PID).
pong() ->
receive
finished ->
io:format("Pong finished~n", []);
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
end.
start() ->
Pong_PID = spawn(tut15, pong, []),
spawn(tut15, ping, [3, Pong_PID]).
1> c(tut15).
{ok,tut15}
2> tut15: start().
<0.36.0>
Pong received ping
Ping received pong
Pong received ping
Ping received pong
Pong received ping
94 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
Ping received pong
ping finished
Pong finished
The function start first creates a process, let us call it "pong":
Pong_PID = spawn(tut15, pong, [])
This process executes tut15:pong(). Pong_PID is the process identity of the "pong" process. The function
start now creates another process "ping":
spawn(tut15, ping, [3, Pong_PID]),
This process executes:
tut15:ping(3, Pong_PID)
<0.36.0> is the return value from the start function.
The process "pong" now does:
receive
finished ->
io:format("Pong finished~n", []);
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
end.
The receive construct is used to allow processes to wait for messages from other processes. It has the following
format:
receive
pattern1 ->
actions1;
pattern2 ->
actions2;
....
patternN
actionsN
end.
Notice there is no ";" before the end.
Messages between Erlang processes are simply valid Erlang terms. That is, they can be lists, tuples, integers, atoms,
pids, and so on.
Each process has its own input queue for messages it receives. New messages received are put at the end of the
queue. When a process executes a receive, the first message in the queue is matched against the first pattern in
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 95
4.3 Concurrent Programming
the receive. If this matches, the message is removed from the queue and the actions corresponding to the pattern
are executed.
However, if the first pattern does not match, the second pattern is tested. If this matches, the message is removed from
the queue and the actions corresponding to the second pattern are executed. If the second pattern does not match, the
third is tried and so on until there are no more patterns to test. If there are no more patterns to test, the first message
is kept in the queue and the second message is tried instead. If this matches any pattern, the appropriate actions are
executed and the second message is removed from the queue (keeping the first message and any other messages in
the queue). If the second message does not match, the third message is tried, and so on, until the end of the queue
is reached. If the end of the queue is reached, the process blocks (stops execution) and waits until a new message is
received and this procedure is repeated.
The Erlang implementation is "clever" and minimizes the number of times each message is tested against the patterns
in each receive.
Now back to the ping pong example.
"Pong" is waiting for messages. If the atom finished is received, "pong" writes "Pong finished" to the output and,
as it has nothing more to do, terminates. If it receives a message with the format:
{ping, Ping_PID}
it writes "Pong received ping" to the output and sends the atom pong to the process "ping":
Ping_PID ! pong
Notice how the operator "!" is used to send messages. The syntax of "!" is:
Pid ! Message
That is, Message (any Erlang term) is sent to the process with identity Pid.
After sending the message pong to the process "ping", "pong" calls the pong function again, which causes it to get
back to the receive again and wait for another message.
Now let us look at the process "ping". Recall that it was started by executing:
tut15:ping(3, Pong_PID)
Looking at the function ping/2, the second clause of ping/2 is executed since the value of the first argument is 3
(not 0) (first clause head is ping(0,Pong_PID), second clause head is ping(N,Pong_PID), so N becomes 3).
The second clause sends a message to "pong":
Pong_PID ! {ping, self()},
self() returns the pid of the process that executes self(), in this case the pid of "ping". (Recall the code for
"pong", this lands up in the variable Ping_PID in the receive previously explained.)
"Ping" now waits for a reply from "pong":
96 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
receive
pong ->
io:format("Ping received pong~n", [])
end,
It writes "Ping received pong" when this reply arrives, after which "ping" calls the ping function again.
ping(N - 1, Pong_PID)
N-1 causes the first argument to be decremented until it becomes 0. When this occurs, the first clause of ping/2
is executed:
ping(0, Pong_PID) ->
Pong_PID ! finished,
io:format("ping finished~n", []);
The atom finished is sent to "pong" (causing it to terminate as described above) and "ping finished" is written to
the output. "Ping" then terminates as it has nothing left to do.
4.3.3 Registered Process Names
In the above example, "pong" was first created to be able to give the identity of "pong" when "ping" was started. That
is, in some way "ping" must be able to know the identity of "pong" to be able to send a message to it. Sometimes
processes which need to know each other's identities are started independently of each other. Erlang thus provides a
mechanism for processes to be given names so that these names can be used as identities instead of pids. This is done
by using the register BIF:
register(some_atom, Pid)
Let us now rewrite the ping pong example using this and give the name pong to the "pong" process:
-module(tut16).
-export([start/0, ping/1, pong/0]).
ping(0) ->
pong ! finished,
io:format("ping finished~n", []);
ping(N) ->
pong ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping(N - 1).
pong() ->
receive
finished ->
io:format("Pong finished~n", []);
{ping, Ping_PID} ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 97
4.3 Concurrent Programming
end.
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
start() ->
register(pong, spawn(tut16, pong, [])),
spawn(tut16, ping, [3]).
2> c(tut16).
{ok, tut16}
3> tut16:start().
<0.38.0>
Pong received ping
Ping received pong
Pong received ping
Ping received pong
Pong received ping
Ping received pong
ping finished
Pong finished
Here the start/0 function,
register(pong, spawn(tut16, pong, [])),
both spawns the "pong" process and gives it the name pong. In the "ping" process, messages can be sent to pong by:
pong ! {ping, self()},
ping/2 now becomes ping/1 as the argument Pong_PID is not needed.
4.3.4 Distributed Programming
Let us rewrite the ping pong program with "ping" and "pong" on different computers. First a few things are needed to
set up to get this to work. The distributed Erlang implementation provides a very basic authentication mechanism to
prevent unintentional access to an Erlang system on another computer. Erlang systems which talk to each other must
have the same magic cookie. The easiest way to achieve this is by having a file called .erlang.cookie in your
home directory on all machines on which you are going to run Erlang systems communicating with each other:
•
•
On Windows systems the home directory is the directory pointed out by the environment variable $HOME you may need to set this.
On Linux or UNIX you can safely ignore this and simply create a file called .erlang.cookie in the
directory you get to after executing the command cd without any argument.
The .erlang.cookie file is to contain a line with the same atom. For example, on Linux or UNIX, in the OS shell:
$ cd
$ cat > .erlang.cookie
this_is_very_secret
$ chmod 400 .erlang.cookie
98 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
The chmod above makes the .erlang.cookie file accessible only by the owner of the file. This is a requirement.
When you start an Erlang system that is going to talk to other Erlang systems, you must give it a name, for example:
$ erl -sname my_name
We will see more details of this later. If you want to experiment with distributed Erlang, but you only have one
computer to work on, you can start two separate Erlang systems on the same computer but give them different names.
Each Erlang system running on a computer is called an Erlang node.
(Note: erl -sname assumes that all nodes are in the same IP domain and we can use only the first component of
the IP address, if we want to use nodes in different domains we use -name instead, but then all IP address must be
given in full.)
Here is the ping pong example modified to run on two separate nodes:
-module(tut17).
-export([start_ping/1, start_pong/0,
ping/2, pong/0]).
ping(0, Pong_Node) ->
{pong, Pong_Node} ! finished,
io:format("ping finished~n", []);
ping(N, Pong_Node) ->
{pong, Pong_Node} ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping(N - 1, Pong_Node).
pong() ->
receive
finished ->
io:format("Pong finished~n", []);
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
end.
start_pong() ->
register(pong, spawn(tut17, pong, [])).
start_ping(Pong_Node) ->
spawn(tut17, ping, [3, Pong_Node]).
Let us assume there are two computers called gollum and kosken. First a node is started on kosken, called ping, and
then a node on gollum, called pong.
On kosken (on a Linux/UNIX system):
kosken> erl -sname ping
Erlang (BEAM) emulator version 5.2.3.7 [hipe] [threads:0]
Eshell V5.2.3.7
(abort with ^G)
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 99
4.3 Concurrent Programming
(ping@kosken)1>
On gollum:
gollum> erl -sname pong
Erlang (BEAM) emulator version 5.2.3.7 [hipe] [threads:0]
Eshell V5.2.3.7
(pong@gollum)1>
(abort with ^G)
Now the "pong" process on gollum is started:
(pong@gollum)1> tut17:start_pong().
true
And the "ping" process on kosken is started (from the code above you can see that a parameter of the start_ping
function is the node name of the Erlang system where "pong" is running):
(ping@kosken)1> tut17:start_ping(pong@gollum).
<0.37.0>
Ping received pong
Ping received pong
Ping received pong
ping finished
As shown, the ping pong program has run. On the "pong" side:
(pong@gollum)2>
Pong received ping
Pong received ping
Pong received ping
Pong finished
(pong@gollum)2>
Looking at the tut17 code, you see that the pong function itself is unchanged, the following lines work in the same
way irrespective of on which node the "ping" process is executes:
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
Thus, Erlang pids contain information about where the process executes. So if you know the pid of a process, the "!"
operator can be used to send it a message disregarding if the process is on the same node or on a different node.
A difference is how messages are sent to a registered process on another node:
{pong, Pong_Node} ! {ping, self()},
100 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
A tuple {registered_name,node_name} is used instead of just the registered_name.
In the previous example, "ping" and "pong" were started from the shells of two separate Erlang nodes. spawn can
also be used to start processes in other nodes.
The next example is the ping pong program, yet again, but this time "ping" is started in another node:
-module(tut18).
-export([start/1,
ping/2, pong/0]).
ping(0, Pong_Node) ->
{pong, Pong_Node} ! finished,
io:format("ping finished~n", []);
ping(N, Pong_Node) ->
{pong, Pong_Node} ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping(N - 1, Pong_Node).
pong() ->
receive
finished ->
io:format("Pong finished~n", []);
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
end.
start(Ping_Node) ->
register(pong, spawn(tut18, pong, [])),
spawn(Ping_Node, tut18, ping, [3, node()]).
Assuming an Erlang system called ping (but not the "ping" process) has already been started on kosken, then on gollum
this is done:
(pong@gollum)1> tut18:start(ping@kosken).
<3934.39.0>
Pong received ping
Ping received pong
Pong received ping
Ping received pong
Pong received ping
Ping received pong
Pong finished
ping finished
Notice that all the output is received on gollum. This is because the I/O system finds out where the process is spawned
from and sends all output there.
4.3.5 A Larger Example
Now for a larger example with a simple "messenger". The messenger is a program that allows users to log in on
different nodes and send simple messages to each other.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 101
4.3 Concurrent Programming
Before starting, notice the following:
•
•
•
This example only shows the message passing logic - no attempt has been made to provide a nice graphical user
interface, although this can also be done in Erlang.
This sort of problem can be solved easier by use of the facilities in OTP, which also provide methods for updating
code on the fly and so on (see OTP Design Principles).
The first program contains some inadequacies regarding handling of nodes which disappear. These are corrected
in a later version of the program.
The messenger is set up by allowing "clients" to connect to a central server and say who and where they are. That is,
a user does not need to know the name of the Erlang node where another user is located to send a message.
File messenger.erl:
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
Message passing utility.
User interface:
logon(Name)
One user at a time can log in from each Erlang node in the
system messenger: and choose a suitable Name. If the Name
is already logged in at another node or if someone else is
already logged in at the same node, login will be rejected
with a suitable error message.
logoff()
Logs off anybody at that node
message(ToName, Message)
sends Message to ToName. Error messages if the user of this
function is not logged on or if ToName is not logged on at
any node.
One node in the network of Erlang nodes runs a server which maintains
data about the logged on users. The server is registered as "messenger"
Each node where there is a user logged on runs a client process registered
as "mess_client"
Protocol between the client processes and the server
---------------------------------------------------To server: {ClientPid, logon, UserName}
Reply {messenger, stop, user_exists_at_other_node} stops the client
Reply {messenger, logged_on} logon was successful
To server: {ClientPid, logoff}
Reply: {messenger, logged_off}
To server: {ClientPid, logoff}
Reply: no reply
To server: {ClientPid, message_to, ToName, Message} send a message
Reply: {messenger, stop, you_are_not_logged_on} stops the client
Reply: {messenger, receiver_not_found} no user with this name logged on
Reply: {messenger, sent} Message has been sent (but no guarantee)
To client: {message_from, Name, Message},
Protocol between the "commands" and the client
---------------------------------------------Started: messenger:client(Server_Node, Name)
To client: logoff
To client: {message_to, ToName, Message}
Configuration: change the server_node() function to return the
102 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
%%% name of the node where the messenger server runs
-module(messenger).
-export([start_server/0, server/1, logon/1, logoff/0, message/2, client/2]).
%%% Change the function below to return the name of the node where the
%%% messenger server runs
server_node() ->
messenger@bill.
%%% This is the server process for the "messenger"
%%% the user list has the format [{ClientPid1, Name1},{ClientPid22, Name2},...]
server(User_List) ->
receive
{From, logon, Name} ->
New_User_List = server_logon(From, Name, User_List),
server(New_User_List);
{From, logoff} ->
New_User_List = server_logoff(From, User_List),
server(New_User_List);
{From, message_to, To, Message} ->
server_transfer(From, To, Message, User_List),
io:format("list is now: ~p~n", [User_List]),
server(User_List)
end.
%%% Start the server
start_server() ->
register(messenger, spawn(messenger, server, [[]])).
%%% Server adds a new user to the user list
server_logon(From, Name, User_List) ->
%% check if logged on anywhere else
case lists:keymember(Name, 2, User_List) of
true ->
From ! {messenger, stop, user_exists_at_other_node}, %reject logon
User_List;
false ->
From ! {messenger, logged_on},
[{From, Name} | User_List]
%add user to the list
end.
%%% Server deletes a user from the user list
server_logoff(From, User_List) ->
lists:keydelete(From, 1, User_List).
%%% Server transfers a message between user
server_transfer(From, To, Message, User_List) ->
%% check that the user is logged on and who he is
case lists:keysearch(From, 1, User_List) of
false ->
From ! {messenger, stop, you_are_not_logged_on};
{value, {From, Name}} ->
server_transfer(From, Name, To, Message, User_List)
end.
%%% If the user exists, send the message
server_transfer(From, Name, To, Message, User_List) ->
%% Find the receiver and send the message
case lists:keysearch(To, 2, User_List) of
false ->
From ! {messenger, receiver_not_found};
{value, {ToPid, To}} ->
ToPid ! {message_from, Name, Message},
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 103
4.3 Concurrent Programming
end.
From ! {messenger, sent}
%%% User Commands
logon(Name) ->
case whereis(mess_client) of
undefined ->
register(mess_client,
spawn(messenger, client, [server_node(), Name]));
_ -> already_logged_on
end.
logoff() ->
mess_client ! logoff.
message(ToName, Message) ->
case whereis(mess_client) of % Test if the client is running
undefined ->
not_logged_on;
_ -> mess_client ! {message_to, ToName, Message},
ok
end.
%%% The client process which runs on each server node
client(Server_Node, Name) ->
{messenger, Server_Node} ! {self(), logon, Name},
await_result(),
client(Server_Node).
client(Server_Node) ->
receive
logoff ->
{messenger, Server_Node} ! {self(), logoff},
exit(normal);
{message_to, ToName, Message} ->
{messenger, Server_Node} ! {self(), message_to, ToName, Message},
await_result();
{message_from, FromName, Message} ->
io:format("Message from ~p: ~p~n", [FromName, Message])
end,
client(Server_Node).
%%% wait for a response from the server
await_result() ->
receive
{messenger, stop, Why} -> % Stop the client
io:format("~p~n", [Why]),
exit(normal);
{messenger, What} -> % Normal response
io:format("~p~n", [What])
end.
To use this program, you need to:
•
•
Configure the server_node() function.
Copy the compiled code (messenger.beam) to the directory on each computer where you start Erlang.
In the following example using this program, nodes are started on four different computers. If you do not have that
many machines available on your network, you can start several nodes on the same machine.
Four Erlang nodes are started up: messenger@super, c1@bilbo, c2@kosken, c3@gollum.
104 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
First the server at messenger@super is started up:
(messenger@super)1> messenger:start_server().
true
Now Peter logs on at c1@bilbo:
(c1@bilbo)1> messenger:logon(peter).
true
logged_on
James logs on at c2@kosken:
(c2@kosken)1> messenger:logon(james).
true
logged_on
And Fred logs on at c3@gollum:
(c3@gollum)1> messenger:logon(fred).
true
logged_on
Now Peter sends Fred a message:
(c1@bilbo)2> messenger:message(fred, "hello").
ok
sent
Fred receives the message and sends a message to Peter and logs off:
Message from peter: "hello"
(c3@gollum)2> messenger:message(peter, "go away, I'm busy").
ok
sent
(c3@gollum)3> messenger:logoff().
logoff
James now tries to send a message to Fred:
(c2@kosken)2> messenger:message(fred, "peter doesn't like you").
ok
receiver_not_found
But this fails as Fred has already logged off.
First let us look at some of the new concepts that have been introduced.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 105
4.3 Concurrent Programming
There are two versions of the server_transfer function: one with four arguments (server_transfer/4)
and one with five (server_transfer/5). These are regarded by Erlang as two separate functions.
Notice how to write the server function so that it calls itself, through server(User_List), and thus creates
a loop. The Erlang compiler is "clever" and optimizes the code so that this really is a sort of loop and not a proper
function call. But this only works if there is no code after the call. Otherwise, the compiler expects the call to return
and make a proper function call. This would result in the process getting bigger and bigger for every loop.
Functions in the lists module are used. This is a very useful module and a study of the manual page is recommended
(erl -man lists). lists:keymember(Key,Position,Lists) looks through a list of tuples and looks
at Position in each tuple to see if it is the same as Key. The first element is position 1. If it finds a tuple where the
element at Position is the same as Key, it returns true, otherwise false.
3> lists:keymember(a, 2, [{x,y,z},{b,b,b},{b,a,c},{q,r,s}]).
true
4> lists:keymember(p, 2, [{x,y,z},{b,b,b},{b,a,c},{q,r,s}]).
false
lists:keydelete works in the same way but deletes the first tuple found (if any) and returns the remaining list:
5> lists:keydelete(a, 2, [{x,y,z},{b,b,b},{b,a,c},{q,r,s}]).
[{x,y,z},{b,b,b},{q,r,s}]
lists:keysearch is like lists:keymember, but it returns {value,Tuple_Found} or the atom false.
There are many very useful functions in the lists module.
An Erlang process (conceptually) runs until it does a receive and there is no message which it wants to receive in
the message queue. "conceptually" is used here because the Erlang system shares the CPU time between the active
processes in the system.
A process terminates when there is nothing more for it to do, that is, the last function it calls simply returns and does
not call another function. Another way for a process to terminate is for it to call exit/1. The argument to exit/1
has a special meaning, which is discussed later. In this example, exit(normal) is done, which has the same effect
as a process running out of functions to call.
The BIF whereis(RegisteredName) checks if a registered process of name RegisteredName exists. If it
exists, the pid of that process is returned. If it does not exist, the atom undefined is returned.
You should by now be able to understand most of the code in the messenger-module. Let us study one case in detail:
a message is sent from one user to another.
The first user "sends" the message in the example above by:
messenger:message(fred, "hello")
After testing that the client process exists:
whereis(mess_client)
And a message is sent to mess_client:
106 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.3 Concurrent Programming
mess_client ! {message_to, fred, "hello"}
The client sends the message to the server by:
{messenger, messenger@super} ! {self(), message_to, fred, "hello"},
And waits for a reply from the server.
The server receives this message and calls:
server_transfer(From, fred, "hello", User_List),
This checks that the pid From is in the User_List:
lists:keysearch(From, 1, User_List)
If keysearch returns the atom false, some error has occurred and the server sends back the message:
From ! {messenger, stop, you_are_not_logged_on}
This is received by the client, which in turn does exit(normal) and terminates. If keysearch returns {value,
{From,Name}} it is certain that the user is logged on and that his name (peter) is in variable Name.
Let us now call:
server_transfer(From, peter, fred, "hello", User_List)
Notice that as this is server_transfer/5, it is not the same as the previous function server_transfer/4.
Another keysearch is done on User_List to find the pid of the client corresponding to fred:
lists:keysearch(fred, 2, User_List)
This time argument 2 is used, which is the second element in the tuple. If this returns the atom false, fred is not
logged on and the following message is sent:
From ! {messenger, receiver_not_found};
This is received by the client.
If keysearch returns:
{value, {ToPid, fred}}
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 107
4.4 Robustness
The following message is sent to fred's client:
ToPid ! {message_from, peter, "hello"},
The following message is sent to peter's client:
From ! {messenger, sent}
Fred's client receives the message and prints it:
{message_from, peter, "hello"} ->
io:format("Message from ~p: ~p~n", [peter, "hello"])
Peter's client receives the message in the await_result function.
4.4 Robustness
Several things are wrong with the messenger example in A Larger Example. For example, if a node where a user is
logged on goes down without doing a logoff, the user remains in the server's User_List, but the client disappears.
This makes it impossible for the user to log on again as the server thinks the user already is logged on.
Or what happens if the server goes down in the middle of sending a message, leaving the sending client hanging forever
in the await_result function?
4.4.1 Time-outs
Before improving the messenger program, let us look at some general principles, using the ping pong program as an
example. Recall that when "ping" finishes, it tells "pong" that it has done so by sending the atom finished as a
message to "pong" so that "pong" can also finish. Another way to let "pong" finish is to make "pong" exit if it does
not receive a message from ping within a certain time. This can be done by adding a time-out to pong as shown in
the following example:
-module(tut19).
-export([start_ping/1, start_pong/0,
ping/2, pong/0]).
ping(0, Pong_Node) ->
io:format("ping finished~n", []);
ping(N, Pong_Node) ->
{pong, Pong_Node} ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping(N - 1, Pong_Node).
pong() ->
receive
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
108 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.4 Robustness
pong()
after 5000 ->
io:format("Pong timed out~n", [])
end.
start_pong() ->
register(pong, spawn(tut19, pong, [])).
start_ping(Pong_Node) ->
spawn(tut19, ping, [3, Pong_Node]).
After this is compiled and the file tut19.beam is copied to the necessary directories, the following is seen on
(pong@kosken):
(pong@kosken)1> tut19:start_pong().
true
Pong received ping
Pong received ping
Pong received ping
Pong timed out
And the following is seen on (ping@gollum):
(ping@gollum)1> tut19:start_ping(pong@kosken).
<0.36.0>
Ping received pong
Ping received pong
Ping received pong
ping finished
The time-out is set in:
pong() ->
receive
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
after 5000 ->
io:format("Pong timed out~n", [])
end.
The time-out (after 5000) is started when receive is entered. The time-out is canceled if {ping,Ping_PID}
is received. If {ping,Ping_PID} is not received, the actions following the time-out are done after 5000
milliseconds. after must be last in the receive, that is, preceded by all other message reception specifications in
the receive. It is also possible to call a function that returned an integer for the time-out:
after pong_timeout() ->
In general, there are better ways than using time-outs to supervise parts of a distributed Erlang system. Time-outs are
usually appropriate to supervise external events, for example, if you have expected a message from some external
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 109
4.4 Robustness
system within a specified time. For example, a time-out can be used to log a user out of the messenger system if they
have not accessed it for, say, ten minutes.
4.4.2 Error Handling
Before going into details of the supervision and error handling in an Erlang system, let us see how Erlang processes
terminate, or in Erlang terminology, exit.
A process which executes exit(normal) or simply runs out of things to do has a normal exit.
A process which encounters a runtime error (for example, divide by zero, bad match, trying to call a function that does
not exist and so on) exits with an error, that is, has an abnormal exit. A process which executes exit(Reason) where
Reason is any Erlang term except the atom normal, also has an abnormal exit.
An Erlang process can set up links to other Erlang processes. If a process calls link(Other_Pid) it sets up a bidirectional
link between itself and the process called Other_Pid. When a process terminates, it sends something called a signal
to all the processes it has links to.
The signal carries information about the pid it was sent from and the exit reason.
The default behaviour of a process that receives a normal exit is to ignore the signal.
The default behaviour in the two other cases (that is, abnormal exit) above is to:
•
•
•
Bypass all messages to the receiving process.
Kill the receiving process.
Propagate the same error signal to the links of the killed process.
In this way you can connect all processes in a transaction together using links. If one of the processes exits abnormally,
all the processes in the transaction are killed. As it is often wanted to create a process and link to it at the same time,
there is a special BIF, spawn_link that does the same as spawn, but also creates a link to the spawned process.
Now an example of the ping pong example using links to terminate "pong":
-module(tut20).
-export([start/1,
ping/2, pong/0]).
ping(N, Pong_Pid) ->
link(Pong_Pid),
ping1(N, Pong_Pid).
ping1(0, _) ->
exit(ping);
ping1(N, Pong_Pid) ->
Pong_Pid ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping1(N - 1, Pong_Pid).
pong() ->
receive
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong()
end.
start(Ping_Node) ->
110 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.4 Robustness
PongPID = spawn(tut20, pong, []),
spawn(Ping_Node, tut20, ping, [3, PongPID]).
(s1@bill)3> tut20:start(s2@kosken).
Pong received ping
<3820.41.0>
Ping received pong
Pong received ping
Ping received pong
Pong received ping
Ping received pong
This is a slight modification of the ping pong program where both processes are spawned from the same start/1
function, and the "ping" process can be spawned on a separate node. Notice the use of the link BIF. "Ping" calls
exit(ping) when it finishes and this causes an exit signal to be sent to "pong", which also terminates.
It is possible to modify the default behaviour of a process so that it does not get killed when it receives abnormal
exit signals. Instead, all signals are turned into normal messages on the format {'EXIT',FromPID,Reason} and
added to the end of the receiving process' message queue. This behaviour is set by:
process_flag(trap_exit, true)
There are several other process flags, see erlang(3). Changing the default behaviour of a process in this way is usually
not done in standard user programs, but is left to the supervisory programs in OTP. However, the ping pong program
is modified to illustrate exit trapping.
-module(tut21).
-export([start/1,
ping/2, pong/0]).
ping(N, Pong_Pid) ->
link(Pong_Pid),
ping1(N, Pong_Pid).
ping1(0, _) ->
exit(ping);
ping1(N, Pong_Pid) ->
Pong_Pid ! {ping, self()},
receive
pong ->
io:format("Ping received pong~n", [])
end,
ping1(N - 1, Pong_Pid).
pong() ->
process_flag(trap_exit, true),
pong1().
pong1() ->
receive
{ping, Ping_PID} ->
io:format("Pong received ping~n", []),
Ping_PID ! pong,
pong1();
{'EXIT', From, Reason} ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 111
4.4 Robustness
end.
io:format("pong exiting, got ~p~n", [{'EXIT', From, Reason}])
start(Ping_Node) ->
PongPID = spawn(tut21, pong, []),
spawn(Ping_Node, tut21, ping, [3, PongPID]).
(s1@bill)1> tut21:start(s2@gollum).
<3820.39.0>
Pong received ping
Ping received pong
Pong received ping
Ping received pong
Pong received ping
Ping received pong
pong exiting, got {'EXIT',<3820.39.0>,ping}
4.4.3 The Larger Example with Robustness Added
Let us return to the messenger program and add changes to make it more robust:
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
%%%
Message passing utility.
User interface:
login(Name)
One user at a time can log in from each Erlang node in the
system messenger: and choose a suitable Name. If the Name
is already logged in at another node or if someone else is
already logged in at the same node, login will be rejected
with a suitable error message.
logoff()
Logs off anybody at that node
message(ToName, Message)
sends Message to ToName. Error messages if the user of this
function is not logged on or if ToName is not logged on at
any node.
One node in the network of Erlang nodes runs a server which maintains
data about the logged on users. The server is registered as "messenger"
Each node where there is a user logged on runs a client process registered
as "mess_client"
Protocol between the client processes and the server
---------------------------------------------------To server: {ClientPid, logon, UserName}
Reply {messenger, stop, user_exists_at_other_node} stops the client
Reply {messenger, logged_on} logon was successful
When the client terminates for some reason
To server: {'EXIT', ClientPid, Reason}
To server: {ClientPid, message_to, ToName, Message} send a message
Reply: {messenger, stop, you_are_not_logged_on} stops the client
Reply: {messenger, receiver_not_found} no user with this name logged on
Reply: {messenger, sent} Message has been sent (but no guarantee)
To client: {message_from, Name, Message},
Protocol between the "commands" and the client
----------------------------------------------
112 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.4 Robustness
%%%
%%%
%%%
%%%
%%%
%%%
%%%
Started: messenger:client(Server_Node, Name)
To client: logoff
To client: {message_to, ToName, Message}
Configuration: change the server_node() function to return the
name of the node where the messenger server runs
-module(messenger).
-export([start_server/0, server/0,
logon/1, logoff/0, message/2, client/2]).
%%% Change the function below to return the name of the node where the
%%% messenger server runs
server_node() ->
messenger@super.
%%% This is the server process for the "messenger"
%%% the user list has the format [{ClientPid1, Name1},{ClientPid22, Name2},...]
server() ->
process_flag(trap_exit, true),
server([]).
server(User_List) ->
receive
{From, logon, Name} ->
New_User_List = server_logon(From, Name, User_List),
server(New_User_List);
{'EXIT', From, _} ->
New_User_List = server_logoff(From, User_List),
server(New_User_List);
{From, message_to, To, Message} ->
server_transfer(From, To, Message, User_List),
io:format("list is now: ~p~n", [User_List]),
server(User_List)
end.
%%% Start the server
start_server() ->
register(messenger, spawn(messenger, server, [])).
%%% Server adds a new user to the user list
server_logon(From, Name, User_List) ->
%% check if logged on anywhere else
case lists:keymember(Name, 2, User_List) of
true ->
From ! {messenger, stop, user_exists_at_other_node}, %reject logon
User_List;
false ->
From ! {messenger, logged_on},
link(From),
[{From, Name} | User_List]
%add user to the list
end.
%%% Server deletes a user from the user list
server_logoff(From, User_List) ->
lists:keydelete(From, 1, User_List).
%%% Server transfers a message between user
server_transfer(From, To, Message, User_List) ->
%% check that the user is logged on and who he is
case lists:keysearch(From, 1, User_List) of
false ->
From ! {messenger, stop, you_are_not_logged_on};
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 113
4.4 Robustness
{value, {_, Name}} ->
server_transfer(From, Name, To, Message, User_List)
end.
%%% If the user exists, send the message
server_transfer(From, Name, To, Message, User_List) ->
%% Find the receiver and send the message
case lists:keysearch(To, 2, User_List) of
false ->
From ! {messenger, receiver_not_found};
{value, {ToPid, To}} ->
ToPid ! {message_from, Name, Message},
From ! {messenger, sent}
end.
%%% User Commands
logon(Name) ->
case whereis(mess_client) of
undefined ->
register(mess_client,
spawn(messenger, client, [server_node(), Name]));
_ -> already_logged_on
end.
logoff() ->
mess_client ! logoff.
message(ToName, Message) ->
case whereis(mess_client) of % Test if the client is running
undefined ->
not_logged_on;
_ -> mess_client ! {message_to, ToName, Message},
ok
end.
%%% The client process which runs on each user node
client(Server_Node, Name) ->
{messenger, Server_Node} ! {self(), logon, Name},
await_result(),
client(Server_Node).
client(Server_Node) ->
receive
logoff ->
exit(normal);
{message_to, ToName, Message} ->
{messenger, Server_Node} ! {self(), message_to, ToName, Message},
await_result();
{message_from, FromName, Message} ->
io:format("Message from ~p: ~p~n", [FromName, Message])
end,
client(Server_Node).
%%% wait for a response from the server
await_result() ->
receive
{messenger, stop, Why} -> % Stop the client
io:format("~p~n", [Why]),
exit(normal);
{messenger, What} -> % Normal response
io:format("~p~n", [What])
after 5000 ->
io:format("No response from server~n", []),
exit(timeout)
114 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.5 Records and Macros
end.
The following changes are added:
The messenger server traps exits. If it receives an exit signal, {'EXIT',From,Reason}, this means that a client
process has terminated or is unreachable for one of the following reasons:
•
•
•
•
The user has logged off (the "logoff" message is removed).
The network connection to the client is broken.
The node on which the client process resides has gone down.
The client processes has done some illegal operation.
If an exit signal is received as above, the tuple {From,Name} is deleted from the servers User_List using the
server_logoff function. If the node on which the server runs goes down, an exit signal (automatically generated
by the system) is sent to all of the client processes: {'EXIT',MessengerPID,noconnection} causing all the
client processes to terminate.
Also, a time-out of five seconds has been introduced in the await_result function. That is, if the server does not
reply within five seconds (5000 ms), the client terminates. This is only needed in the logon sequence before the client
and the server are linked.
An interesting case is if the client terminates before the server links to it. This is taken care of because linking to a
non-existent process causes an exit signal, {'EXIT',From,noproc}, to be automatically generated. This is as if
the process terminated immediately after the link operation.
4.5 Records and Macros
Larger programs are usually written as a collection of files with a well-defined interface between the various parts.
4.5.1 The Larger Example Divided into Several Files
To illustrate this, the messenger example from the previous section is divided into the following five files:
•
mess_config.hrl
•
Header file for configuration data
mess_interface.hrl
•
Interface definitions between the client and the messenger
user_interface.erl
•
Functions for the user interface
mess_client.erl
•
Functions for the client side of the messenger
mess_server.erl
Functions for the server side of the messenger
While doing this, the message passing interface between the shell, the client, and the server is cleaned up and is defined
using records. Also, macros are introduced:
%%%----FILE mess_config.hrl---%%% Configure the location of the server node,
-define(server_node, messenger@super).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 115
4.5 Records and Macros
%%%----END FILE----
%%%----FILE mess_interface.hrl---%%% Message interface between client and server and client shell for
%%% messenger program
%%%Messages from Client to server received in server/1 function.
-record(logon,{client_pid, username}).
-record(message,{client_pid, to_name, message}).
%%% {'EXIT', ClientPid, Reason} (client terminated or unreachable.
%%% Messages from Server to Client, received in await_result/0 function
-record(abort_client,{message}).
%%% Messages are: user_exists_at_other_node,
%%%
you_are_not_logged_on
-record(server_reply,{message}).
%%% Messages are: logged_on
%%%
receiver_not_found
%%%
sent (Message has been sent (no guarantee)
%%% Messages from Server to Client received in client/1 function
-record(message_from,{from_name, message}).
%%% Messages from shell to Client received in client/1 function
%%% spawn(mess_client, client, [server_node(), Name])
-record(message_to,{to_name, message}).
%%% logoff
%%%----END FILE----
%%%----FILE user_interface.erl---%%% User interface to the messenger program
%%% login(Name)
%%%
One user at a time can log in from each Erlang node in the
%%%
system messenger: and choose a suitable Name. If the Name
%%%
is already logged in at another node or if someone else is
%%%
already logged in at the same node, login will be rejected
%%%
with a suitable error message.
%%% logoff()
%%%
Logs off anybody at that node
%%% message(ToName, Message)
%%%
sends Message to ToName. Error messages if the user of this
%%%
function is not logged on or if ToName is not logged on at
%%%
any node.
-module(user_interface).
-export([logon/1, logoff/0, message/2]).
-include("mess_interface.hrl").
-include("mess_config.hrl").
logon(Name) ->
case whereis(mess_client) of
undefined ->
register(mess_client,
spawn(mess_client, client, [?server_node, Name]));
_ -> already_logged_on
end.
116 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.5 Records and Macros
logoff() ->
mess_client ! logoff.
message(ToName, Message) ->
case whereis(mess_client) of % Test if the client is running
undefined ->
not_logged_on;
_ -> mess_client ! #message_to{to_name=ToName, message=Message},
ok
end.
%%%----END FILE----
%%%----FILE mess_client.erl---%%% The client process which runs on each user node
-module(mess_client).
-export([client/2]).
-include("mess_interface.hrl").
client(Server_Node, Name) ->
{messenger, Server_Node} ! #logon{client_pid=self(), username=Name},
await_result(),
client(Server_Node).
client(Server_Node) ->
receive
logoff ->
exit(normal);
#message_to{to_name=ToName, message=Message} ->
{messenger, Server_Node} !
#message{client_pid=self(), to_name=ToName, message=Message},
await_result();
{message_from, FromName, Message} ->
io:format("Message from ~p: ~p~n", [FromName, Message])
end,
client(Server_Node).
%%% wait for a response from the server
await_result() ->
receive
#abort_client{message=Why} ->
io:format("~p~n", [Why]),
exit(normal);
#server_reply{message=What} ->
io:format("~p~n", [What])
after 5000 ->
io:format("No response from server~n", []),
exit(timeout)
end.
%%%----END FILE---
%%%----FILE mess_server.erl---%%% This is the server process of the messenger service
-module(mess_server).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 117
4.5 Records and Macros
-export([start_server/0, server/0]).
-include("mess_interface.hrl").
server() ->
process_flag(trap_exit, true),
server([]).
%%% the user list has the format [{ClientPid1, Name1},{ClientPid22, Name2},...]
server(User_List) ->
io:format("User list = ~p~n", [User_List]),
receive
#logon{client_pid=From, username=Name} ->
New_User_List = server_logon(From, Name, User_List),
server(New_User_List);
{'EXIT', From, _} ->
New_User_List = server_logoff(From, User_List),
server(New_User_List);
#message{client_pid=From, to_name=To, message=Message} ->
server_transfer(From, To, Message, User_List),
server(User_List)
end.
%%% Start the server
start_server() ->
register(messenger, spawn(?MODULE, server, [])).
%%% Server adds a new user to the user list
server_logon(From, Name, User_List) ->
%% check if logged on anywhere else
case lists:keymember(Name, 2, User_List) of
true ->
From ! #abort_client{message=user_exists_at_other_node},
User_List;
false ->
From ! #server_reply{message=logged_on},
link(From),
[{From, Name} | User_List]
%add user to the list
end.
%%% Server deletes a user from the user list
server_logoff(From, User_List) ->
lists:keydelete(From, 1, User_List).
%%% Server transfers a message between user
server_transfer(From, To, Message, User_List) ->
%% check that the user is logged on and who he is
case lists:keysearch(From, 1, User_List) of
false ->
From ! #abort_client{message=you_are_not_logged_on};
{value, {_, Name}} ->
server_transfer(From, Name, To, Message, User_List)
end.
%%% If the user exists, send the message
server_transfer(From, Name, To, Message, User_List) ->
%% Find the receiver and send the message
case lists:keysearch(To, 2, User_List) of
false ->
From ! #server_reply{message=receiver_not_found};
{value, {ToPid, To}} ->
ToPid ! #message_from{from_name=Name, message=Message},
From ! #server_reply{message=sent}
end.
%%%----END FILE---
118 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
4.5 Records and Macros
4.5.2 Header Files
As shown above, some files have extension .hrl. These are header files that are included in the .erl files by:
-include("File_Name").
for example:
-include("mess_interface.hrl").
In the case above the file is fetched from the same directory as all the other files in the messenger example. (*manual*).
.hrl files can contain any valid Erlang code but are most often used for record and macro definitions.
4.5.3 Records
A record is defined as:
-record(name_of_record,{field_name1, field_name2, field_name3, ......}).
For example:
-record(message_to,{to_name, message}).
This is equivalent to:
{message_to, To_Name, Message}
Creating a record is best illustrated by an example:
#message_to{message="hello", to_name=fred)
This creates:
{message_to, fred, "hello"}
Notice that you do not have to worry about the order you assign values to the various parts of the records when you
create it. The advantage of using records is that by placing their definitions in header files you can conveniently define
interfaces that are easy to change. For example, if you want to add a new field to the record, you only have to change
the code where the new field is used and not at every place the record is referred to. If you leave out a field when
creating a record, it gets the value of the atom undefined. (*manual*)
Pattern matching with records is very similar to creating records. For example, inside a case or receive:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 119
4.5 Records and Macros
#message_to{to_name=ToName, message=Message} ->
This is the same as:
{message_to, ToName, Message}
4.5.4 Macros
Another thing that has been added to the messenger is a macro. The file mess_config.hrl contains the definition:
%%% Configure the location of the server node,
-define(server_node, messenger@super).
This file is included in mess_server.erl:
-include("mess_config.hrl").
Every occurrence of ?server_node in mess_server.erl is now replaced by messenger@super.
A macro is also used when spawning the server process:
spawn(?MODULE, server, [])
This is a standard macro (that is, defined by the system, not by the user). ?MODULE is always replaced by the name
of the current module (that is, the -module definition near the start of the file). There are more advanced ways of
using macros with, for example, parameters (*manual*).
The three Erlang (.erl) files in the messenger example are individually compiled into object code file (.beam). The
Erlang system loads and links these files into the system when they are referred to during execution of the code. In
this case, they are simply put in our current working directory (that is, the place you have done "cd" to). There are
ways of putting the .beam files in other directories.
In the messenger example, no assumptions have been made about what the message being sent is. It can be any valid
Erlang term.
120 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.1 Introduction
5 Erlang Reference Manual
5.1 Introduction
This section is the Erlang reference manual. It describes the Erlang programming language.
5.1.1 Purpose
The focus of the Erlang reference manual is on the language itself, not the implementation of it. The language constructs
are described in text and with examples rather than formally specified. This is to make the manual more readable. The
Erlang reference manual is not intended as a tutorial.
Information about implementation of Erlang can, for example, be found, in the following:
•
System Principles
•
Starting and stopping, boot scripts, code loading, error logging, creating target systems
Efficiency Guide
•
Memory consumption, system limits
ERTS User's Guide
Crash dumps, drivers
5.1.2 Prerequisites
It is assumed that the reader has done some programming and is familiar with concepts such as data types and
programming language syntax.
5.1.3 Document Conventions
In this section, the following terminology is used:
•
•
A sequence is one or more items. For example, a clause body consists of a sequence of expressions. This means
that there must be at least one expression.
A list is any number of items. For example, an argument list can consist of zero, one, or more arguments.
If a feature has been added in R13A or later, this is mentioned in the text.
5.1.4 Complete List of BIFs
For a complete list of BIFs, their arguments and return values, see erlang(3) manual page in ERTS.
5.1.5 Reserved Words
The following are reserved words in Erlang:
after and andalso band begin bnot bor bsl bsr bxor case catch cond div end fun
if let not of or orelse receive rem try when xor
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 121
5.2 Character Set and Source File Encoding
5.2 Character Set and Source File Encoding
5.2.1 Character Set
The syntax of Erlang tokens allow the use of the full ISO-8859-1 (Latin-1) character set. This is noticeable in the
following ways:
•
•
All the Latin-1 printable characters can be used and are shown without the escape backslash convention.
Atoms and variables can use all Latin-1 letters.
Octal
Decimal
Class
200 - 237
128 - 159
Control characters
240 - 277
160 - 191
- ¿ Punctuation characters
300 - 326
192 - 214
327
À-Ö
215
×
Uppercase letters
Punctuation character
330 - 336
216 - 222
Ø-Þ
Uppercase letters
337 - 366
223 - 246
ß-ö
Lowercase letters
367
370 - 377
247
248 - 255
÷
ø-ÿ
Punctuation character
Lowercase letters
Table 2.1: Character Classes
In Erlang/OTP R16B the syntax of Erlang tokens was extended to handle Unicode. The support was limited to string
literals and comments. More about the usage of Unicode in Erlang source files can be found in STDLIB's User's Guide.
From Erlang/OTP 20, atoms and function names are also allowed to contain Unicode characters outside the ISOLatin-1 range. Module names, application names, and node names are still restricted to the ISO-Latin-1 range.
5.2.2 Source File Encoding
The Erlang source file encoding is selected by a comment in one of the first two lines of the source file. The first
string that matches the regular expression coding\s*[:=]\s*([-a-zA-Z0-9])+ selects the encoding. If the
matching string is an invalid encoding, it is ignored. The valid encodings are Latin-1 and UTF-8, where the case
of the characters can be chosen freely.
The following example selects UTF-8 as default encoding:
%% coding: utf-8
Two more examples, both selecting Latin-1 as default encoding:
%% For this file we have chosen encoding = Latin-1
122 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.3 Data Types
%% -*- coding: latin-1 -*-
The default encoding for Erlang source files is changed from Latin-1 to UTF-8 since Erlang/OTP 17.0.
5.3 Data Types
Erlang provides a number of data types, which are listed in this section.
5.3.1 Terms
A piece of data of any data type is called a term.
5.3.2 Number
There are two types of numeric literals, integers and floats. Besides the conventional notation, there are two Erlangspecific notations:
•
•
$char
ASCII value or unicode code-point of the character char.
base#value
Integer with the base base, that must be an integer in the range 2..36.
Examples:
1> 42.
42
2> $A.
65
3> $\n.
10
4> 2#101.
5
5> 16#1f.
31
6> 2.3.
2.3
7> 2.3e3.
2.3e3
8> 2.3e-3.
0.0023
5.3.3 Atom
An atom is a literal, a constant with name. An atom is to be enclosed in single quotes (') if it does not begin with a
lower-case letter or if it contains other characters than alphanumeric characters, underscore (_), or @.
Examples:
hello
phone_number
'Monday'
'phone number'
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 123
5.3 Data Types
5.3.4 Bit Strings and Binaries
A bit string is used to store an area of untyped memory.
Bit strings are expressed using the bit syntax.
Bit strings that consist of a number of bits that are evenly divisible by eight, are called binaries
Examples:
1> <<10,20>>.
<<10,20>>
2> <<"ABC">>.
<<"ABC">>
1> <<1:1,0:1>>.
<<2:2>>
For more examples, see Programming Examples.
5.3.5 Reference
A reference is a term that is unique in an Erlang runtime system, created by calling make_ref/0.
5.3.6 Fun
A fun is a functional object. Funs make it possible to create an anonymous function and pass the function itself -- not
its name -- as argument to other functions.
Example:
1> Fun1 = fun (X) -> X+1 end.
#Fun<erl_eval.6.39074546>
2> Fun1(2).
3
Read more about funs in Fun Expressions. For more examples, see Programming Examples.
5.3.7 Port Identifier
A port identifier identifies an Erlang port.
open_port/2, which is used to create ports, returns a value of this data type.
Read more about ports in Ports and Port Drivers.
5.3.8 Pid
A process identifier, pid, identifies a process.
The following BIFs, which are used to create processes, return values of this data type:
•
•
•
spawn/1,2,3,4
spawn_link/1,2,3,4
spawn_opt/4
Example:
124 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.3 Data Types
1> spawn(m, f, []).
<0.51.0>
In the following example, the BIF self() returns the pid of the calling process:
-module(m).
-export([loop/0]).
loop() ->
receive
who_are_you ->
io:format("I am ~p~n", [self()]),
loop()
end.
1> P = spawn(m, loop, []).
<0.58.0>
2> P ! who_are_you.
I am <0.58.0>
who_are_you
Read more about processes in Processes.
5.3.9 Tuple
A tuple is a compound data type with a fixed number of terms:
{Term1,...,TermN}
Each term Term in the tuple is called an element. The number of elements is said to be the size of the tuple.
There exists a number of BIFs to manipulate tuples.
Examples:
1> P = {adam,24,{july,29}}.
{adam,24,{july,29}}
2> element(1,P).
adam
3> element(3,P).
{july,29}
4> P2 = setelement(2,P,25).
{adam,25,{july,29}}
5> tuple_size(P).
3
6> tuple_size({}).
0
5.3.10 Map
A map is a compound data type with a variable number of key-value associations:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 125
5.3 Data Types
#{Key1=>Value1,...,KeyN=>ValueN}
Each key-value association in the map is called an association pair. The key and value parts of the pair are called
elements. The number of association pairs is said to be the size of the map.
There exists a number of BIFs to manipulate maps.
Examples:
1> M1 = #{name=>adam,age=>24,date=>{july,29}}.
#{age => 24,date => {july,29},name => adam}
2> maps:get(name,M1).
adam
3> maps:get(date,M1).
{july,29}
4> M2 = maps:update(age,25,M1).
#{age => 25,date => {july,29},name => adam}
5> map_size(M).
3
6> map_size(#{}).
0
A collection of maps processing functions can be found in maps manual page in STDLIB.
Read more about maps in Map Expressions.
Note:
Maps are considered to be experimental during Erlang/OTP R17.
5.3.11 List
A list is a compound data type with a variable number of terms.
[Term1,...,TermN]
Each term Term in the list is called an element. The number of elements is said to be the length of the list.
Formally, a list is either the empty list [] or consists of a head (first element) and a tail (remainder of the list). The
tail is also a list. The latter can be expressed as [H|T]. The notation [Term1,...,TermN] above is equivalent
with the list [Term1|[...|[TermN|[]]]].
Example:
[] is a list, thus
[c|[]] is a list, thus
[b|[c|[]]] is a list, thus
[a|[b|[c|[]]]] is a list, or in short [a,b,c]
A list where the tail is a list is sometimes called a proper list. It is allowed to have a list where the tail is not a list,
for example, [a|b]. However, this type of list is of little practical use.
Examples:
126 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.3 Data Types
1> L1 = [a,2,{c,4}].
[a,2,{c,4}]
2> [H|T] = L1.
[a,2,{c,4}]
3> H.
a
4> T.
[2,{c,4}]
5> L2 = [d|T].
[d,2,{c,4}]
6> length(L1).
3
7> length([]).
0
A collection of list processing functions can be found in the lists manual page in STDLIB.
5.3.12 String
Strings are enclosed in double quotes ("), but is not a data type in Erlang. Instead, a string "hello" is shorthand for
the list [$h,$e,$l,$l,$o], that is, [104,101,108,108,111].
Two adjacent string literals are concatenated into one. This is done in the compilation, thus, does not incur any runtime
overhead.
Example:
"string" "42"
is equivalent to
"string42"
5.3.13 Record
A record is a data structure for storing a fixed number of elements. It has named fields and is similar to a struct in
C. However, a record is not a true data type. Instead, record expressions are translated to tuple expressions during
compilation. Therefore, record expressions are not understood by the shell unless special actions are taken. For details,
see the shell(3) manual page in STDLIB).
Examples:
-module(person).
-export([new/2]).
-record(person, {name, age}).
new(Name, Age) ->
#person{name=Name, age=Age}.
1> person:new(ernie, 44).
{person,ernie,44}
Read more about records in Records. More examples can be found in Programming Examples.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 127
5.3 Data Types
5.3.14 Boolean
There is no Boolean data type in Erlang. Instead the atoms true and false are used to denote Boolean values.
Examples:
1> 2 =< 3.
true
2> true or false.
true
5.3.15 Escape Sequences
Within strings and quoted atoms, the following escape sequences are recognized:
Sequence
Description
\b
Backspace
\d
Delete
\e
Escape
\f
Form feed
\n
Newline
\r
Carriage return
\s
Space
\t
Tab
\v
Vertical tab
\XYZ, \YZ, \Z
Character with octal representation XYZ, YZ or Z
\xXY
Character with hexadecimal representation XY
\x{X...}
Character with hexadecimal representation; X... is one
or more hexadecimal characters
\^a...\^z
\^A...\^Z
Control A to control Z
\'
Single quote
\"
Double quote
\\
Backslash
Table 3.1: Recognized Escape Sequences
128 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.4 Pattern Matching
5.3.16 Type Conversions
There are a number of BIFs for type conversions.
Examples:
1> atom_to_list(hello).
"hello"
2> list_to_atom("hello").
hello
3> binary_to_list(<<"hello">>).
"hello"
4> binary_to_list(<<104,101,108,108,111>>).
"hello"
5> list_to_binary("hello").
<<104,101,108,108,111>>
6> float_to_list(7.0).
"7.00000000000000000000e+00"
7> list_to_float("7.000e+00").
7.0
8> integer_to_list(77).
"77"
9> list_to_integer("77").
77
10> tuple_to_list({a,b,c}).
[a,b,c]
11> list_to_tuple([a,b,c]).
{a,b,c}
12> term_to_binary({a,b,c}).
<<131,104,3,100,0,1,97,100,0,1,98,100,0,1,99>>
13> binary_to_term(<<131,104,3,100,0,1,97,100,0,1,98,100,0,1,99>>).
{a,b,c}
14> binary_to_integer(<<"77">>).
77
15> integer_to_binary(77).
<<"77">>
16> float_to_binary(7.0).
<<"7.00000000000000000000e+00">>
17> binary_to_float(<<"7.000e+00">>).
7.0
5.4 Pattern Matching
5.4.1 Pattern Matching
Variables are bound to values through the pattern matching mechanism. Pattern matching occurs when evaluating a
function call, case- receive- try- expressions and match operator (=) expressions.
In a pattern matching, a left-hand side pattern is matched against a right-hand side term. If the matching succeeds, any
unbound variables in the pattern become bound. If the matching fails, a run-time error occurs.
Examples:
1>
**
2>
2
3>
3
X.
1: variable 'X' is unbound **
X = 2.
X + 1.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 129
5.5 Modules
4> {X, Y} = {1, 2}.
** exception error: no match of right hand side value {1,2}
5> {X, Y} = {2, 3}.
{2,3}
6> Y.
3
5.5 Modules
5.5.1 Module Syntax
Erlang code is divided into modules. A module consists of a sequence of attributes and function declarations, each
terminated by period (.).
Example:
-module(m).
-export([fact/1]).
% module attribute
% module attribute
fact(N) when N>0 ->
N * fact(N-1);
fact(0) ->
1.
% beginning of function declaration
% |
% |
% end of function declaration
For a description of function declarations, see Function Declaration Syntax.
5.5.2 Module Attributes
A module attribute defines a certain property of a module.
A module attribute consists of a tag and a value:
-Tag(Value).
Tag must be an atom, while Value must be a literal term. As a convenience in user-defined attributes, if the literal
term Value has the syntax Name/Arity (where Name is an atom and Arity a positive integer), the term Name/
Arity is translated to {Name,Arity}.
Any module attribute can be specified. The attributes are stored in the compiled code and can be retrieved by calling
Module:module_info(attributes), or by using the module beam_lib(3) in STDLIB.
Several module attributes have predefined meanings. Some of them have arity two, but user-defined module attributes
must have arity one.
Pre-Defined Module Attributes
Pre-defined module attributes is to be placed before any function declaration.
-module(Module).
Module declaration, defining the name of the module. The name Module, an atom, is to be same as the file name
minus the extension .erl. Otherwise code loading does not work as intended.
This attribute is to be specified first and is the only mandatory attribute.
130 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.5 Modules
-export(Functions).
Exported functions. Specifies which of the functions, defined within the module, that are visible from outside
the module.
Functions is a list [Name1/Arity1, ..., NameN/ArityN], where each NameI is an atom and
ArityI an integer.
-import(Module,Functions).
Imported functions. Can be called the same way as local functions, that is, without any module prefix.
Module, an atom, specifies which module to import functions from. Functions is a list similar as for export.
-compile(Options).
Compiler options. Options is a single option or a list of options. This attribute is added to the option list when
compiling the module. See the compile(3) manual page in Compiler.
-vsn(Vsn).
Module version. Vsn is any literal term and can be retrieved using beam_lib:version/1, see the
beam_lib(3) manual page in STDLIB.
If this attribute is not specified, the version defaults to the MD5 checksum of the module.
-on_load(Function).
This attribute names a function that is to be run automatically when a module is loaded. For more information,
see Running a Function When a Module is Loaded.
Behaviour Module Attribute
It is possible to specify that the module is the callback module for a behaviour:
-behaviour(Behaviour).
The atom Behaviour gives the name of the behaviour, which can be a user-defined behaviour or one of the following
OTP standard behaviours:
•
•
•
•
gen_server
gen_statem
gen_event
supervisor
The spelling behavior is also accepted.
The callback functions of the module can be specified either directly by the exported function behaviour_info/1:
behaviour_info(callbacks) -> Callbacks.
or by a -callback attribute for each callback function:
-callback Name(Arguments) -> Result.
Here, Arguments is a list of zero or more arguments. The -callback attribute is to be preferred since the extra
type information can be used by tools to produce documentation or find discrepancies.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 131
5.5 Modules
Read more about behaviours and callback modules in OTP Design Principles.
Record Definitions
The same syntax as for module attributes is used for record definitions:
-record(Record,Fields).
Record definitions are allowed anywhere in a module, also among the function declarations. Read more in Records.
Preprocessor
The same syntax as for module attributes is used by the preprocessor, which supports file inclusion, macros, and
conditional compilation:
-include("SomeFile.hrl").
-define(Macro,Replacement).
Read more in Preprocessor.
Setting File and Line
The same syntax as for module attributes is used for changing the pre-defined macros ?FILE and ?LINE:
-file(File, Line).
This attribute is used by tools, such as Yecc, to inform the compiler that the source program is generated by another
tool. It also indicates the correspondence of source files to lines of the original user-written file, from which the source
program is produced.
Types and function specifications
A similar syntax as for module attributes is used for specifying types and function specifications:
-type my_type() :: atom() | integer().
-spec my_function(integer()) -> integer().
Read more in Types and Function specifications.
The description is based on EEP8 - Types and function specifications, which is not to be further updated.
5.5.3 Comments
Comments can be placed anywhere in a module except within strings and quoted atoms. A comment begins with the
character "%", continues up to, but does not include the next end-of-line, and has no effect. Notice that the terminating
end-of-line has the effect of white space.
5.5.4 module_info/0 and module_info/1 functions
The compiler automatically inserts the two special, exported functions into each module:
•
Module:module_info/0
132 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.6 Functions
•
Module:module_info/1
These functions can be called to retrieve information about the module.
module_info/0
The module_info/0 function in each module, returns a list of {Key,Value} tuples with information about the
module. Currently, the list contain tuples with the following Keys: module, attributes, compile, exports,
md5 and native. The order and number of tuples may change without prior notice.
module_info/1
The call module_info(Key), where Key is an atom, returns a single piece of information about the module.
The following values are allowed for Key:
module
Returns an atom representing the module name.
attributes
Returns a list of {AttributeName,ValueList} tuples, where AttributeName is the name of an
attribute, and ValueList is a list of values. Notice that a given attribute can occur more than once in the list
with different values if the attribute occurs more than once in the module.
The list of attributes becomes empty if the module is stripped with the beam_lib(3) module (in STDLIB).
compile
Returns a list of tuples with information about how the module was compiled. This list is empty if the module
has been stripped with the beam_lib(3) module (in STDLIB).
md5
Returns a binary representing the MD5 checksum of the module. If the module has native code loaded, this will
be the MD5 of the native code, not the BEAM bytecode.
exports
Returns a list of {Name,Arity} tuples with all exported functions in the module.
functions
Returns a list of {Name,Arity} tuples with all functions in the module.
native
Return true if the module has native compiled code. Return false otherwise. In a system compiled without
HiPE support, the result is always false
5.6 Functions
5.6.1 Function Declaration Syntax
A function declaration is a sequence of function clauses separated by semicolons, and terminated by period (.).
A function clause consists of a clause head and a clause body, separated by ->.
A clause head consists of the function name, an argument list, and an optional guard sequence beginning with the
keyword when:
Name(Pattern11,...,Pattern1N) [when GuardSeq1] ->
Body1;
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 133
5.6 Functions
...;
Name(PatternK1,...,PatternKN) [when GuardSeqK] ->
BodyK.
The function name is an atom. Each argument is a pattern.
The number of arguments N is the arity of the function. A function is uniquely defined by the module name, function
name, and arity. That is, two functions with the same name and in the same module, but with different arities are two
different functions.
A function named f in the module m and with arity N is often denoted as m:f/N.
A clause body consists of a sequence of expressions separated by comma (,):
Expr1,
...,
ExprN
Valid Erlang expressions and guard sequences are described in Expressions.
Example:
fact(N) when N>0 ->
N * fact(N-1);
% first clause head
% first clause body
fact(0) ->
1.
% second clause head
% second clause body
5.6.2 Function Evaluation
When a function m:f/N is called, first the code for the function is located. If the function cannot be found, an undef
runtime error occurs. Notice that the function must be exported to be visible outside the module it is defined in.
If the function is found, the function clauses are scanned sequentially until a clause is found that fulfills both of the
following two conditions:
•
•
The patterns in the clause head can be successfully matched against the given arguments.
The guard sequence, if any, is true.
If such a clause cannot be found, a function_clause runtime error occurs.
If such a clause is found, the corresponding clause body is evaluated. That is, the expressions in the body are evaluated
sequentially and the value of the last expression is returned.
Consider the function fact:
-module(m).
-export([fact/1]).
fact(N) when N>0 ->
N * fact(N-1);
fact(0) ->
1.
Assume that you want to calculate the factorial for 1:
134 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.6 Functions
1> m:fact(1).
Evaluation starts at the first clause. The pattern N is matched against argument 1. The matching succeeds and the guard
(N>0) is true, thus N is bound to 1, and the corresponding body is evaluated:
N * fact(N-1) => (N is bound to 1)
1 * fact(0)
Now, fact(0) is called, and the function clauses are scanned sequentially again. First, the pattern N is matched
against 0. The matching succeeds, but the guard (N>0) is false. Second, the pattern 0 is matched against 0. The matching
succeeds and the body is evaluated:
1 * fact(0) =>
1 * 1 =>
1
Evaluation has succeed and m:fact(1) returns 1.
If m:fact/1 is called with a negative number as argument, no clause head matches. A function_clause runtime
error occurs.
5.6.3 Tail recursion
If the last expression of a function body is a function call, a tail recursive call is done. This is to ensure that no
system resources, for example, call stack, are consumed. This means that an infinite loop can be done if it uses tailrecursive calls.
Example:
loop(N) ->
io:format("~w~n", [N]),
loop(N+1).
The earlier factorial example can act as a counter-example. It is not tail-recursive, since a multiplication is done on
the result of the recursive call to fact(N-1).
5.6.4 Built-In Functions (BIFs)
BIFs are implemented in C code in the runtime system. BIFs do things that are difficult or impossible to implement
in Erlang. Most of the BIFs belong to the module erlang but there are also BIFs belonging to a few other modules,
for example lists and ets.
The most commonly used BIFs belonging to erlang(3) are auto-imported. They do not need to be prefixed with
the module name. Which BIFs that are auto-imported is specified in the erlang(3) module in ERTS. For example,
standard-type conversion BIFs like atom_to_list and BIFs allowed in guards can be called without specifying
the module name.
Examples:
1> tuple_size({a,b,c}).
3
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 135
5.7 Types and Function Specifications
2> atom_to_list('Erlang').
"Erlang"
Notice that it is normally the set of auto-imported BIFs that are referred to when talking about 'BIFs'.
5.7 Types and Function Specifications
5.7.1 The Erlang Type Language
Erlang is a dynamically typed language. Still, it comes with a notation for declaring sets of Erlang terms to form a
particular type. This effectively forms specific subtypes of the set of all Erlang terms.
Subsequently, these types can be used to specify types of record fields and also the argument and return types of
functions.
Type information can be used for the following:
•
•
•
To document function interfaces
To provide more information for bug detection tools, such as Dialyzer
To be exploited by documentation tools, such as EDoc, for generating program documentation of various forms
It is expected that the type language described in this section supersedes and replaces the purely comment-based
@type and @spec declarations used by EDoc.
5.7.2 Types and their Syntax
Types describe sets of Erlang terms. Types consist of, and are built from, a set of predefined types, for example,
integer(), atom(), and pid(). Predefined types represent a typically infinite set of Erlang terms that belong to
this type. For example, the type atom() denotes the set of all Erlang atoms.
For integers and atoms, it is allowed for singleton types; for example, the integers -1 and 42, or the atoms 'foo' and
'bar'. All other types are built using unions of either predefined types or singleton types. In a type union between a
type and one of its subtypes, the subtype is absorbed by the supertype. Thus, the union is then treated as if the subtype
was not a constituent of the union. For example, the type union:
atom() | 'bar' | integer() | 42
describes the same set of terms as the type union:
atom() | integer()
Because of subtype relations that exist between types, types form a lattice where the top-most element, any(), denotes
the set of all Erlang terms and the bottom-most element, none(), denotes the empty set of terms.
The set of predefined types and the syntax for types follows:
Type ::
|
|
|
|
|
|
|
|
any()
none()
pid()
port()
reference()
[]
Atom
Bitstring
float()
%% The top type, the set of all Erlang terms
%% The bottom type, contains no terms
%% nil
136 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.7 Types and Function Specifications
|
|
|
|
|
|
|
Fun
Integer
List
Map
Tuple
Union
UserDefined
%% described in Type Declarations of User-Defined Types
Atom :: atom()
| Erlang_Atom
Bitstring ::
|
|
|
Fun ::
|
|
|
%% 'foo', 'bar', ...
<<>>
<<_:M>>
<<_:_*N>>
<<_:M, _:_*N>>
fun()
fun((...) -> Type)
fun(() -> Type)
fun((TList) -> Type)
%% M is a positive integer
%% N is a positive integer
%% any function
%% any arity, returning Type
Integer :: integer()
| Erlang_Integer
| Erlang_Integer..Erlang_Integer
%% ..., -1, 0, 1, ... 42 ...
%% specifies an integer range
List ::
|
|
|
%%
%%
%%
%%
list(Type)
maybe_improper_list(Type1, Type2)
nonempty_improper_list(Type1, Type2)
nonempty_list(Type)
Proper list ([]-terminated)
Type1=contents, Type2=termination
Type1 and Type2 as above
Proper non-empty list
Map :: map()
| #{}
| #{PairList}
%% denotes a map of any size
%% denotes the empty map
Tuple :: tuple()
| {}
| {TList}
%% denotes a tuple of any size
PairList :: Pair
| Pair, PairList
Pair :: Type := Type
| Type => Type
%% denotes a mandatory pair
%% denotes an optional pair
TList :: Type
| Type, TList
Union :: Type1 | Type2
The general form of bit strings is <<_:M, _:_*N>>, where M and N are positive integers. It denotes a bit string that
is M + (k*N) bits long (that is, a bit string that starts with M bits and continues with k segments of N bits each,
where k is also a positive integer). The notations <<_:_*N>>, <<_:M>>, and <<>> are convenient shorthands for
the cases that M or N, or both, are zero.
Because lists are commonly used, they have shorthand type notations. The types list(T) and
nonempty_list(T) have the shorthands [T] and [T,...], respectively. The only difference between the two
shorthands is that [T] can be an empty list but [T,...] cannot.
Notice that the shorthand for list(), that is, the list of elements of unknown type, is [_] (or [any()]), not [].
The notation [] specifies the singleton type for the empty list.
The general form of maps is #{PairList}. The key types in PairList are allowed to overlap, and if they do,
the leftmost pair takes precedence. A map pair has a key in PairList if it belongs to this type. A PairList may
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 137
5.7 Types and Function Specifications
contain both 'mandatory' and 'optional' pairs where 'mandatory' denotes that a key type, and its associated value type,
must be present. In the case of an 'optional' pair it is not required for the key type to be present.
Notice that the syntactic representation of map() is #{any() => any()} (or #{_ => _}), not #{}. The notation
#{} specifies the singleton type for the empty map.
For convenience, the following types are also built-in. They can be thought as predefined aliases for the type unions
also shown in the table.
Built-in type
Defined as
term()
any()
binary()
<<_:_*8>>
bitstring()
<<_:_*1>>
boolean()
'false' | 'true'
byte()
0..255
char()
0..16#10ffff
nil()
[]
number()
integer() | float()
list()
[any()]
maybe_improper_list()
maybe_improper_list(any(), any())
nonempty_list()
nonempty_list(any())
string()
[char()]
nonempty_string()
[char(),...]
iodata()
iolist() | binary()
iolist()
maybe_improper_list(byte() | binary()
| iolist(), binary() | [])
function()
fun()
module()
atom()
mfa()
{module(),atom(),arity()}
arity()
0..255
identifier()
pid() | port() | reference()
node()
atom()
138 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.7 Types and Function Specifications
timeout()
'infinity' | non_neg_integer()
no_return()
none()
Table 7.1: Built-in types, predefined aliases
In addition, the following three built-in types exist and can be thought as defined below, though strictly their "type
definition" is not valid syntax according to the type language defined above.
Built-in type
Can be thought defined by the syntax
non_neg_integer()
0..
pos_integer()
1..
neg_integer()
..-1
Table 7.2: Additional built-in types
Users are not allowed to define types with the same names as the predefined or built-in ones. This is checked by the
compiler and its violation results in a compilation error.
Note:
The following built-in list types also exist, but they are expected to be rarely used. Hence, they have long names:
nonempty_maybe_improper_list() :: nonempty_maybe_improper_list(any(), any())
nonempty_improper_list(Type1, Type2)
nonempty_maybe_improper_list(Type1, Type2)
where the last two types define the set of Erlang terms one would expect.
Also for convenience, record notation is allowed to be used. Records are shorthands for the corresponding tuples:
Record :: #Erlang_Atom{}
| #Erlang_Atom{Fields}
Records are extended to possibly contain type information. This is described in Type Information in Record
Declarations.
5.7.3 Type Declarations of User-Defined Types
As seen, the basic syntax of a type is an atom followed by closed parentheses. New types are declared using -type
and -opaque attributes as in the following:
-type my_struct_type() :: Type.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 139
5.7 Types and Function Specifications
-opaque my_opaq_type() :: Type.
The type name is the atom my_struct_type, followed by parentheses. Type is a type as defined in the previous
section. A current restriction is that Type can contain only predefined types, or user-defined types which are either
of the following:
•
•
Module-local type, that is, with a definition that is present in the code of the module
Remote type, that is, type defined in, and exported by, other modules; more about this soon.
For module-local types, the restriction that their definition exists in the module is enforced by the compiler and results
in a compilation error. (A similar restriction currently exists for records.)
Type declarations can also be parameterized by including type variables between the parentheses. The syntax of type
variables is the same as Erlang variables, that is, starts with an upper-case letter. Naturally, these variables can - and
is to - appear on the RHS of the definition. A concrete example follows:
-type orddict(Key, Val) :: [{Key, Val}].
A module can export some types to declare that other modules are allowed to refer to them as remote types. This
declaration has the following form:
-export_type([T1/A1, ..., Tk/Ak]).
Here the Ti's are atoms (the name of the type) and the Ai's are their arguments
Example:
-export_type([my_struct_type/0, orddict/2]).
Assuming that these types are exported from module 'mod', you can refer to them from other modules using remote
type expressions like the following:
mod:my_struct_type()
mod:orddict(atom(), term())
It is not allowed to refer to types that are not declared as exported.
Types declared as opaque represent sets of terms whose structure is not supposed to be visible from outside of their
defining module. That is, only the module defining them is allowed to depend on their term structure. Consequently,
such types do not make much sense as module local - module local types are not accessible by other modules anyway
- and is always to be exported.
5.7.4 Type Information in Record Declarations
The types of record fields can be specified in the declaration of the record. The syntax for this is as follows:
-record(rec, {field1 :: Type1, field2, field3 :: Type3}).
140 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.7 Types and Function Specifications
For fields without type annotations, their type defaults to any(). That is, the previous example is a shorthand for the
following:
-record(rec, {field1 :: Type1, field2 :: any(), field3 :: Type3}).
In the presence of initial values for fields, the type must be declared after the initialization, as follows:
-record(rec, {field1 = [] :: Type1, field2, field3 = 42 :: Type3}).
The initial values for fields are to be compatible with (that is, a member of) the corresponding types. This is checked
by the compiler and results in a compilation error if a violation is detected.
Note:
Before Erlang/OTP 19, for fields without initial values, the singleton type 'undefined' was added to all
declared types. In other words, the following two record declarations had identical effects:
-record(rec, {f1 = 42 :: integer(),
f2
:: float(),
f3
:: 'a' | 'b'}).
-record(rec, {f1 = 42 :: integer(),
f2
:: 'undefined' | float(),
f3
:: 'undefined' | 'a' | 'b'}).
This is no longer the case. If you require 'undefined' in your record field type, you must explicitly add it
to the typespec, as in the 2nd example.
Any record, containing type information or not, once defined, can be used as a type using the following syntax:
#rec{}
In addition, the record fields can be further specified when using a record type by adding type information about the
field as follows:
#rec{some_field :: Type}
Any unspecified fields are assumed to have the type in the original record declaration.
5.7.5 Specifications for Functions
A specification (or contract) for a function is given using the -spec attribute. The general format is as follows:
-spec Module:Function(ArgType1, ..., ArgTypeN) -> ReturnType.
The arity of the function must match the number of arguments, else a compilation error occurs.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 141
5.7 Types and Function Specifications
This form can also be used in header files (.hrl) to declare type information for exported functions. Then these header
files can be included in files that (implicitly or explicitly) import these functions.
Within a given module, the following shorthand suffices in most cases:
-spec Function(ArgType1, ..., ArgTypeN) -> ReturnType.
Also, for documentation purposes, argument names can be given:
-spec Function(ArgName1 :: Type1, ..., ArgNameN :: TypeN) -> RT.
A function specification can be overloaded. That is, it can have several types, separated by a semicolon (;):
-spec foo(T1, T2) -> T3
; (T4, T5) -> T6.
A current restriction, which currently results in a warning (not an error) by the compiler, is that the domains of the
argument types cannot overlap. For example, the following specification results in a warning:
-spec foo(pos_integer()) -> pos_integer()
; (integer()) -> integer().
Type variables can be used in specifications to specify relations for the input and output arguments of a function. For
example, the following specification defines the type of a polymorphic identity function:
-spec id(X) -> X.
Notice that the above specification does not restrict the input and output type in any way. These types can be constrained
by guard-like subtype constraints and provide bounded quantification:
-spec id(X) -> X when X :: tuple().
Currently, the :: constraint (read as «is a subtype of») is the only guard constraint that can be used in the 'when'
part of a '-spec' attribute.
142 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
Note:
The above function specification uses multiple occurrences of the same type variable. That provides more type
information than the following function specification, where the type variables are missing:
-spec id(tuple()) -> tuple().
The latter specification says that the function takes some tuple and returns some tuple. The specification with the
X type variable specifies that the function takes a tuple and returns the same tuple.
However, it is up to the tools that process the specifications to choose whether to take this extra information into
account or not.
The scope of a :: constraint is the (...) -> RetType specification after which it appears. To avoid confusion,
it is suggested that different variables are used in different constituents of an overloaded contract, as shown in the
following example:
-spec foo({X, integer()}) -> X when X :: atom()
; ([Y]) -> Y when Y :: number().
Some functions in Erlang are not meant to return; either because they define servers or because they are used to throw
exceptions, as in the following function:
my_error(Err) -> erlang:throw({error, Err}).
For such functions, it is recommended to use the special no_return() type for their "return", through a contract
of the following form:
-spec my_error(term()) -> no_return().
5.8 Expressions
In this section, all valid Erlang expressions are listed. When writing Erlang programs, it is also allowed to use macroand record expressions. However, these expressions are expanded during compilation and are in that sense not true
Erlang expressions. Macro- and record expressions are covered in separate sections:
•
•
Preprocessor
Records
5.8.1 Expression Evaluation
All subexpressions are evaluated before an expression itself is evaluated, unless explicitly stated otherwise. For
example, consider the expression:
Expr1 + Expr2
Expr1 and Expr2, which are also expressions, are evaluated first - in any order - before the addition is performed.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 143
5.8 Expressions
Many of the operators can only be applied to arguments of a certain type. For example, arithmetic operators can only
be applied to numbers. An argument of the wrong type causes a badarg runtime error.
5.8.2 Terms
The simplest form of expression is a term, that is an integer, float, atom, string, list, map, or tuple. The return value
is the term itself.
5.8.3 Variables
A variable is an expression. If a variable is bound to a value, the return value is this value. Unbound variables are
only allowed in patterns.
Variables start with an uppercase letter or underscore (_). Variables can contain alphanumeric characters, underscore
and @.
Examples:
X
Name1
PhoneNumber
Phone_number
_
_Height
Variables are bound to values using pattern matching. Erlang uses single assignment, that is, a variable can only be
bound once.
The anonymous variable is denoted by underscore (_) and can be used when a variable is required but its value can
be ignored.
Example:
[H|_] = [1,2,3]
Variables starting with underscore (_), for example, _Height, are normal variables, not anonymous. They are
however ignored by the compiler in the sense that they do not generate any warnings for unused variables.
Example:
The following code:
member(_, []) ->
[].
can be rewritten to be more readable:
member(Elem, []) ->
[].
This causes a warning for an unused variable, Elem, if the code is compiled with the flag warn_unused_vars
set. Instead, the code can be rewritten to:
144 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
member(_Elem, []) ->
[].
Notice that since variables starting with an underscore are not anonymous, this matches:
{_,_} = {1,2}
But this fails:
{_N,_N} = {1,2}
The scope for a variable is its function clause. Variables bound in a branch of an if, case, or receive expression
must be bound in all branches to have a value outside the expression. Otherwise they are regarded as 'unsafe' outside
the expression.
For the try expression variable scoping is limited so that variables bound in the expression are always 'unsafe' outside
the expression.
5.8.4 Patterns
A pattern has the same structure as a term but can contain unbound variables.
Example:
Name1
[H|T]
{error,Reason}
Patterns are allowed in clause heads, case and receive expressions, and match expressions.
Match Operator = in Patterns
If Pattern1 and Pattern2 are valid patterns, the following is also a valid pattern:
Pattern1 = Pattern2
When matched against a term, both Pattern1 and Pattern2 are matched against the term. The idea behind this
feature is to avoid reconstruction of terms.
Example:
f({connect,From,To,Number,Options}, To) ->
Signal = {connect,From,To,Number,Options},
...;
f(Signal, To) ->
ignore.
can instead be written as
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 145
5.8 Expressions
f({connect,_,To,_,_} = Signal, To) ->
...;
f(Signal, To) ->
ignore.
String Prefix in Patterns
When matching strings, the following is a valid pattern:
f("prefix" ++ Str) -> ...
This is syntactic sugar for the equivalent, but harder to read:
f([$p,$r,$e,$f,$i,$x | Str]) -> ...
Expressions in Patterns
An arithmetic expression can be used within a pattern if it meets both of the following two conditions:
•
•
It uses only numeric or bitwise operators.
Its value can be evaluated to a constant when complied.
Example:
case {Value, Result} of
{?THRESHOLD+1, ok} -> ...
5.8.5 Match
The following matches Expr1, a pattern, against Expr2:
Expr1 = Expr2
If the matching succeeds, any unbound variable in the pattern becomes bound and the value of Expr2 is returned.
If the matching fails, a badmatch run-time error occurs.
Examples:
1> {A, B} = {answer, 42}.
{answer,42}
2> A.
answer
3> {C, D} = [1, 2].
** exception error: no match of right-hand side value [1,2]
5.8.6 Function Calls
146 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
ExprF(Expr1,...,ExprN)
ExprM:ExprF(Expr1,...,ExprN)
In the first form of function calls, ExprM:ExprF(Expr1,...,ExprN), each of ExprM and ExprF must be an
atom or an expression that evaluates to an atom. The function is said to be called by using the fully qualified function
name. This is often referred to as a remote or external function call.
Example:
lists:keysearch(Name, 1, List)
In the second form of function calls, ExprF(Expr1,...,ExprN), ExprF must be an atom or evaluate to a fun.
If ExprF is an atom, the function is said to be called by using the implicitly qualified function name. If the
function ExprF is locally defined, it is called. Alternatively, if ExprF is explicitly imported from the M module,
M:ExprF(Expr1,...,ExprN) is called. If ExprF is neither declared locally nor explicitly imported, ExprF
must be the name of an automatically imported BIF.
Examples:
handle(Msg, State)
spawn(m, init, [])
Examples where ExprF is a fun:
1> Fun1 = fun(X) -> X+1 end,
Fun1(3).
4
2> fun lists:append/2([1,2], [3,4]).
[1,2,3,4]
3>
Notice that when calling a local function, there is a difference between using the implicitly or fully qualified function
name. The latter always refers to the latest version of the module. See Compilation and Code Loading and Function
Evaluation.
Local Function Names Clashing With Auto-Imported BIFs
If a local function has the same name as an auto-imported BIF, the semantics is that implicitly qualified function calls
are directed to the locally defined function, not to the BIF. To avoid confusion, there is a compiler directive available,
-compile({no_auto_import,[F/A]}), that makes a BIF not being auto-imported. In certain situations, such
a compile-directive is mandatory.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 147
5.8 Expressions
Warning:
Before OTP R14A (ERTS version 5.8), an implicitly qualified function call to a function having the same name
as an auto-imported BIF always resulted in the BIF being called. In newer versions of the compiler, the local
function is called instead. This is to avoid that future additions to the set of auto-imported BIFs do not silently
change the behavior of old code.
However, to avoid that old (pre R14) code changed its behavior when compiled with OTP version R14A or later,
the following restriction applies: If you override the name of a BIF that was auto-imported in OTP versions prior
to R14A (ERTS version 5.8) and have an implicitly qualified call to that function in your code, you either need
to explicitly remove the auto-import using a compiler directive, or replace the call with a fully qualified function
call. Otherwise you get a compilation error. See the following example:
-export([length/1,f/1]).
-compile({no_auto_import,[length/1]}). % erlang:length/1 no longer autoimported
length([]) ->
0;
length([H|T]) ->
1 + length(T). %% Calls the local function length/1
f(X) when erlang:length(X) > 3 -> %% Calls erlang:length/1,
%% which is allowed in guards
long.
The same logic applies to explicitly imported functions from other modules, as to locally defined functions. It is not
allowed to both import a function from another module and have the function declared in the module at the same time:
-export([f/1]).
-compile({no_auto_import,[length/1]}). % erlang:length/1 no longer autoimported
-import(mod,[length/1]).
f(X) when erlang:length(X) > 33 -> %% Calls erlang:length/1,
%% which is allowed in guards
erlang:length(X);
f(X) ->
length(X).
%% Explicit call to erlang:length in body
%% mod:length/1 is called
For auto-imported BIFs added in Erlang/OTP R14A and thereafter, overriding the name with a local function or explicit
import is always allowed. However, if the -compile({no_auto_import,[F/A]) directive is not used, the
compiler issues a warning whenever the function is called in the module using the implicitly qualified function name.
5.8.7 If
if
GuardSeq1 ->
Body1;
...;
148 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
end
GuardSeqN ->
BodyN
The branches of an if-expression are scanned sequentially until a guard sequence GuardSeq that evaluates to true
is found. Then the corresponding Body (sequence of expressions separated by ',') is evaluated.
The return value of Body is the return value of the if expression.
If no guard sequence is evaluated as true, an if_clause run-time error occurs. If necessary, the guard expression
true can be used in the last branch, as that guard sequence is always true.
Example:
is_greater_than(X, Y) ->
if
X>Y ->
true;
true -> % works as an 'else' branch
false
end
5.8.8 Case
case Expr of
Pattern1 [when GuardSeq1] ->
Body1;
...;
PatternN [when GuardSeqN] ->
BodyN
end
The expression Expr is evaluated and the patterns Pattern are sequentially matched against the result. If a match
succeeds and the optional guard sequence GuardSeq is true, the corresponding Body is evaluated.
The return value of Body is the return value of the case expression.
If there is no matching pattern with a true guard sequence, a case_clause run-time error occurs.
Example:
is_valid_signal(Signal) ->
case Signal of
{signal, _What, _From, _To} ->
true;
{signal, _What, _To} ->
true;
_Else ->
false
end.
5.8.9 Send
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 149
5.8 Expressions
Expr1 ! Expr2
Sends the value of Expr2 as a message to the process specified by Expr1. The value of Expr2 is also the return
value of the expression.
Expr1 must evaluate to a pid, a registered name (atom), or a tuple {Name,Node}. Name is an atom and Node is
a node name, also an atom.
•
•
•
If Expr1 evaluates to a name, but this name is not registered, a badarg run-time error occurs.
Sending a message to a pid never fails, even if the pid identifies a non-existing process.
Distributed message sending, that is, if Expr1 evaluates to a tuple {Name,Node} (or a pid located at another
node), also never fails.
5.8.10 Receive
receive
Pattern1 [when GuardSeq1] ->
Body1;
...;
PatternN [when GuardSeqN] ->
BodyN
end
Receives messages sent to the process using the send operator (!). The patterns Pattern are sequentially matched
against the first message in time order in the mailbox, then the second, and so on. If a match succeeds and the optional
guard sequence GuardSeq is true, the corresponding Body is evaluated. The matching message is consumed, that
is, removed from the mailbox, while any other messages in the mailbox remain unchanged.
The return value of Body is the return value of the receive expression.
receive never fails. The execution is suspended, possibly indefinitely, until a message arrives that matches one of
the patterns and with a true guard sequence.
Example:
wait_for_onhook() ->
receive
onhook ->
disconnect(),
idle();
{connect, B} ->
B ! {busy, self()},
wait_for_onhook()
end.
The receive expression can be augmented with a timeout:
receive
Pattern1 [when GuardSeq1] ->
Body1;
...;
PatternN [when GuardSeqN] ->
BodyN
after
ExprT ->
150 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
end
BodyT
ExprT is to evaluate to an integer. The highest allowed value is 16#FFFFFFFF, that is, the value must fit in 32
bits. receive..after works exactly as receive, except that if no matching message has arrived within ExprT
milliseconds, then BodyT is evaluated instead. The return value of BodyT then becomes the return value of the
receive..after expression.
Example:
wait_for_onhook() ->
receive
onhook ->
disconnect(),
idle();
{connect, B} ->
B ! {busy, self()},
wait_for_onhook()
after
60000 ->
disconnect(),
error()
end.
It is legal to use a receive..after expression with no branches:
receive
after
ExprT ->
BodyT
end
This construction does not consume any messages, only suspends execution in the process for ExprT milliseconds.
This can be used to implement simple timers.
Example:
timer() ->
spawn(m, timer, [self()]).
timer(Pid) ->
receive
after
5000 ->
Pid ! timeout
end.
There are two special cases for the timeout value ExprT:
infinity
The process is to wait indefinitely for a matching message; this is the same as not using a timeout. This can be
useful for timeout values that are calculated at runtime.
0
If there is no matching message in the mailbox, the timeout occurs immediately.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 151
5.8 Expressions
5.8.11 Term Comparisons
Expr1 op Expr2
op
Description
==
Equal to
/=
Not equal to
=<
Less than or equal to
<
Less than
>=
Greater than or equal to
>
Greater than
=:=
Exactly equal to
=/=
Exactly not equal to
Table 8.1: Term Comparison Operators.
The arguments can be of different data types. The following order is defined:
number < atom < reference < fun < port < pid < tuple < map < nil < list < bit string
Lists are compared element by element. Tuples are ordered by size, two tuples with the same size are compared element
by element.
Maps are ordered by size, two maps with the same size are compared by keys in ascending term order and then by
values in key order. In maps key order integers types are considered less than floats types.
When comparing an integer to a float, the term with the lesser precision is converted into the type of the other term,
unless the operator is one of =:= or =/=. A float is more precise than an integer until all significant figures of the
float are to the left of the decimal point. This happens when the float is larger/smaller than +/-9007199254740992.0.
The conversion strategy is changed depending on the size of the float because otherwise comparison of large floats
and integers would lose their transitivity.
Term comparison operators return the Boolean value of the expression, true or false.
Examples:
1> 1==1.0.
true
2> 1=:=1.0.
false
3> 1 > a.
false
4> #{c => 3} > #{a => 1, b => 2}.
152 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
false
4> #{a => 1, b => 2} == #{a => 1.0, b => 2.0}.
true
5.8.12 Arithmetic Expressions
op Expr
Expr1 op Expr2
Operator
Description
Argument Type
+
Unary +
Number
-
Unary -
Number
+
number
-
Number
*
Number
/
Floating point division
Number
bnot
Unary bitwise NOT
Integer
div
Integer division
Integer
rem
Integer remainder of X/Y
Integer
band
Bitwise AND
Integer
bor
Bitwise OR
Integer
bxor
Arithmetic bitwise XOR
Integer
bsl
Arithmetic bitshift left
Integer
bsr
Bitshift right
Integer
Table 8.2: Arithmetic Operators.
Examples:
1> +1.
1
2> -1.
-1
3> 1+1.
2
4> 4/2.
2.0
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 153
5.8 Expressions
5>
2
6>
1
7>
0
8>
3
9>
**
5 div 2.
5 rem 2.
2#10 band 2#01.
2#10 bor 2#01.
a + 10.
exception error: an error occurred when evaluating an arithmetic expression
in operator +/2
called as a + 10
10> 1 bsl (1 bsl 64).
** exception error: a system limit has been reached
in operator bsl/2
called as 1 bsl 18446744073709551616
5.8.13 Boolean Expressions
op Expr
Expr1 op Expr2
Operator
Description
not
Unary logical NOT
and
Logical AND
or
Logical OR
xor
Logical XOR
Table 8.3: Logical Operators.
Examples:
1> not true.
false
2> true and false.
false
3> true xor false.
true
4> true or garbage.
** exception error: bad argument
in operator or/2
called as true or garbage
5.8.14 Short-Circuit Expressions
Expr1 orelse Expr2
Expr1 andalso Expr2
154 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
Expr2 is evaluated only if necessary. That is, Expr2 is evaluated only if:
•
Expr1 evaluates to false in an orelse expression.
or
•
Expr1 evaluates to true in an andalso expression.
Returns either the value of Expr1 (that is, true or false) or the value of Expr2 (if Expr2 is evaluated).
Example 1:
case A >= -1.0 andalso math:sqrt(A+1) > B of
This works even if A is less than -1.0, since in that case, math:sqrt/1 is never evaluated.
Example 2:
OnlyOne = is_atom(L) orelse
(is_list(L) andalso length(L) == 1),
From Erlang/OTP R13A, Expr2 is no longer required to evaluate to a Boolean value. As a consequence, andalso
and orelse are now tail-recursive. For instance, the following function is tail-recursive in Erlang/OTP R13A and
later:
all(Pred, [Hd|Tail]) ->
Pred(Hd) andalso all(Pred, Tail);
all(_, []) ->
true.
5.8.15 List Operations
Expr1 ++ Expr2
Expr1 -- Expr2
The list concatenation operator ++ appends its second argument to its first and returns the resulting list.
The list subtraction operator -- produces a list that is a copy of the first argument. The procedure is a follows: for
each element in the second argument, the first occurrence of this element (if any) is removed.
Example:
1> [1,2,3]++[4,5].
[1,2,3,4,5]
2> [1,2,3,2,1,2]--[2,1,2].
[3,1,2]
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 155
5.8 Expressions
Warning:
The complexity of A -- B is proportional to length(A)*length(B). That is, it becomes very slow if both
A and B are long lists.
5.8.16 Map Expressions
Creating Maps
Constructing a new map is done by letting an expression K be associated with another expression V:
#{ K => V }
New maps can include multiple associations at construction by listing every association:
#{ K1 => V1, .., Kn => Vn }
An empty map is constructed by not associating any terms with each other:
#{}
All keys and values in the map are terms. Any expression is first evaluated and then the resulting terms are used as
key and value respectively.
Keys and values are separated by the => arrow and associations are separated by a comma ,.
Examples:
M0
M1
M2
M3
M4
=
=
=
=
=
#{},
#{a => <<"hello">>},
#{1 => 2, b => b},
#{k => {A,B}},
#{{"w", 1} => f()}.
%
%
%
%
%
empty map
single association with literals
multiple associations with literals
single association with variables
compound key associated with an evaluated expression
Here, A and B are any expressions and M0 through M4 are the resulting map terms.
If two matching keys are declared, the latter key takes precedence.
Example:
1> #{1 => a, 1 => b}.
#{1 => b }
2> #{1.0 => a, 1 => b}.
#{1 => b, 1.0 => a}
The order in which the expressions constructing the keys (and their associated values) are evaluated is not defined.
The syntactic order of the key-value pairs in the construction is of no relevance, except in the recently mentioned case
of two matching keys.
Updating Maps
Updating a map has a similar syntax as constructing it.
156 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
An expression defining the map to be updated, is put in front of the expression defining the keys to be updated and
their respective values:
M#{ K => V }
Here M is a term of type map and K and V are any expression.
If key K does not match any existing key in the map, a new association is created from key K to value V.
If key K matches an existing key in map M, its associated value is replaced by the new value V. In both cases, the
evaluated map expression returns a new map.
If M is not of type map, an exception of type badmap is thrown.
To only update an existing value, the following syntax is used:
M#{ K := V }
Here M is a term of type map, V is an expression and K is an expression that evaluates to an existing key in M.
If key K does not match any existing keys in map M, an exception of type badarg is triggered at runtime. If a matching
key K is present in map M, its associated value is replaced by the new value V, and the evaluated map expression
returns a new map.
If M is not of type map, an exception of type badmap is thrown.
Examples:
M0
M1
M2
M3
M4
=
=
=
=
=
#{},
M0#{a => 0},
M1#{a => 1, b => 2},
M2#{"function" => fun() -> f() end},
M3#{a := 2, b := 3}. % 'a' and 'b' was added in `M1` and `M2`.
Here M0 is any map. It follows that M1 .. M4 are maps as well.
More Examples:
1> M = #{1 => a}.
#{1 => a }
2> M#{1.0 => b}.
#{1 => a, 1.0 => b}.
3> M#{1 := b}.
#{1 => b}
4> M#{1.0 := b}.
** exception error: bad argument
As in construction, the order in which the key and value expressions are evaluated is not defined. The syntactic order
of the key-value pairs in the update is of no relevance, except in the case where two keys match. In that case, the
latter value is used.
Maps in Patterns
Matching of key-value associations from maps is done as follows:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 157
5.8 Expressions
#{ K := V } = M
Here M is any map. The key K must be an expression with bound variables or literals. V can be any pattern with either
bound or unbound variables.
If the variable V is unbound, it becomes bound to the value associated with the key K, which must exist in the map M.
If the variable V is bound, it must match the value associated with K in M.
Example:
1> M = #{"tuple" => {1,2}}.
#{"tuple" => {1,2}}
2> #{"tuple" := {1,B}} = M.
#{"tuple" => {1,2}}
3> B.
2.
This binds variable B to integer 2.
Similarly, multiple values from the map can be matched:
#{ K1 := V1, .., Kn := Vn } = M
Here keys K1 .. Kn are any expressions with literals or bound variables. If all keys exist in map M, all variables in
V1 .. Vn is matched to the associated values of their respective keys.
If the matching conditions are not met, the match fails, either with:
•
A badmatch exception.
•
This is if it is used in the context of the match operator as in the example.
Or resulting in the next clause being tested in function heads and case expressions.
Matching in maps only allows for := as delimiters of associations.
The order in which keys are declared in matching has no relevance.
Duplicate keys are allowed in matching and match each pattern associated to the keys:
#{ K := V1, K := V2 } = M
Matching an expression against an empty map literal, matches its type but no variables are bound:
#{} = Expr
This expression matches if the expression Expr is of type map, otherwise it fails with an exception badmatch.
Matching Syntax
Matching of literals as keys are allowed in function heads:
%% only start if not_started
handle_call(start, From, #{ state := not_started } = S) ->
...
{reply, ok, S#{ state := start }};
158 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
%% only change if started
handle_call(change, From, #{ state := start } = S) ->
...
{reply, ok, S#{ state := changed }};
Maps in Guards
Maps are allowed in guards as long as all subexpressions are valid guard expressions.
Two guard BIFs handle maps:
•
•
is_map/1 in the erlang module
map_size/1 in the erlang module
5.8.17 Bit Syntax Expressions
<<>>
<<E1,...,En>>
Each element Ei specifies a segment of the bit string. Each element Ei is a value, followed by an optional size
expression and an optional type specifier list.
Ei = Value |
Value:Size |
Value/TypeSpecifierList |
Value:Size/TypeSpecifierList
Used in a bit string construction, Value is an expression that is to evaluate to an integer, float, or bit string. If the
expression is not a single literal or variable, it is to be enclosed in parentheses.
Used in a bit string matching, Value must be a variable, or an integer, float, or string.
Notice that, for example, using a string literal as in <<"abc">> is syntactic sugar for <<$a,$b,$c>>.
Used in a bit string construction, Size is an expression that is to evaluate to an integer.
Used in a bit string matching, Size must be an integer, or a variable bound to an integer.
The value of Size specifies the size of the segment in units (see below). The default value depends on the type (see
below):
•
•
•
For integer it is 8.
For float it is 64.
For binary and bitstring it is the whole binary or bit string.
In matching, this default value is only valid for the last element. All other bit string or binary elements in the matching
must have a size specification.
For the utf8, utf16, and utf32 types, Size must not be given. The size of the segment is implicitly determined
by the type and value itself.
TypeSpecifierList is a list of type specifiers, in any order, separated by hyphens (-). Default values are used
for any omitted type specifiers.
Type= integer | float | binary | bytes | bitstring | bits | utf8 | utf16 | utf32
The default is integer. bytes is a shorthand for binary and bits is a shorthand for bitstring. See
below for more information about the utf types.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 159
5.8 Expressions
Signedness= signed | unsigned
Only matters for matching and when the type is integer. The default is unsigned.
Endianness= big | little | native
Native-endian means that the endianness is resolved at load time to be either big-endian or little-endian,
depending on what is native for the CPU that the Erlang machine is run on. Endianness only matters when the
Type is either integer, utf16, utf32, or float. The default is big.
Unit= unit:IntegerLiteral
The allowed range is 1..256. Defaults to 1 for integer, float, and bitstring, and to 8 for binary. No
unit specifier must be given for the types utf8, utf16, and utf32.
The value of Size multiplied with the unit gives the number of bits. A segment of type binary must have a size
that is evenly divisible by 8.
Note:
When constructing binaries, if the size N of an integer segment is too small to contain the given integer, the most
significant bits of the integer are silently discarded and only the N least significant bits are put into the binary.
The types utf8, utf16, and utf32 specifies encoding/decoding of the Unicode Transformation Formats UTF-8,
UTF-16, and UTF-32, respectively.
When constructing a segment of a utf type, Value must be an integer in the range 0..16#D7FF or
16#E000....16#10FFFF. Construction fails with a badarg exception if Value is outside the allowed ranges. The size
of the resulting binary segment depends on the type or Value, or both:
•
•
•
For utf8, Value is encoded in 1-4 bytes.
For utf16, Value is encoded in 2 or 4 bytes.
For utf32, Value is always be encoded in 4 bytes.
When constructing, a literal string can be given followed by one of the UTF types, for example: <<"abc"/utf8>>
which is syntactic sugar for <<$a/utf8,$b/utf8,$c/utf8>>.
A successful match of a segment of a utf type, results in an integer in the range 0..16#D7FF or 16#E000..16#10FFFF.
The match fails if the returned value falls outside those ranges.
A segment of type utf8 matches 1-4 bytes in the binary, if the binary at the match position contains a valid UTF-8
sequence. (See RFC-3629 or the Unicode standard.)
A segment of type utf16 can match 2 or 4 bytes in the binary. The match fails if the binary at the match position
does not contain a legal UTF-16 encoding of a Unicode code point. (See RFC-2781 or the Unicode standard.)
A segment of type utf32 can match 4 bytes in the binary in the same way as an integer segment matches 32 bits.
The match fails if the resulting integer is outside the legal ranges mentioned above.
Examples:
1> Bin1 = <<1,17,42>>.
<<1,17,42>>
2> Bin2 = <<"abc">>.
<<97,98,99>>
3> Bin3 = <<1,17,42:16>>.
<<1,17,0,42>>
4> <<A,B,C:16>> = <<1,17,42:16>>.
<<1,17,0,42>>
5> C.
42
160 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
6> <<D:16,E,F>> = <<1,17,42:16>>.
<<1,17,0,42>>
7> D.
273
8> F.
42
9> <<G,H/binary>> = <<1,17,42:16>>.
<<1,17,0,42>>
10> H.
<<17,0,42>>
11> <<G,H/bitstring>> = <<1,17,42:12>>.
<<1,17,1,10:4>>
12> H.
<<17,1,10:4>>
13> <<1024/utf8>>.
<<208,128>>
Notice that bit string patterns cannot be nested.
Notice also that "B=<<1>>" is interpreted as "B =<<1>>" which is a syntax error. The correct way is to write a
space after '=': "B= <<1>>.
More examples are provided in Programming Examples.
5.8.18 Fun Expressions
fun
end
[Name](Pattern11,...,Pattern1N) [when GuardSeq1] ->
Body1;
...;
[Name](PatternK1,...,PatternKN) [when GuardSeqK] ->
BodyK
A fun expression begins with the keyword fun and ends with the keyword end. Between them is to be a function
declaration, similar to a regular function declaration, except that the function name is optional and is to be a variable,
if any.
Variables in a fun head shadow the function name and both shadow variables in the function clause surrounding the
fun expression. Variables bound in a fun body are local to the fun body.
The return value of the expression is the resulting fun.
Examples:
1> Fun1 = fun (X) -> X+1 end.
#Fun<erl_eval.6.39074546>
2> Fun1(2).
3
3> Fun2 = fun (X) when X>=5 -> gt; (X) -> lt end.
#Fun<erl_eval.6.39074546>
4> Fun2(7).
gt
5> Fun3 = fun Fact(1) -> 1; Fact(X) when X > 1 -> X * Fact(X - 1) end.
#Fun<erl_eval.6.39074546>
6> Fun3(4).
24
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 161
5.8 Expressions
The following fun expressions are also allowed:
fun Name/Arity
fun Module:Name/Arity
In Name/Arity, Name is an atom and Arity is an integer. Name/Arity must specify an existing local function.
The expression is syntactic sugar for:
fun (Arg1,...,ArgN) -> Name(Arg1,...,ArgN) end
In Module:Name/Arity, Module, and Name are atoms and Arity is an integer. Starting from Erlang/OTP R15,
Module, Name, and Arity can also be variables. A fun defined in this way refers to the function Name with arity
Arity in the latest version of module Module. A fun defined in this way is not dependent on the code for the module
in which it is defined.
More examples are provided in Programming Examples.
5.8.19 Catch and Throw
catch Expr
Returns the value of Expr unless an exception occurs during the evaluation. In that case, the exception is caught.
For exceptions of class error, that is, run-time errors, {'EXIT',{Reason,Stack}} is returned.
For exceptions of class exit, that is, the code called exit(Term), {'EXIT',Term} is returned.
For exceptions of class throw, that is the code called throw(Term), Term is returned.
Reason depends on the type of error that occurred, and Stack is the stack of recent function calls, see Exit Reasons.
Examples:
1> catch 1+2.
3
2> catch 1+a.
{'EXIT',{badarith,[...]}}
Notice that catch has low precedence and catch subexpressions often needs to be enclosed in a block expression
or in parentheses:
3> A = catch 1+2.
** 1: syntax error before: 'catch' **
4> A = (catch 1+2).
3
The BIF throw(Any) can be used for non-local return from a function. It must be evaluated within a catch, which
returns the value Any.
Example:
162 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
5> catch throw(hello).
hello
If throw/1 is not evaluated within a catch, a nocatch run-time error occurs.
5.8.20 Try
try Exprs
catch
[Class1:]ExceptionPattern1 [when ExceptionGuardSeq1] ->
ExceptionBody1;
[ClassN:]ExceptionPatternN [when ExceptionGuardSeqN] ->
ExceptionBodyN
end
This is an enhancement of catch. It gives the possibility to:
•
•
•
Distinguish between different exception classes.
Choose to handle only the desired ones.
Passing the others on to an enclosing try or catch, or to default error handling.
Notice that although the keyword catch is used in the try expression, there is not a catch expression within the
try expression.
It returns the value of Exprs (a sequence of expressions Expr1, ..., ExprN) unless an exception occurs
during the evaluation. In that case the exception is caught and the patterns ExceptionPattern with the right
exception class Class are sequentially matched against the caught exception. An omitted Class is shorthand for
throw. If a match succeeds and the optional guard sequence ExceptionGuardSeq is true, the corresponding
ExceptionBody is evaluated to become the return value.
If an exception occurs during evaluation of Exprs but there is no matching ExceptionPattern of the right
Class with a true guard sequence, the exception is passed on as if Exprs had not been enclosed in a try expression.
If an exception occurs during evaluation of ExceptionBody, it is not caught.
The try expression can have an of section:
try Exprs of
Pattern1 [when GuardSeq1] ->
Body1;
...;
PatternN [when GuardSeqN] ->
BodyN
catch
[Class1:]ExceptionPattern1 [when ExceptionGuardSeq1] ->
ExceptionBody1;
...;
[ClassN:]ExceptionPatternN [when ExceptionGuardSeqN] ->
ExceptionBodyN
end
If the evaluation of Exprs succeeds without an exception, the patterns Pattern are sequentially matched against the
result in the same way as for a case expression, except that if the matching fails, a try_clause run-time error occurs.
An exception occurring during the evaluation of Body is not caught.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 163
5.8 Expressions
The try expression can also be augmented with an after section, intended to be used for cleanup with side effects:
try Exprs of
Pattern1 [when GuardSeq1] ->
Body1;
...;
PatternN [when GuardSeqN] ->
BodyN
catch
[Class1:]ExceptionPattern1 [when ExceptionGuardSeq1] ->
ExceptionBody1;
...;
[ClassN:]ExceptionPatternN [when ExceptionGuardSeqN] ->
ExceptionBodyN
after
AfterBody
end
AfterBody is evaluated after either Body or ExceptionBody, no matter which one. The evaluated value of
AfterBody is lost; the return value of the try expression is the same with an after section as without.
Even if an exception occurs during evaluation of Body or ExceptionBody, AfterBody is evaluated. In this case
the exception is passed on after AfterBody has been evaluated, so the exception from the try expression is the
same with an after section as without.
If an exception occurs during evaluation of AfterBody itself, it is not caught. So if AfterBody is evaluated after an
exception in Exprs, Body, or ExceptionBody, that exception is lost and masked by the exception in AfterBody.
The of, catch, and after sections are all optional, as long as there is at least a catch or an after section. So
the following are valid try expressions:
try Exprs of
Pattern when GuardSeq ->
Body
after
AfterBody
end
try Exprs
catch
ExpressionPattern ->
ExpressionBody
after
AfterBody
end
try Exprs after AfterBody end
Next is an example of using after. This closes the file, even in the event of exceptions in file:read/2 or in
binary_to_term/1. The exceptions are the same as without the try...after...end expression:
termize_file(Name) ->
{ok,F} = file:open(Name, [read,binary]),
try
{ok,Bin} = file:read(F, 1024*1024),
binary_to_term(Bin)
after
164 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
file:close(F)
end.
Next is an example of using try to emulate catch Expr:
try Expr
catch
throw:Term -> Term;
exit:Reason -> {'EXIT',Reason}
error:Reason -> {'EXIT',{Reason,erlang:get_stacktrace()}}
end
5.8.21 Parenthesized Expressions
(Expr)
Parenthesized expressions are useful to override operator precedences, for example, in arithmetic expressions:
1> 1 + 2 * 3.
7
2> (1 + 2) * 3.
9
5.8.22 Block Expressions
begin
Expr1,
...,
ExprN
end
Block expressions provide a way to group a sequence of expressions, similar to a clause body. The return value is the
value of the last expression ExprN.
5.8.23 List Comprehensions
List comprehensions is a feature of many modern functional programming languages. Subject to certain rules, they
provide a succinct notation for generating elements in a list.
List comprehensions are analogous to set comprehensions in Zermelo-Frankel set theory and are called ZF expressions
in Miranda. They are analogous to the setof and findall predicates in Prolog.
List comprehensions are written with the following syntax:
[Expr || Qualifier1,...,QualifierN]
Here, Expr is an arbitrary expression, and each Qualifier is either a generator or a filter.
•
A generator is written as:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 165
5.8 Expressions
•
•
Pattern <- ListExpr.
ListExpr must be an expression, which evaluates to a list of terms.
A bit string generator is written as:
BitstringPattern <= BitStringExpr.
BitStringExpr must be an expression, which evaluates to a bitstring.
A filter is an expression, which evaluates to true or false.
The variables in the generator patterns, shadow variables in the function clause, surrounding the list comprehensions.
A list comprehension returns a list, where the elements are the result of evaluating Expr for each combination of
generator list elements and bit string generator elements, for which all filters are true.
Example:
1> [X*2 || X <- [1,2,3]].
[2,4,6]
When there are no generators or bit string generators, a list comprehension returns either a list with one element (the
result of evaluating Expr) if all filters are true or an empty list otherwise.
Example:
1> [2 || is_integer(2)].
[2]
2> [x || is_integer(x)].
[]
More examples are provided in Programming Examples.
5.8.24 Bit String Comprehensions
Bit string comprehensions are analogous to List Comprehensions. They are used to generate bit strings efficiently and
succinctly.
Bit string comprehensions are written with the following syntax:
<< BitStringExpr || Qualifier1,...,QualifierN >>
BitStringExpr is an expression that evalutes to a bit string. If BitStringExpr is a function call, it must be
enclosed in parentheses. Each Qualifier is either a generator, a bit string generator or a filter.
•
•
•
A generator is written as:
Pattern <- ListExpr.
ListExpr must be an expression that evaluates to a list of terms.
A bit string generator is written as:
BitstringPattern <= BitStringExpr.
BitStringExpr must be an expression that evaluates to a bitstring.
A filter is an expression that evaluates to true or false.
The variables in the generator patterns, shadow variables in the function clause, surrounding the bit string
comprehensions.
166 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
A bit string comprehension returns a bit string, which is created by concatenating the results of evaluating BitString
for each combination of bit string generator elements, for which all filters are true.
Example:
1> << << (X*2) >> ||
<<X>> <= << 1,2,3 >> >>.
<<2,4,6>>
More examples are provided in Programming Examples.
5.8.25 Guard Sequences
A guard sequence is a sequence of guards, separated by semicolon (;). The guard sequence is true if at least one of
the guards is true. (The remaining guards, if any, are not evaluated.)
Guard1;...;GuardK
A guard is a sequence of guard expressions, separated by comma (,). The guard is true if all guard expressions evaluate
to true.
GuardExpr1,...,GuardExprN
The set of valid guard expressions (sometimes called guard tests) is a subset of the set of valid Erlang expressions.
The reason for restricting the set of valid expressions is that evaluation of a guard expression must be guaranteed to
be free of side effects. Valid guard expressions are the following:
•
•
•
•
•
•
•
The atom true
Other constants (terms and bound variables), all regarded as false
Calls to the BIFs specified in table Type Test BIFs
Term comparisons
Arithmetic expressions
Boolean expressions
Short-circuit expressions (andalso/orelse)
is_atom/1
is_binary/1
is_bitstring/1
is_boolean/1
is_float/1
is_function/1
is_function/2
is_integer/1
is_list/1
is_map/1
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 167
5.8 Expressions
is_number/1
is_pid/1
is_port/1
is_record/2
is_record/3
is_reference/1
is_tuple/1
Table 8.4: Type Test BIFs
Notice that most type test BIFs have older equivalents, without the is_ prefix. These old BIFs are retained for
backwards compatibility only and are not to be used in new code. They are also only allowed at top level. For example,
they are not allowed in Boolean expressions in guards.
abs(Number)
bit_size(Bitstring)
byte_size(Bitstring)
element(N, Tuple)
float(Term)
hd(List)
length(List)
map_size(Map)
node()
node(Pid|Ref|Port)
round(Number)
self()
size(Tuple|Bitstring)
tl(List)
trunc(Number)
168 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.8 Expressions
tuple_size(Tuple)
Table 8.5: Other BIFs Allowed in Guard Expressions
If an arithmetic expression, a Boolean expression, a short-circuit expression, or a call to a guard BIF fails (because of
invalid arguments), the entire guard fails. If the guard was part of a guard sequence, the next guard in the sequence
(that is, the guard following the next semicolon) is evaluated.
5.8.26 Operator Precedence
Operator precedence in falling priority:
:
#
Unary + - bnot not
/ * div rem band and
Left associative
+ - bor bxor bsl bsr or xor
Left associative
++ --
Right associative
== /= =< < >= > =:= =/=
andalso
orelse
=!
Right associative
catch
Table 8.6: Operator Precedence
When evaluating an expression, the operator with the highest priority is evaluated first. Operators with the same priority
are evaluated according to their associativity.
Example:
The left associative arithmetic operators are evaluated left to right:
6 + 5 * 4 - 3 / 2 evaluates to
6 + 20 - 1.5 evaluates to
26 - 1.5 evaluates to
24.5
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 169
5.9 Preprocessor
5.9 Preprocessor
5.9.1 File Inclusion
A file can be included as follows:
-include(File).
-include_lib(File).
File, a string, is to point out a file. The contents of this file are included as is, at the position of the directive.
Include files are typically used for record and macro definitions that are shared by several modules. It is recommended
to use the file name extension .hrl for include files.
File can start with a path component $VAR, for some string VAR. If that is the case, the value of the environment
variable VAR as returned by os:getenv(VAR) is substituted for $VAR. If os:getenv(VAR) returns false,
$VAR is left as is.
If the filename File is absolute (possibly after variable substitution), the include file with that name is included.
Otherwise, the specified file is searched for in the following directories, and in this order:
•
•
•
The current working directory
The directory where the module is being compiled
The directories given by the include option
For details, see the erlc(1) manual page in ERTS and compile(3) manual page in Compiler.
Examples:
-include("my_records.hrl").
-include("incdir/my_records.hrl").
-include("/home/user/proj/my_records.hrl").
-include("$PROJ_ROOT/my_records.hrl").
include_lib is similar to include, but is not to point out an absolute file. Instead, the first path component
(possibly after variable substitution) is assumed to be the name of an application.
Example:
-include_lib("kernel/include/file.hrl").
The code server uses code:lib_dir(kernel) to find the directory of the current (latest) version of Kernel, and
then the subdirectory include is searched for the file file.hrl.
5.9.2 Defining and Using Macros
A macro is defined as follows:
-define(Const, Replacement).
-define(Func(Var1,...,VarN), Replacement).
170 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.9 Preprocessor
A macro definition can be placed anywhere among the attributes and function declarations of a module, but the
definition must come before any usage of the macro.
If a macro is used in several modules, it is recommended that the macro definition is placed in an include file.
A macro is used as follows:
?Const
?Func(Arg1,...,ArgN)
Macros are expanded during compilation. A simple macro ?Const is replaced with Replacement.
Example:
-define(TIMEOUT, 200).
...
call(Request) ->
server:call(refserver, Request, ?TIMEOUT).
This is expanded to:
call(Request) ->
server:call(refserver, Request, 200).
A macro ?Func(Arg1,...,ArgN) is replaced with Replacement, where all occurrences of a variable Var
from the macro definition are replaced with the corresponding argument Arg.
Example:
-define(MACRO1(X, Y), {a, X, b, Y}).
...
bar(X) ->
?MACRO1(a, b),
?MACRO1(X, 123)
This is expanded to:
bar(X) ->
{a,a,b,b},
{a,X,b,123}.
It is good programming practice, but not mandatory, to ensure that a macro definition is a valid Erlang syntactic form.
To view the result of macro expansion, a module can be compiled with the 'P' option. compile:file(File,
['P']). This produces a listing of the parsed code after preprocessing and parse transforms, in the file File.P.
5.9.3 Predefined Macros
The following macros are predefined:
?MODULE
The name of the current module.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 171
5.9 Preprocessor
?MODULE_STRING.
The name of the current module, as a string.
?FILE.
The file name of the current module.
?LINE.
The current line number.
?MACHINE.
The machine name, 'BEAM'.
?FUNCTION_NAME
The name of the current function.
?FUNCTION_ARITY
The arity (number of arguments) for the current function.
5.9.4 Macros Overloading
It is possible to overload macros, except for predefined macros. An overloaded macro has more than one definition,
each with a different number of arguments.
The feature was added in Erlang 5.7.5/OTP R13B04.
A macro ?Func(Arg1,...,ArgN) with a (possibly empty) list of arguments results in an error message if there
is at least one definition of Func with arguments, but none with N arguments.
Assuming these definitions:
-define(F0(), c).
-define(F1(A), A).
-define(C, m:f).
the following does not work:
f0() ->
?F0. % No, an empty list of arguments expected.
f1(A) ->
?F1(A, A). % No, exactly one argument expected.
On the other hand,
f() ->
?C().
is expanded to
f() ->
m:f().
5.9.5 Flow Control in Macros
The following macro directives are supplied:
172 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.9 Preprocessor
-undef(Macro).
Causes the macro to behave as if it had never been defined.
-ifdef(Macro).
Evaluate the following lines only if Macro is defined.
-ifndef(Macro).
Evaluate the following lines only if Macro is not defined.
-else.
Only allowed after an ifdef or ifndef directive. If that condition is false, the lines following else are
evaluated instead.
-endif.
Specifies the end of an ifdef or ifndef directive.
Note:
The macro directives cannot be used inside functions.
Example:
-module(m).
...
-ifdef(debug).
-define(LOG(X), io:format("{~p,~p}: ~p~n", [?MODULE,?LINE,X])).
-else.
-define(LOG(X), true).
-endif.
...
When trace output is desired, debug is to be defined when the module m is compiled:
% erlc -Ddebug m.erl
or
1> c(m, {d, debug}).
{ok,m}
?LOG(Arg) is then expanded to a call to io:format/2 and provide the user with some simple trace output.
5.9.6 -error() and -warning() directives
The directive -error(Term) causes a compilation error.
Example:
-module(t).
-export([version/0]).
-ifdef(VERSION).
version() -> ?VERSION.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 173
5.9 Preprocessor
-else.
-error("Macro VERSION must be defined.").
version() -> "".
-endif.
The error message will look like this:
% erlc t.erl
t.erl:7: -error("Macro VERSION must be defined.").
The directive -warning(Term) causes a compilation warning.
Example:
-module(t).
-export([version/0]).
-ifndef(VERSION).
-warning("Macro VERSION not defined -- using default version.").
-define(VERSION, "0").
-endif.
version() -> ?VERSION.
The warning message will look like this:
% erlc t.erl
t.erl:5: Warning: -warning("Macro VERSION not defined -- using default version.").
The -error() and -warning() directives were added in OTP 19.
5.9.7 Stringifying Macro Arguments
The construction ??Arg, where Arg is a macro argument, is expanded to a string containing the tokens of the
argument. This is similar to the #arg stringifying construction in C.
Example:
-define(TESTCALL(Call), io:format("Call ~s: ~w~n", [??Call, Call])).
?TESTCALL(myfunction(1,2)),
?TESTCALL(you:function(2,1)).
results in
io:format("Call ~s: ~w~n",["myfunction ( 1 , 2 )",myfunction(1,2)]),
io:format("Call ~s: ~w~n",["you : function ( 2 , 1 )",you:function(2,1)]).
That is, a trace output, with both the function called and the resulting value.
174 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.10 Records
5.10 Records
A record is a data structure for storing a fixed number of elements. It has named fields and is similar to a struct in
C. Record expressions are translated to tuple expressions during compilation. Therefore, record expressions are not
understood by the shell unless special actions are taken. For details, see the shell(3) manual page in STDLIB.
More examples are provided in Programming Examples.
5.10.1 Defining Records
A record definition consists of the name of the record, followed by the field names of the record. Record and field names
must be atoms. Each field can be given an optional default value. If no default value is supplied, undefined is used.
-record(Name, {Field1 [= Value1],
...
FieldN [= ValueN]}).
A record definition can be placed anywhere among the attributes and function declarations of a module, but the
definition must come before any usage of the record.
If a record is used in several modules, it is recommended that the record definition is placed in an include file.
5.10.2 Creating Records
The following expression creates a new Name record where the value of each field FieldI is the value of evaluating
the corresponding expression ExprI:
#Name{Field1=Expr1,...,FieldK=ExprK}
The fields can be in any order, not necessarily the same order as in the record definition, and fields can be omitted.
Omitted fields get their respective default value instead.
If several fields are to be assigned the same value, the following construction can be used:
#Name{Field1=Expr1,...,FieldK=ExprK, _=ExprL}
Omitted fields then get the value of evaluating ExprL instead of their default values. This feature is primarily intended
to be used to create patterns for ETS and Mnesia match functions.
Example:
-record(person, {name, phone, address}).
...
lookup(Name, Tab) ->
ets:match_object(Tab, #person{name=Name, _='_'}).
5.10.3 Accessing Record Fields
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 175
5.10 Records
Expr#Name.Field
Returns the value of the specified field. Expr is to evaluate to a Name record.
The following expression returns the position of the specified field in the tuple representation of the record:
#Name.Field
Example:
-record(person, {name, phone, address}).
...
lookup(Name, List) ->
lists:keysearch(Name, #person.name, List).
5.10.4 Updating Records
Expr#Name{Field1=Expr1,...,FieldK=ExprK}
Expr is to evaluate to a Name record. A copy of this record is returned, with the value of each specified field FieldI
changed to the value of evaluating the corresponding expression ExprI. All other fields retain their old values.
5.10.5 Records in Guards
Since record expressions are expanded to tuple expressions, creating records and accessing record fields are allowed
in guards. However all subexpressions, for example, for field initiations, must be valid guard expressions as well.
Examples:
handle(Msg, State) when Msg==#msg{to=void, no=3} ->
...
handle(Msg, State) when State#state.running==true ->
...
There is also a type test BIF is_record(Term, RecordTag).
Example:
is_person(P) when is_record(P, person) ->
true;
is_person(_P) ->
false.
5.10.6 Records in Patterns
A pattern that matches a certain record is created in the same way as a record is created:
176 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.10 Records
#Name{Field1=Expr1,...,FieldK=ExprK}
In this case, one or more of Expr1...ExprK can be unbound variables.
5.10.7 Nested Records
Beginning with Erlang/OTP R14, parentheses when accessing or updating nested records can be omitted. Assume the
following record definitions:
-record(nrec0, {name = "nested0"}).
-record(nrec1, {name = "nested1", nrec0=#nrec0{}}).
-record(nrec2, {name = "nested2", nrec1=#nrec1{}}).
N2 = #nrec2{},
Before R14, parentheses were needed as follows:
"nested0" = ((N2#nrec2.nrec1)#nrec1.nrec0)#nrec0.name,
N0n = ((N2#nrec2.nrec1)#nrec1.nrec0)#nrec0{name = "nested0a"},
Since R14, the following can also be written:
"nested0" = N2#nrec2.nrec1#nrec1.nrec0#nrec0.name,
N0n = N2#nrec2.nrec1#nrec1.nrec0#nrec0{name = "nested0a"},
5.10.8 Internal Representation of Records
Record expressions are translated to tuple expressions during compilation. A record defined as:
-record(Name, {Field1,...,FieldN}).
is internally represented by the tuple:
{Name,Value1,...,ValueN}
Here each ValueI is the default value for FieldI.
To each module using records, a pseudo function is added during compilation to obtain information about records:
record_info(fields, Record) -> [Field]
record_info(size, Record) -> Size
Size is the size of the tuple representation, that is, one more than the number of fields.
In addition, #Record.Name returns the index in the tuple representation of Name of the record Record.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 177
5.11 Errors and Error Handling
Name must be an atom.
5.11 Errors and Error Handling
5.11.1 Terminology
Errors can roughly be divided into four different types:
•
•
•
•
Compile-time errors
Logical errors
Run-time errors
Generated errors
A compile-time error, for example a syntax error, does not cause much trouble as it is caught by the compiler.
A logical error is when a program does not behave as intended, but does not crash. An example is that nothing happens
when a button in a graphical user interface is clicked.
A run-time error is when a crash occurs. An example is when an operator is applied to arguments of the wrong type.
The Erlang programming language has built-in features for handling of run-time errors.
A run-time error can also be emulated by calling erlang:error(Reason) or erlang:error(Reason,
Args).
A run-time error is another name for an exception of class error.
A generated error is when the code itself calls exit/1 or throw/1. Notice that emulated run-time errors are not
denoted as generated errors here.
Generated errors are exceptions of classes exit and throw.
When a run-time error or generated error occurs in Erlang, execution for the process that evaluated the erroneous
expression is stopped. This is referred to as a failure, that execution or evaluation fails, or that the process fails,
terminates, or exits. Notice that a process can terminate/exit for other reasons than a failure.
A process that terminates emits an exit signal with an exit reason that says something about which error has occurred.
Normally, some information about the error is printed to the terminal.
5.11.2 Exceptions
Exceptions are run-time errors or generated errors and are of three different classes, with different origins. The try
expression can distinguish between the different classes, whereas the catch expression cannot. They are described in
Expressions .
Class
Origin
error
Run-time error, for example, 1+a, or the process called
erlang:error/1,2
exit
The process called exit/1
throw
The process called throw/1
Table 11.1: Exception Classes.
178 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.11 Errors and Error Handling
An exception consists of its class, an exit reason (see Exit Reason), and a stack trace (which aids in finding the code
location of the exception).
The stack trace can be retrieved using erlang:get_stacktrace/0 from within a try expression, and is returned
for exceptions of class error from a catch expression.
An exception of class error is also known as a run-time error.
5.11.3 Handling of Run-time Errors in Erlang
Error Handling Within Processes
It is possible to prevent run-time errors and other exceptions from causing the process to terminate by using catch
or try, see Expressions about catch and try.
Error Handling Between Processes
Processes can monitor other processes and detect process terminations, see Processes.
5.11.4 Exit Reasons
When a run-time error occurs, that is an exception of class error. The exit reason is a tuple {Reason,Stack},
where Reason is a term indicating the type of error:
Reason
Type of Error
badarg
Bad argument. The argument is of wrong data type, or is
otherwise badly formed.
badarith
Bad argument in an arithmetic expression.
{badmatch,V}
Evaluation of a match expression failed. The value V
did not match.
function_clause
No matching function clause is found when evaluating a
function call.
{case_clause,V}
No matching branch is found when evaluating a case
expression. The value V did not match.
if_clause
No true branch is found when evaluating an if
expression.
{try_clause,V}
No matching branch is found when evaluating the ofsection of a try expression. The value V did not match.
undef
The function cannot be found when evaluating a
function call.
{badfun,F}
Something is wrong with a fun F.
{badarity,F}
A fun is applied to the wrong number of arguments. F
describes the fun and the arguments.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 179
5.12 Processes
timeout_value
The timeout value in a receive..after expression
is evaluated to something else than an integer or
infinity.
noproc
Trying to link to a non-existing process.
{nocatch,V}
Trying to evaluate a throw outside a catch. V is the
thrown term.
system_limit
A system limit has been reached. See Efficiency Guide
for information about system limits.
Table 11.2: Exit Reasons
Stack is the stack of function calls being evaluated when the error occurred, given as a list of tuples
{Module,Name,Arity} with the most recent function call first. The most recent function call tuple can in some
cases be {Module,Name,[Arg]}.
5.12 Processes
5.12.1 Processes
Erlang is designed for massive concurrency. Erlang processes are lightweight (grow and shrink dynamically) with
small memory footprint, fast to create and terminate, and the scheduling overhead is low.
5.12.2 Process Creation
A process is created by calling spawn:
spawn(Module, Name, Args) -> pid()
Module = Name = atom()
Args = [Arg1,...,ArgN]
ArgI = term()
spawn creates a new process and returns the pid.
The new process starts executing in Module:Name(Arg1,...,ArgN) where the arguments are the elements of
the (possible empty) Args argument list.
There exist a number of other spawn BIFs, for example, spawn/4 for spawning a process at another node.
5.12.3 Registered Processes
Besides addressing a process by using its pid, there are also BIFs for registering a process under a name. The name
must be an atom and is automatically unregistered if the process terminates:
BIF
Description
register(Name, Pid)
Associates the name Name, an atom, with the process
Pid.
180 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.12 Processes
registered()
Returns a list of names that have been registered using
register/2.
whereis(Name)
Returns the pid registered under Name, or undefined
if the name is not registered.
Table 12.1: Name Registration BIFs
5.12.4 Process Termination
When a process terminates, it always terminates with an exit reason. The reason can be any term.
A process is said to terminate normally, if the exit reason is the atom normal. A process with no more code to
execute terminates normally.
A process terminates with an exit reason {Reason,Stack} when a run-time error occurs. See Exit Reasons.
A process can terminate itself by calling one of the following BIFs:
•
•
•
exit(Reason)
erlang:error(Reason)
erlang:error(Reason, Args)
The process then terminates with reason Reason for exit/1 or {Reason,Stack} for the others.
A process can also be terminated if it receives an exit signal with another exit reason than normal, see Error Handling.
5.12.5 Message Sending
Processes communicate by sending and receiving messages. Messages are sent by using the send operator ! and
received by calling receive.
Message sending is asynchronous and safe, the message is guaranteed to eventually reach the recipient, provided that
the recipient exists.
5.12.6 Links
Two processes can be linked to each other. A link between two processes Pid1 and Pid2 is created by Pid1 calling
the BIF link(Pid2) (or conversely). There also exist a number of spawn_link BIFs, which spawn and link to
a process in one operation.
Links are bidirectional and there can only be one link between two processes. Repeated calls to link(Pid) have
no effect.
A link can be removed by calling the BIF unlink(Pid).
Links are used to monitor the behaviour of other processes, see Error Handling.
5.12.7 Error Handling
Erlang has a built-in feature for error handling between processes. Terminating processes emit exit signals to all
linked processes, which can terminate as well or handle the exit in some way. This feature can be used to build
hierarchical program structures where some processes are supervising other processes, for example, restarting them
if they terminate abnormally.
See OTP Design Principles for more information about OTP supervision trees, which use this feature.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 181
5.12 Processes
Emitting Exit Signals
When a process terminates, it terminates with an exit reason as explained in Process Termination. This exit reason
is emitted in an exit signal to all linked processes.
A process can also call the function exit(Pid,Reason). This results in an exit signal with exit reason Reason
being emitted to Pid, but does not affect the calling process.
Receiving Exit Signals
The default behaviour when a process receives an exit signal with an exit reason other than normal, is to terminate
and in turn emit exit signals with the same exit reason to its linked processes. An exit signal with reason normal
is ignored.
A process can be set to trap exit signals by calling:
process_flag(trap_exit, true)
When a process is trapping exits, it does not terminate when an exit signal is received. Instead, the signal is transformed
into a message {'EXIT',FromPid,Reason}, which is put into the mailbox of the process, just like a regular
message.
An exception to the above is if the exit reason is kill, that is if exit(Pid,kill) has been called. This
unconditionally terminates the process, regardless of if it is trapping exit signals.
5.12.8 Monitors
An alternative to links are monitors. A process Pid1 can create a monitor for Pid2 by calling the BIF
erlang:monitor(process, Pid2). The function returns a reference Ref.
If Pid2 terminates with exit reason Reason, a 'DOWN' message is sent to Pid1:
{'DOWN', Ref, process, Pid2, Reason}
If Pid2 does not exist, the 'DOWN' message is sent immediately with Reason set to noproc.
Monitors are unidirectional. Repeated calls to erlang:monitor(process, Pid) creates several independent
monitors, and each one sends a 'DOWN' message when Pid terminates.
A monitor can be removed by calling erlang:demonitor(Ref).
Monitors can be created for processes with registered names, also at other nodes.
5.12.9 Process Dictionary
Each process has its own process dictionary, accessed by calling the following BIFs:
put(Key, Value)
get(Key)
get()
get_keys(Value)
erase(Key)
erase()
182 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.13 Distributed Erlang
5.13 Distributed Erlang
5.13.1 Distributed Erlang System
A distributed Erlang system consists of a number of Erlang runtime systems communicating with each other. Each
such runtime system is called a node. Message passing between processes at different nodes, as well as links and
monitors, are transparent when pids are used. Registered names, however, are local to each node. This means that the
node must be specified as well when sending messages, and so on, using registered names.
The distribution mechanism is implemented using TCP/IP sockets. How to implement an alternative carrier is described
in the ERTS User's Guide.
Warning:
Starting a distributed node without also specifying -proto_dist inet_tls will expose the node to attacks
that may give the attacker complete access to the node and in extension the cluster. When using un-secure
distributed nodes, make sure that the network is configured to keep potential attackers out. See the Using SSL
for Erlang Distribution User's Guide for details on how to setup a secure distributed node.
5.13.2 Nodes
A node is an executing Erlang runtime system that has been given a name, using the command-line flag -name (long
names) or -sname (short names).
The format of the node name is an atom name@host. name is the name given by the user. host is the full host name
if long names are used, or the first part of the host name if short names are used. node() returns the name of the node.
Example:
% erl -name dilbert
(dilbert@uab.ericsson.se)1> node().
'dilbert@uab.ericsson.se'
% erl -sname dilbert
(dilbert@uab)1> node().
dilbert@uab
Note:
A node with a long node name cannot communicate with a node with a short node name.
5.13.3 Node Connections
The nodes in a distributed Erlang system are loosely connected. The first time the name of another node is used,
for example, if spawn(Node,M,F,A) or net_adm:ping(Node) is called, a connection attempt to that node
is made.
Connections are by default transitive. If a node A connects to node B, and node B has a connection to node C, then node
A also tries to connect to node C. This feature can be turned off by using the command-line flag -connect_all
false, see the erl(1) manual page in ERTS.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 183
5.13 Distributed Erlang
If a node goes down, all connections to that node are removed. Calling erlang:disconnect_node(Node)
forces disconnection of a node.
The list of (visible) nodes currently connected to is returned by nodes().
5.13.4 epmd
The Erlang Port Mapper Daemon epmd is automatically started at every host where an Erlang node is started. It is
responsible for mapping the symbolic node names to machine addresses. See the epmd(1) manual page in ERTS.
5.13.5 Hidden Nodes
In a distributed Erlang system, it is sometimes useful to connect to a node without also connecting to all other nodes.
An example is some kind of O&M functionality used to inspect the status of a system, without disturbing it. For this
purpose, a hidden node can be used.
A hidden node is a node started with the command-line flag -hidden. Connections between hidden nodes and other
nodes are not transitive, they must be set up explicitly. Also, hidden nodes does not show up in the list of nodes returned
by nodes(). Instead, nodes(hidden) or nodes(connected) must be used. This means, for example, that
the hidden node is not added to the set of nodes that global is keeping track of.
This feature was added in Erlang 5.0/OTP R7.
5.13.6 C Nodes
A C node is a C program written to act as a hidden node in a distributed Erlang system. The library Erl_Interface
contains functions for this purpose. For more information about C nodes, see the Erl_Interface application and
Interoperability Tutorial..
5.13.7 Security
Authentication determines which nodes are allowed to communicate with each other. In a network of different Erlang
nodes, it is built into the system at the lowest possible level. Each node has its own magic cookie, which is an Erlang
atom.
When a node tries to connect to another node, the magic cookies are compared. If they do not match, the connected
node rejects the connection.
At start-up, a node has a random atom assigned as its magic cookie and the cookie of other nodes is assumed to
be nocookie. The first action of the Erlang network authentication server (auth) is then to read a file named
$HOME/.erlang.cookie. If the file does not exist, it is created. The UNIX permissions mode of the file is set
to octal 400 (read-only by user) and its contents are a random string. An atom Cookie is created from the contents
of the file and the cookie of the local node is set to this using erlang:set_cookie(node(), Cookie). This
also makes the local node assume that all other nodes have the same cookie Cookie.
Thus, groups of users with identical cookie files get Erlang nodes that can communicate freely and without interference
from the magic cookie system. Users who want to run nodes on separate file systems must make certain that their
cookie files are identical on the different file systems.
For a node Node1 with magic cookie Cookie to be able to connect to, or accept a connection from, another node
Node2 with a different cookie DiffCookie, the function erlang:set_cookie(Node2, DiffCookie)
must first be called at Node1. Distributed systems with multiple user IDs can be handled in this way.
The default when a connection is established between two nodes, is to immediately connect all other visible nodes as
well. This way, there is always a fully connected network. If there are nodes with different cookies, this method can be
inappropriate and the command-line flag -connect_all false must be set, see the erl(1) manual page in ERTS.
The magic cookie of the local node is retrieved by calling erlang:get_cookie().
184 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.13 Distributed Erlang
5.13.8 Distribution BIFs
Some useful BIFs for distributed programming (for more information, see the erlang(3) manual page in ERTS:
BIF
Description
erlang:disconnect_node(Node)
Forces the disconnection of a node.
erlang:get_cookie()
Returns the magic cookie of the current node.
is_alive()
Returns true if the runtime system is a node and can
connect to other nodes, false otherwise.
monitor_node(Node, true|false)
Monitors the status of Node. A message{nodedown,
Node} is received if the connection to it is lost.
node()
Returns the name of the current node. Allowed in
guards.
node(Arg)
Returns the node where Arg, a pid, reference, or port, is
located.
nodes()
Returns a list of all visible nodes this node is connected
to.
nodes(Arg)
Depending on Arg, this function can return a list
not only of visible nodes, but also hidden nodes and
previously known nodes, and so on.
erlang:set_cookie(Node, Cookie)
Sets the magic cookie used when connecting to Node.
If Node is the current node, Cookie is used when
connecting to all new nodes.
spawn[_link|_opt](Node, Fun)
Creates a process at a remote node.
spawn[_link|opt](Node, Module,
FunctionName, Args)
Creates a process at a remote node.
Table 13.1: Distribution BIFs
5.13.9 Distribution Command-Line Flags
Examples of command-line flags used for distributed programming (for more information, see the erl(1) manual page
in ERTS:
Command-Line Flag
Description
-connect_all false
Only explicit connection set-ups are used.
-hidden
Makes a node into a hidden node.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 185
5.14 Compilation and Code Loading
-name Name
Makes a runtime system into a node, using long node
names.
-setcookie Cookie
Same as calling erlang:set_cookie(node(),
Cookie).
-sname Name
Makes a runtime system into a node, using short node
names.
Table 13.2: Distribution Command-Line Flags
5.13.10 Distribution Modules
Examples of modules useful for distributed programming:
In the Kernel application:
Module
Description
global
A global name registration facility.
global_group
Grouping nodes to global name registration groups.
net_adm
Various Erlang net administration routines.
net_kernel
Erlang networking kernel.
Table 13.3: Kernel Modules Useful For Distribution.
In the STDLIB application:
Module
Description
slave
Start and control of slave nodes.
Table 13.4: STDLIB Modules Useful For Distribution.
5.14 Compilation and Code Loading
How code is compiled and loaded is not a language issue, but is system-dependent. This section describes compilation
and code loading in Erlang/OTP with references to relevant parts of the documentation.
5.14.1 Compilation
Erlang programs must be compiled to object code. The compiler can generate a new file that contains the object code.
The current abstract machine, which runs the object code, is called BEAM, therefore the object files get the suffix
.beam. The compiler can also generate a binary which can be loaded directly.
The compiler is located in the module compile (see the compile(3) manual page in Compiler).
186 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.14 Compilation and Code Loading
compile:file(Module)
compile:file(Module, Options)
The Erlang shell understands the command c(Module) which both compiles and loads Module.
There is also a module make, which provides a set of functions similar to the UNIX type Make functions, see the
make(3) manual page in Tools.
The compiler can also be accessed from the OS prompt, see the erl(1) manual page in ERTS.
% erl -compile Module1...ModuleN
% erl -make
The erlc program provides an even better way to compile modules from the shell, see the erlc(1) manual page in
ERTS. It understands a number of flags that can be used to define macros, add search paths for include files, and more.
% erlc <flags> File1.erl...FileN.erl
5.14.2 Code Loading
The object code must be loaded into the Erlang runtime system. This is handled by the code server, see the code(3)
manual page in Kernel.
The code server loads code according to a code loading strategy, which is either interactive (default) or embedded.
In interactive mode, code is searched for in a code path and loaded when first referenced. In embedded mode, code
is loaded at start-up according to a boot script. This is described in System Principles .
5.14.3 Code Replacement
Erlang supports change of code in a running system. Code replacement is done on module level.
The code of a module can exist in two variants in a system: current and old. When a module is loaded into the system
for the first time, the code becomes 'current'. If then a new instance of the module is loaded, the code of the previous
instance becomes 'old' and the new instance becomes 'current'.
Both old and current code is valid, and can be evaluated concurrently. Fully qualified function calls always refer to
current code. Old code can still be evaluated because of processes lingering in the old code.
If a third instance of the module is loaded, the code server removes (purges) the old code and any processes lingering
in it is terminated. Then the third instance becomes 'current' and the previously current code becomes 'old'.
To change from old code to current code, a process must make a fully qualified function call.
Example:
-module(m).
-export([loop/0]).
loop() ->
receive
code_switch ->
m:loop();
Msg ->
...
loop()
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 187
5.14 Compilation and Code Loading
end.
To make the process change code, send the message code_switch to it. The process then makes a fully qualified
call to m:loop() and changes to current code. Notice that m:loop/0 must be exported.
For code replacement of funs to work, use the syntax fun Module:FunctionName/Arity.
5.14.4 Running a Function When a Module is Loaded
The -on_load() directive names a function that is to be run automatically when a module is loaded.
Its syntax is as follows:
-on_load(Name/0).
It is not necessary to export the function. It is called in a freshly spawned process (which terminates as soon as the
function returns).
The function must return ok if the module is to become the new current code for the module and become callable.
Returning any other value or generating an exception causes the new code to be unloaded. If the return value is not
an atom, a warning error report is sent to the error logger.
If there already is current code for the module, that code will remain current and can be called until the on_load
function has returned. If the on_load function fails, the current code (if any) will remain current. If there is no current
code for a module, any process that makes an external call to the module before the on_load function has finished
will be suspended until the on_load function have finished.
Note:
Before OTP 19, if the on_load function failed, any previously current code would become old, essentially
leaving the system without any working and reachable instance of the module. That problem has been eliminated
in OTP 19.
In embedded mode, first all modules are loaded. Then all on_load functions are called. The system is terminated
unless all of the on_load functions return ok.
Example:
-module(m).
-on_load(load_my_nifs/0).
load_my_nifs() ->
NifPath = ...,
%Set up the path to the NIF library.
Info = ...,
%Initialize the Info term
erlang:load_nif(NifPath, Info).
If the call to erlang:load_nif/2 fails, the module is unloaded and a warning report is sent to the error loader.
188 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.15 Ports and Port Drivers
5.15 Ports and Port Drivers
Examples of how to use ports and port drivers are provided in Interoperability Tutorial. For information about the
BIFs mentioned, see the erlang(3) manual page in ERTS.
5.15.1 Ports
Ports provide the basic mechanism for communication with the external world, from Erlang's point of view. They
provide a byte-oriented interface to an external program. When a port has been created, Erlang can communicate with
it by sending and receiving lists of bytes, including binaries.
The Erlang process creating a port is said to be the port owner, or the connected process of the port. All
communication to and from the port must go through the port owner. If the port owner terminates, so does the port
(and the external program, if it is written correctly).
The external program resides in another OS process. By default, it reads from standard input (file descriptor 0) and
writes to standard output (file descriptor 1). The external program is to terminate when the port is closed.
5.15.2 Port Drivers
It is possible to write a driver in C according to certain principles and dynamically link it to the Erlang runtime system.
The linked-in driver looks like a port from the Erlang programmer's point of view and is called a port driver.
Warning:
An erroneous port driver causes the entire Erlang runtime system to leak memory, hang or crash.
For information about port drivers, see the erl_driver(4) manual page in ERTS, driver_entry(1) manual page in ERTS,
and erl_ddll(3) manual page in Kernel.
5.15.3 Port BIFs
To create a port:
open_port(PortName, PortSettings
Returns a port identifier Port as the result of opening a
new Erlang port. Messages can be sent to, and received
from, a port identifier, just like a pid. Port identifiers
can also be linked to using link/1, or registered under
a name using register/2.
Table 15.1: Port Creation BIF
PortName is usually a tuple {spawn,Command}, where the string Command is the name of the external program.
The external program runs outside the Erlang workspace, unless a port driver with the name Command is found. If
Command is found, that driver is started.
PortSettings is a list of settings (options) for the port. The list typically contains at least a tuple {packet,N},
which specifies that data sent between the port and the external program are preceded by an N-byte length indicator.
Valid values for N are 1, 2, or 4. If binaries are to be used instead of lists of bytes, the option binary must be included.
The port owner Pid can communicate with the port Port by sending and receiving messages. (In fact, any process
can send the messages to the port, but the port owner must be identified in the message).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 189
5.15 Ports and Port Drivers
As of Erlang/OTP R16, messages sent to ports are delivered truly asynchronously. The underlying implementation
previously delivered messages to ports synchronously. Message passing has however always been documented as an
asynchronous operation. Hence, this is not to be an issue for an Erlang program communicating with ports, unless
false assumptions about ports have been made.
In the following tables of examples, Data must be an I/O list. An I/O list is a binary or a (possibly deep) list of binaries
or integers in the range 0..255:
Message
Description
{Pid,{command,Data}}
Sends Data to the port.
{Pid,close}
Closes the port. Unless the port is already closed, the
port replies with {Port,closed} when all buffers
have been flushed and the port really closes.
{Pid,{connect,NewPid}}
Sets the port owner of Portto NewPid. Unless
the port is already closed, the port replies
with{Port,connected} to the old port owner. Note
that the old port owner is still linked to the port, but the
new port owner is not.
Table 15.2: Messages Sent To a Port
Message
Description
{Port,{data,Data}}
Data is received from the external program.
{Port,closed}
Reply to Port ! {Pid,close}.
{Port,connected}
Reply to Port ! {Pid,{connect,NewPid}}.
{'EXIT',Port,Reason}
If the port has terminated for some reason.
Table 15.3: Messages Received From a Port
Instead of sending and receiving messages, there are also a number of BIFs that can be used:
Port BIF
Description
port_command(Port,Data)
Sends Data to the port.
port_close(Port)
Closes the port.
port_connect(Port,NewPid)
Sets the port owner of Portto NewPid. The old
port owner Pid stays linked to the port and must call
unlink(Port) if this is not desired.
erlang:port_info(Port,Item)
Returns information as specified by Item.
190 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
5.15 Ports and Port Drivers
erlang:ports()
Returns a list of all ports on the current node.
Table 15.4: Port BIFs
Some additional BIFs that apply to port drivers: port_control/3 and erlang:port_call/3.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 191
6.1 Records
6 Programming Examples
This section contains examples on using records, funs, list comprehensions, and the bit syntax.
6.1 Records
6.1.1 Records and Tuples
The main advantage of using records rather than tuples is that fields in a record are accessed by name, whereas fields
in a tuple are accessed by position. To illustrate these differences, suppose that you want to represent a person with
the tuple {Name, Address, Phone}.
To write functions that manipulate this data, remember the following:
•
•
•
The Name field is the first element of the tuple.
The Address field is the second element.
The Phone field is the third element.
For example, to extract data from a variable P that contains such a tuple, you can write the following code and then
use pattern matching to extract the relevant fields:
Name = element(1, P),
Address = element(2, P),
...
Such code is difficult to read and understand, and errors occur if the numbering of the elements in the tuple is wrong. If
the data representation of the fields is changed, by re-ordering, adding, or removing fields, all references to the person
tuple must be checked and possibly modified.
Records allow references to the fields by name, instead of by position. In the following example, a record instead of
a tuple is used to store the data:
-record(person, {name, phone, address}).
This enables references to the fields of the record by name. For example, if P is a variable whose value is a person
record, the following code access the name and address fields of the records:
Name = P#person.name,
Address = P#person.address,
...
Internally, records are represented using tagged tuples:
{person, Name, Phone, Address}
192 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.1 Records
6.1.2 Defining a Record
This following definition of a person is used in several examples in this section. Three fields are included, name,
phone, and address. The default values for name and phone is "" and [], respectively. The default value for
address is the atom undefined, since no default value is supplied for this field:
-record(person, {name = "", phone = [], address}).
The record must be defined in the shell to enable use of the record syntax in the examples:
> rd(person, {name = "", phone = [], address}).
person
This is because record definitions are only available at compile time, not at runtime. For details on records in the shell,
see the shell(3) manual page in STDLIB.
6.1.3 Creating a Record
A new person record is created as follows:
> #person{phone=[0,8,2,3,4,3,1,2], name="Robert"}.
#person{name = "Robert",phone = [0,8,2,3,4,3,1,2],address = undefined}
As the address field was omitted, its default value is used.
From Erlang 5.1/OTP R8B, a value to all fields in a record can be set with the special field _. _ means "all fields
not explicitly specified".
Example:
> #person{name = "Jakob", _ = '_'}.
#person{name = "Jakob",phone = '_',address = '_'}
It is primarily intended to be used in ets:match/2 and mnesia:match_object/3, to set record fields to the
atom '_'. (This is a wildcard in ets:match/2.)
6.1.4 Accessing a Record Field
The following example shows how to access a record field:
> P = #person{name = "Joe", phone = [0,8,2,3,4,3,1,2]}.
#person{name = "Joe",phone = [0,8,2,3,4,3,1,2],address = undefined}
> P#person.name.
"Joe"
6.1.5 Updating a Record
The following example shows how to update a record:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 193
6.1 Records
> P1 = #person{name="Joe", phone=[1,2,3], address="A street"}.
#person{name = "Joe",phone = [1,2,3],address = "A street"}
> P2 = P1#person{name="Robert"}.
#person{name = "Robert",phone = [1,2,3],address = "A street"}
6.1.6 Type Testing
The following example shows that the guard succeeds if P is record of type person:
foo(P) when is_record(P, person) -> a_person;
foo(_) -> not_a_person.
6.1.7 Pattern Matching
Matching can be used in combination with records, as shown in the following example:
> P3 = #person{name="Joe", phone=[0,0,7], address="A street"}.
#person{name = "Joe",phone = [0,0,7],address = "A street"}
> #person{name = Name} = P3, Name.
"Joe"
The following function takes a list of person records and searches for the phone number of a person with a particular
name:
find_phone([#person{name=Name, phone=Phone} | _], Name) ->
{found, Phone};
find_phone([_| T], Name) ->
find_phone(T, Name);
find_phone([], Name) ->
not_found.
The fields referred to in the pattern can be given in any order.
6.1.8 Nested Records
The value of a field in a record can be an instance of a record. Retrieval of nested data can be done stepwise, or in
a single step, as shown in the following example:
-record(name, {first = "Robert", last = "Ericsson"}).
-record(person, {name = #name{}, phone}).
demo() ->
P = #person{name= #name{first="Robert",last="Virding"}, phone=123},
First = (P#person.name)#name.first.
Here, demo() evaluates to "Robert".
6.1.9 A Longer Example
Comments are embedded in the following example:
194 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.2 Funs
%% File: person.hrl
%%----------------------------------------------------------%% Data Type: person
%% where:
%%
name: A string (default is undefined).
%%
age:
An integer (default is undefined).
%%
phone: A list of integers (default is []).
%%
dict: A dictionary containing various information
%%
about the person.
%%
A {Key, Value} list (default is the empty list).
%%------------------------------------------------------------record(person, {name, age, phone = [], dict = []}).
-module(person).
-include("person.hrl").
-compile(export_all). % For test purposes only.
%% This creates an instance of a person.
%%
Note: The phone number is not supplied so the
%%
default value [] will be used.
make_hacker_without_phone(Name, Age) ->
#person{name = Name, age = Age,
dict = [{computer_knowledge, excellent},
{drinks, coke}]}.
%% This demonstrates matching in arguments
print(#person{name = Name, age = Age,
phone = Phone, dict = Dict}) ->
io:format("Name: ~s, Age: ~w, Phone: ~w ~n"
"Dictionary: ~w.~n", [Name, Age, Phone, Dict]).
%% Demonstrates type testing, selector, updating.
birthday(P) when record(P, person) ->
P#person{age = P#person.age + 1}.
register_two_hackers() ->
Hacker1 = make_hacker_without_phone("Joe", 29),
OldHacker = birthday(Hacker1),
% The central_register_server should have
% an interface function for this.
central_register_server ! {register_person, Hacker1},
central_register_server ! {register_person,
OldHacker#person{name = "Robert",
phone = [0,8,3,2,4,5,3,1]}}.
6.2 Funs
6.2.1 map
The following function, double, doubles every element in a list:
double([H|T]) -> [2*H|double(T)];
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 195
6.2 Funs
double([])
-> [].
Hence, the argument entered as input is doubled as follows:
> double([1,2,3,4]).
[2,4,6,8]
The following function, add_one, adds one to every element in a list:
add_one([H|T]) -> [H+1|add_one(T)];
add_one([])
-> [].
The functions double and add_one have a similar structure. This can be used by writing a function map that
expresses this similarity:
map(F, [H|T]) -> [F(H)|map(F, T)];
map(F, [])
-> [].
The functions double and add_one can now be expressed in terms of map as follows:
double(L) -> map(fun(X) -> 2*X end, L).
add_one(L) -> map(fun(X) -> 1 + X end, L).
map(F, List) is a function that takes a function F and a list L as arguments and returns a new list, obtained by
applying F to each of the elements in L.
The process of abstracting out the common features of a number of different programs is called procedural
abstraction. Procedural abstraction can be used to write several different functions that have a similar structure, but
differ in some minor detail. This is done as follows:
•
•
Step 1. Write one function that represents the common features of these functions.
Step 2. Parameterize the difference in terms of functions that are passed as arguments to the common function.
6.2.2 foreach
This section illustrates procedural abstraction. Initially, the following two examples are written as conventional
functions.
This function prints all elements of a list onto a stream:
print_list(Stream, [H|T]) ->
io:format(Stream, "~p~n", [H]),
print_list(Stream, T);
print_list(Stream, []) ->
true.
This function broadcasts a message to a list of processes:
196 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.2 Funs
broadcast(Msg, [Pid|Pids]) ->
Pid ! Msg,
broadcast(Msg, Pids);
broadcast(_, []) ->
true.
These two functions have a similar structure. They both iterate over a list and do something to each element in the list.
The "something" is passed on as an extra argument to the function that does this.
The function foreach expresses this similarity:
foreach(F, [H|T]) ->
F(H),
foreach(F, T);
foreach(F, []) ->
ok.
Using the function foreach, the function print_list becomes:
foreach(fun(H) -> io:format(S, "~p~n",[H]) end, L)
Using the function foreach, the function broadcast becomes:
foreach(fun(Pid) -> Pid ! M end, L)
foreach is evaluated for its side-effect and not its value. foreach(Fun ,L) calls Fun(X) for each element
X in L and the processing occurs in the order that the elements were defined in L. map does not define the order in
which its elements are processed.
6.2.3 Syntax of Funs
Funs are written with the following syntax (see Fun Expressions for full description):
F = fun (Arg1, Arg2, ... ArgN) ->
...
end
This creates an anonymous function of N arguments and binds it to the variable F.
Another function, FunctionName, written in the same module, can be passed as an argument, using the following
syntax:
F = fun FunctionName/Arity
With this form of function reference, the function that is referred to does not need to be exported from the module.
It is also possible to refer to a function defined in a different module, with the following syntax:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 197
6.2 Funs
F = fun Module:FunctionName/Arity
In this case, the function must be exported from the module in question.
The following program illustrates the different ways of creating funs:
-module(fun_test).
-export([t1/0, t2/0]).
-import(lists, [map/2]).
t1() -> map(fun(X) -> 2 * X end, [1,2,3,4,5]).
t2() -> map(fun double/1, [1,2,3,4,5]).
double(X) -> X * 2.
The fun F can be evaluated with the following syntax:
F(Arg1, Arg2, ..., Argn)
To check whether a term is a fun, use the test is_function/1 in a guard.
Example:
f(F, Args) when is_function(F) ->
apply(F, Args);
f(N, _) when is_integer(N) ->
N.
Funs are a distinct type. The BIFs erlang:fun_info/1,2 can be used to retrieve information about a fun, and
the BIF erlang:fun_to_list/1 returns a textual representation of a fun. The check_process_code/2 BIF
returns true if the process contains funs that depend on the old version of a module.
6.2.4 Variable Bindings Within a Fun
The scope rules for variables that occur in funs are as follows:
•
•
•
All variables that occur in the head of a fun are assumed to be "fresh" variables.
Variables that are defined before the fun, and that occur in function calls or guard tests within the fun, have the
values they had outside the fun.
Variables cannot be exported from a fun.
The following examples illustrate these rules:
print_list(File, List) ->
{ok, Stream} = file:open(File, write),
foreach(fun(X) -> io:format(Stream,"~p~n",[X]) end, List),
file:close(Stream).
198 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.2 Funs
Here, the variable X, defined in the head of the fun, is a new variable. The variable Stream, which is used within
the fun, gets its value from the file:open line.
As any variable that occurs in the head of a fun is considered a new variable, it is equally valid to write as follows:
print_list(File, List) ->
{ok, Stream} = file:open(File, write),
foreach(fun(File) ->
io:format(Stream,"~p~n",[File])
end, List),
file:close(Stream).
Here, File is used as the new variable instead of X. This is not so wise because code in the fun body cannot refer to
the variable File, which is defined outside of the fun. Compiling this example gives the following diagnostic:
./FileName.erl:Line: Warning: variable 'File'
shadowed in 'fun'
This indicates that the variable File, which is defined inside the fun, collides with the variable File, which is
defined outside the fun.
The rules for importing variables into a fun has the consequence that certain pattern matching operations must be
moved into guard expressions and cannot be written in the head of the fun. For example, you might write the following
code if you intend the first clause of F to be evaluated when the value of its argument is Y:
f(...) ->
Y = ...
map(fun(X) when X == Y ->
;
(_) ->
...
end, ...)
...
instead of writing the following code:
f(...) ->
Y = ...
map(fun(Y) ->
;
(_) ->
...
end, ...)
...
6.2.5 Funs and Module Lists
The following examples show a dialogue with the Erlang shell. All the higher order functions discussed are exported
from the module lists.
map
map takes a function of one argument and a list of terms:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 199
6.2 Funs
map(F, [H|T]) -> [F(H)|map(F, T)];
map(F, [])
-> [].
It returns the list obtained by applying the function to every argument in the list.
When a new fun is defined in the shell, the value of the fun is printed as Fun#<erl_eval>:
> Double = fun(X) -> 2 * X end.
#Fun<erl_eval.6.72228031>
> lists:map(Double, [1,2,3,4,5]).
[2,4,6,8,10]
any
any takes a predicate P of one argument and a list of terms:
any(Pred, [H|T]) ->
case Pred(H) of
true -> true;
false -> any(Pred, T)
end;
any(Pred, []) ->
false.
A predicate is a function that returns true or false. any is true if there is a term X in the list such that P(X)
is true.
A predicate Big(X) is defined, which is true if its argument is greater that 10:
> Big = fun(X) -> if X > 10 -> true; true -> false end end.
#Fun<erl_eval.6.72228031>
> lists:any(Big, [1,2,3,4]).
false
> lists:any(Big, [1,2,3,12,5]).
true
all
all has the same arguments as any:
all(Pred, [H|T]) ->
case Pred(H) of
true -> all(Pred, T);
false -> false
end;
all(Pred, []) ->
true.
It is true if the predicate applied to all elements in the list is true.
200 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.2 Funs
> lists:all(Big, [1,2,3,4,12,6]).
false
> lists:all(Big, [12,13,14,15]).
true
foreach
foreach takes a function of one argument and a list of terms:
foreach(F, [H|T]) ->
F(H),
foreach(F, T);
foreach(F, []) ->
ok.
The function is applied to each argument in the list. foreach returns ok. It is only used for its side-effect:
> lists:foreach(fun(X) -> io:format("~w~n",[X]) end, [1,2,3,4]).
1
2
3
4
ok
foldl
foldl takes a function of two arguments, an accumulator and a list:
foldl(F, Accu, [Hd|Tail]) ->
foldl(F, F(Hd, Accu), Tail);
foldl(F, Accu, []) -> Accu.
The function is called with two arguments. The first argument is the successive elements in the list. The second
argument is the accumulator. The function must return a new accumulator, which is used the next time the function
is called.
If you have a list of lists L = ["I","like","Erlang"], then you can sum the lengths of all the strings in L
as follows:
> L = ["I","like","Erlang"].
["I","like","Erlang"]
10> lists:foldl(fun(X, Sum) -> length(X) + Sum end, 0, L).
11
foldl works like a while loop in an imperative language:
L = ["I","like","Erlang"],
Sum = 0,
while( L != []){
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 201
6.2 Funs
end
Sum += length(head(L)),
L = tail(L)
mapfoldl
mapfoldl simultaneously maps and folds over a list:
mapfoldl(F, Accu0, [Hd|Tail]) ->
{R,Accu1} = F(Hd, Accu0),
{Rs,Accu2} = mapfoldl(F, Accu1, Tail),
{[R|Rs], Accu2};
mapfoldl(F, Accu, []) -> {[], Accu}.
The following example shows how to change all letters in L to upper case and then count them.
First the change to upper case:
> Upcase = fun(X) when $a =< X,
(X) -> X
end.
#Fun<erl_eval.6.72228031>
> Upcase_word =
fun(X) ->
lists:map(Upcase, X)
end.
#Fun<erl_eval.6.72228031>
> Upcase_word("Erlang").
"ERLANG"
> lists:map(Upcase_word, L).
["I","LIKE","ERLANG"]
X =< $z -> X + $A - $a;
Now, the fold and the map can be done at the same time:
> lists:mapfoldl(fun(Word, Sum) ->
{Upcase_word(Word), Sum + length(Word)}
end, 0, L).
{["I","LIKE","ERLANG"],11}
filter
filter takes a predicate of one argument and a list and returns all elements in the list that satisfy the predicate:
filter(F, [H|T])
case F(H) of
true ->
false ->
end;
filter(F, []) ->
->
[H|filter(F, T)];
filter(F, T)
[].
202 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.2 Funs
> lists:filter(Big, [500,12,2,45,6,7]).
[500,12,45]
Combining maps and filters enables writing of very succinct code. For example, to define a set difference function
diff(L1, L2) to be the difference between the lists L1 and L2, the code can be written as follows:
diff(L1, L2) ->
filter(fun(X) -> not member(X, L2) end, L1).
This gives the list of all elements in L1 that are not contained in L2.
The AND intersection of the list L1 and L2 is also easily defined:
intersection(L1,L2) -> filter(fun(X) -> member(X,L1) end, L2).
takewhile
takewhile(P, L) takes elements X from a list L as long as the predicate P(X) is true:
takewhile(Pred, [H|T]) ->
case Pred(H) of
true -> [H|takewhile(Pred, T)];
false -> []
end;
takewhile(Pred, []) ->
[].
> lists:takewhile(Big, [200,500,45,5,3,45,6]).
[200,500,45]
dropwhile
dropwhile is the complement of takewhile:
dropwhile(Pred, [H|T]) ->
case Pred(H) of
true -> dropwhile(Pred, T);
false -> [H|T]
end;
dropwhile(Pred, []) ->
[].
> lists:dropwhile(Big, [200,500,45,5,3,45,6]).
[5,3,45,6]
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 203
6.2 Funs
splitwith
splitwith(P, L) splits the list L into the two sublists {L1, L2}, where L = takewhile(P, L) and L2
= dropwhile(P, L):
splitwith(Pred, L) ->
splitwith(Pred, L, []).
splitwith(Pred, [H|T], L) ->
case Pred(H) of
true -> splitwith(Pred, T, [H|L]);
false -> {reverse(L), [H|T]}
end;
splitwith(Pred, [], L) ->
{reverse(L), []}.
> lists:splitwith(Big, [200,500,45,5,3,45,6]).
{[200,500,45],[5,3,45,6]}
6.2.6 Funs Returning Funs
So far, only functions that take funs as arguments have been described. More powerful functions, that themselves
return funs, can also be written. The following examples illustrate these type of functions.
Simple Higher Order Functions
Adder(X) is a function that given X, returns a new function G such that G(K) returns K + X:
> Adder = fun(X) -> fun(Y) -> X + Y end end.
#Fun<erl_eval.6.72228031>
> Add6 = Adder(6).
#Fun<erl_eval.6.72228031>
> Add6(10).
16
Infinite Lists
The idea is to write something like:
-module(lazy).
-export([ints_from/1]).
ints_from(N) ->
fun() ->
[N|ints_from(N+1)]
end.
Then proceed as follows:
> XX = lazy:ints_from(1).
#Fun<lazy.0.29874839>
> XX().
204 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.2 Funs
[1|#Fun<lazy.0.29874839>]
> hd(XX()).
1
> Y = tl(XX()).
#Fun<lazy.0.29874839>
> hd(Y()).
2
And so on. This is an example of "lazy embedding".
Parsing
The following examples show parsers of the following type:
Parser(Toks) -> {ok, Tree, Toks1} | fail
Toks is the list of tokens to be parsed. A successful parse returns {ok, Tree, Toks1}.
•
•
Tree is a parse tree.
Toks1 is a tail of Tree that contains symbols encountered after the structure that was correctly parsed.
An unsuccessful parse returns fail.
The following example illustrates a simple, functional parser that parses the grammar:
(a | b) & (c | d)
The following code defines a function pconst(X) in the module funparse, which returns a fun that parses a list
of tokens:
pconst(X) ->
fun (T) ->
case T of
[X|T1] -> {ok, {const, X}, T1};
_
-> fail
end
end.
This function can be used as follows:
> P1 = funparse:pconst(a).
#Fun<funparse.0.22674075>
> P1([a,b,c]).
{ok,{const,a},[b,c]}
> P1([x,y,z]).
fail
Next, the two higher order functions pand and por are defined. They combine primitive parsers to produce more
complex parsers.
First pand:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 205
6.2 Funs
pand(P1, P2) ->
fun (T) ->
case P1(T) of
{ok, R1, T1} ->
case P2(T1) of
{ok, R2, T2} ->
{ok, {'and', R1, R2}};
fail ->
fail
end;
fail ->
fail
end
end.
Given a parser P1 for grammar G1, and a parser P2 for grammar G2, pand(P1, P2) returns a parser for the
grammar, which consists of sequences of tokens that satisfy G1, followed by sequences of tokens that satisfy G2.
por(P1, P2) returns a parser for the language described by the grammar G1 or G2:
por(P1, P2) ->
fun (T) ->
case P1(T) of
{ok, R, T1} ->
{ok, {'or',1,R}, T1};
fail ->
case P2(T) of
{ok, R1, T1} ->
{ok, {'or',2,R1}, T1};
fail ->
fail
end
end
end.
The original problem was to parse the grammar (a | b) & (c | d). The following code addresses this problem:
grammar() ->
pand(
por(pconst(a), pconst(b)),
por(pconst(c), pconst(d))).
The following code adds a parser interface to the grammar:
parse(List) ->
(grammar())(List).
The parser can be tested as follows:
> funparse:parse([a,c]).
206 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.3 List Comprehensions
{ok,{'and',{'or',1,{const,a}},{'or',1,{const,c}}}}
> funparse:parse([a,d]).
{ok,{'and',{'or',1,{const,a}},{'or',2,{const,d}}}}
> funparse:parse([b,c]).
{ok,{'and',{'or',2,{const,b}},{'or',1,{const,c}}}}
> funparse:parse([b,d]).
{ok,{'and',{'or',2,{const,b}},{'or',2,{const,d}}}}
> funparse:parse([a,b]).
fail
6.3 List Comprehensions
6.3.1 Simple Examples
This section starts with a simple example, showing a generator and a filter:
> [X || X <- [1,2,a,3,4,b,5,6], X > 3].
[a,4,b,5,6]
This is read as follows: The list of X such that X is taken from the list [1,2,a,...] and X is greater than 3.
The notation X <- [1,2,a,...] is a generator and the expression X > 3 is a filter.
An additional filter, integer(X), can be added to restrict the result to integers:
> [X || X <- [1,2,a,3,4,b,5,6], integer(X), X > 3].
[4,5,6]
Generators can be combined. For example, the Cartesian product of two lists can be written as follows:
> [{X, Y} || X <- [1,2,3], Y <- [a,b]].
[{1,a},{1,b},{2,a},{2,b},{3,a},{3,b}]
6.3.2 Quick Sort
The well-known quick sort routine can be written as follows:
sort([Pivot|T]) ->
sort([ X || X <- T, X < Pivot]) ++
[Pivot] ++
sort([ X || X <- T, X >= Pivot]);
sort([]) -> [].
The expression [X || X <- T, X < Pivot] is the list of all elements in T that are less than Pivot.
[X || X <- T, X >= Pivot] is the list of all elements in T that are greater than or equal to Pivot.
A list sorted as follows:
•
•
•
The first element in the list is isolated and the list is split into two sublists.
The first sublist contains all elements that are smaller than the first element in the list.
The second sublist contains all elements that are greater than, or equal to, the first element in the list.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 207
6.3 List Comprehensions
•
Then the sublists are sorted and the results are combined.
6.3.3 Permutations
The following example generates all permutations of the elements in a list:
perms([]) -> [[]];
perms(L) -> [[H|T] || H <- L, T <- perms(L--[H])].
This takes H from L in all possible ways. The result is the set of all lists [H|T], where T is the set of all possible
permutations of L, with H removed:
> perms([b,u,g]).
[[b,u,g],[b,g,u],[u,b,g],[u,g,b],[g,b,u],[g,u,b]]
6.3.4 Pythagorean Triplets
Pythagorean triplets are sets of integers {A,B,C} such that A**2 + B**2 = C**2.
The function pyth(N) generates a list of all integers {A,B,C} such that A**2 + B**2 = C**2 and where the
sum of the sides is equal to, or less than, N:
pyth(N) ->
[ {A,B,C} ||
A <- lists:seq(1,N),
B <- lists:seq(1,N),
C <- lists:seq(1,N),
A+B+C =< N,
A*A+B*B == C*C
].
> pyth(3).
[].
> pyth(11).
[].
> pyth(12).
[{3,4,5},{4,3,5}]
> pyth(50).
[{3,4,5},
{4,3,5},
{5,12,13},
{6,8,10},
{8,6,10},
{8,15,17},
{9,12,15},
{12,5,13},
{12,9,15},
{12,16,20},
{15,8,17},
{16,12,20}]
The following code reduces the search space and is more efficient:
208 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.3 List Comprehensions
pyth1(N) ->
[{A,B,C} ||
A <- lists:seq(1,N-2),
B <- lists:seq(A+1,N-1),
C <- lists:seq(B+1,N),
A+B+C =< N,
A*A+B*B == C*C ].
6.3.5 Simplifications With List Comprehensions
As an example, list comprehensions can be used to simplify some of the functions in lists.erl:
append(L)
-> [X || L1 <- L, X <- L1].
map(Fun, L) -> [Fun(X) || X <- L].
filter(Pred, L) -> [X || X <- L, Pred(X)].
6.3.6 Variable Bindings in List Comprehensions
The scope rules for variables that occur in list comprehensions are as follows:
•
•
•
All variables that occur in a generator pattern are assumed to be "fresh" variables.
Any variables that are defined before the list comprehension, and that are used in filters, have the values they
had before the list comprehension.
Variables cannot be exported from a list comprehension.
As an example of these rules, suppose you want to write the function select, which selects certain elements from a
list of tuples. Suppose you write select(X, L) -> [Y || {X, Y} <- L]. with the intention of extracting
all tuples from L, where the first item is X.
Compiling this gives the following diagnostic:
./FileName.erl:Line: Warning: variable 'X' shadowed in generate
This diagnostic warns that the variable X in the pattern is not the same as the variable X that occurs in the function head.
Evaluating select gives the following result:
> select(b,[{a,1},{b,2},{c,3},{b,7}]).
[1,2,3,7]
This is not the wanted result. To achieve the desired effect, select must be written as follows:
select(X, L) ->
[Y || {X1, Y} <- L, X == X1].
The generator now contains unbound variables and the test has been moved into the filter.
This now works as expected:
> select(b,[{a,1},{b,2},{c,3},{b,7}]).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 209
6.4 Bit Syntax
[2,7]
A consequence of the rules for importing variables into a list comprehensions is that certain pattern matching operations
must be moved into the filters and cannot be written directly in the generators.
To illustrate this, do not write as follows:
f(...) ->
Y = ...
[ Expression || PatternInvolving Y
...
<- Expr, ...]
Instead, write as follows:
f(...) ->
Y = ...
[ Expression || PatternInvolving Y1
...
<- Expr, Y == Y1, ...]
6.4 Bit Syntax
6.4.1 Introduction
The complete specification for the bit syntax appears in the Reference Manual.
In Erlang, a Bin is used for constructing binaries and matching binary patterns. A Bin is written with the following
syntax:
<<E1, E2, ... En>>
A Bin is a low-level sequence of bits or bytes. The purpose of a Bin is to enable construction of binaries:
Bin = <<E1, E2, ... En>>
All elements must be bound. Or match a binary:
<<E1, E2, ... En>> = Bin
Here, Bin is bound and the elements are bound or unbound, as in any match.
A Bin does not need to consist of a whole number of bytes.
A bitstring is a sequence of zero or more bits, where the number of bits does not need to be divisible by 8. If the
number of bits is divisible by 8, the bitstring is also a binary.
Each element specifies a certain segment of the bitstring. A segment is a set of contiguous bits of the binary (not
necessarily on a byte boundary). The first element specifies the initial segment, the second element specifies the
following segment, and so on.
210 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.4 Bit Syntax
The following examples illustrate how binaries are constructed, or matched, and how elements and tails are specified.
Examples
Example 1: A binary can be constructed from a set of constants or a string literal:
Bin11 = <<1, 17, 42>>,
Bin12 = <<"abc">>
This gives two binaries of size 3, with the following evaluations:
•
•
binary_to_list(Bin11) evaluates to [1, 17, 42].
binary_to_list(Bin12) evaluates to [97, 98, 99].
Example 2:Similarly, a binary can be constructed from a set of bound variables:
A = 1, B = 17, C = 42,
Bin2 = <<A, B, C:16>>
This gives a binary of size 4. Here, a size expression is used for the variable C to specify a 16-bits segment of Bin2.
binary_to_list(Bin2) evaluates to [1, 17, 00, 42].
Example 3: A Bin can also be used for matching. D, E, and F are unbound variables, and Bin2 is bound, as in
Example 2:
<<D:16, E, F/binary>> = Bin2
This gives D = 273, E = 00, and F binds to a binary of size 1: binary_to_list(F) = [42].
Example 4: The following is a more elaborate example of matching. Here, Dgram is bound to the consecutive bytes
of an IP datagram of IP protocol version 4. The ambition is to extract the header and the data of the datagram:
-define(IP_VERSION, 4).
-define(IP_MIN_HDR_LEN, 5).
DgramSize = byte_size(Dgram),
case Dgram of
<<?IP_VERSION:4, HLen:4, SrvcType:8, TotLen:16,
ID:16, Flgs:3, FragOff:13,
TTL:8, Proto:8, HdrChkSum:16,
SrcIP:32,
DestIP:32, RestDgram/binary>> when HLen>=5, 4*HLen=<DgramSize ->
OptsLen = 4*(HLen - ?IP_MIN_HDR_LEN),
<<Opts:OptsLen/binary,Data/binary>> = RestDgram,
...
end.
Here, the segment corresponding to the Opts variable has a type modifier, specifying that Opts is to bind to a binary.
All other variables have the default type equal to unsigned integer.
An IP datagram header is of variable length. This length is measured in the number of 32-bit words and is given in the
segment corresponding to HLen. The minimum value of HLen is 5. It is the segment corresponding to Opts that is
variable, so if HLen is equal to 5, Opts becomes an empty binary.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 211
6.4 Bit Syntax
The tail variables RestDgram and Data bind to binaries, as all tail variables do. Both can bind to empty binaries.
The match of Dgram fails if one of the following occurs:
•
•
•
The first 4-bits segment of Dgram is not equal to 4.
HLen is less than 5.
The size of Dgram is less than 4*HLen.
6.4.2 Lexical Note
Notice that "B=<<1>>" will be interpreted as "B =< <1>>", which is a syntax error. The correct way to write the
expression is: B = <<1>>.
6.4.3 Segments
Each segment has the following general syntax:
Value:Size/TypeSpecifierList
The Size or the TypeSpecifier, or both, can be omitted. Thus, the following variants are allowed:
•
•
•
Value
Value:Size
Value/TypeSpecifierList
Default values are used when specifications are missing. The default values are described in Defaults.
The Value part is any expression, when used in binary construction. Used in binary matching, the Value part must
be a literal or a variable. For more information about the Value part, see Constructing Binaries and Bitstrings and
Matching Binaries.
The Size part of the segment multiplied by the unit in TypeSpecifierList (described later) gives the number
of bits for the segment. In construction, Size is any expression that evaluates to an integer. In matching, Size must
be a constant expression or a variable.
The TypeSpecifierList is a list of type specifiers separated by hyphens.
Type
The most commonly used types are integer, float, and binary. See Bit Syntax Expressions in the
Reference Manual for a complete description.
Signedness
The signedness specification can be either signed or unsigned. Notice that signedness only matters for
matching.
Endianness
The endianness specification can be either big, little, or native. Native-endian means that the endian is
resolved at load time, to be either big-endian or little-endian, depending on what is "native" for the CPU that
the Erlang machine is run on.
Unit
The unit size is given as unit:IntegerLiteral. The allowed range is 1-256. It is multiplied by the Size
specifier to give the effective size of the segment. The unit size specifies the alignment for binary segments
without size.
Example:
X:4/little-signed-integer-unit:8
This element has a total size of 4*8 = 32 bits, and it contains a signed integer in little-endian order.
212 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.4 Bit Syntax
6.4.4 Defaults
The default type for a segment is integer. The default type does not depend on the value, even if the value is a literal.
For example, the default type in <<3.14>> is integer, not float.
The default Size depends on the type. For integer it is 8. For float it is 64. For binary it is all of the binary. In matching,
this default value is only valid for the last element. All other binary elements in matching must have a size specification.
The default unit depends on the the type. For integer, float, and bitstring it is 1. For binary it is 8.
The default signedness is unsigned.
The default endianness is big.
6.4.5 Constructing Binaries and Bitstrings
This section describes the rules for constructing binaries using the bit syntax. Unlike when constructing lists or tuples,
the construction of a binary can fail with a badarg exception.
There can be zero or more segments in a binary to be constructed. The expression <<>> constructs a zero length binary.
Each segment in a binary can consist of zero or more bits. There are no alignment rules for individual segments of
type integer and float. For binaries and bitstrings without size, the unit specifies the alignment. Since the default
alignment for the binary type is 8, the size of a binary segment must be a multiple of 8 bits, that is, only whole bytes.
Example:
<<Bin/binary,Bitstring/bitstring>>
The variable Bin must contain a whole number of bytes, because the binary type defaults to unit:8. A badarg
exception is generated if Bin consist of, for example, 17 bits.
The Bitstring variable can consist of any number of bits, for example, 0, 1, 8, 11, 17, 42, and so on. This is because
the default unit for bitstrings is 1.
For clarity, it is recommended not to change the unit size for binaries. Instead, use binary when you need byte
alignment and bitstring when you need bit alignment.
The following example successfully constructs a bitstring of 7 bits, provided that all of X and Y are integers:
<<X:1,Y:6>>
As mentioned earlier, segments have the following general syntax:
Value:Size/TypeSpecifierList
When constructing binaries, Value and Size can be any Erlang expression. However, for syntactical reasons, both
Value and Size must be enclosed in parenthesis if the expression consists of anything more than a single literal or
a variable. The following gives a compiler syntax error:
<<X+1:8>>
This expression must be rewritten into the following, to be accepted by the compiler:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 213
6.4 Bit Syntax
<<(X+1):8>>
Including Literal Strings
A literal string can be written instead of an element:
<<"hello">>
This is syntactic sugar for the following:
<<$h,$e,$l,$l,$o>>
6.4.6 Matching Binaries
This section describes the rules for matching binaries, using the bit syntax.
There can be zero or more segments in a binary pattern. A binary pattern can occur wherever patterns are allowed,
including inside other patterns. Binary patterns cannot be nested. The pattern <<>> matches a zero length binary.
Each segment in a binary can consist of zero or more bits. A segment of type binary must have a size evenly divisible
by 8 (or divisible by the unit size, if the unit size has been changed). A segment of type bitstring has no restrictions
on the size.
As mentioned earlier, segments have the following general syntax:
Value:Size/TypeSpecifierList
When matching Value, value must be either a variable or an integer, or a floating point literal. Expressions are not
allowed.
Size must be an integer literal, or a previously bound variable. The following is not allowed:
foo(N, <<X:N,T/binary>>) ->
{X,T}.
The two occurrences of N are not related. The compiler will complain that the N in the size field is unbound.
The correct way to write this example is as follows:
foo(N, Bin) ->
<<X:N,T/binary>> = Bin,
{X,T}.
Getting the Rest of the Binary or Bitstring
To match out the rest of a binary, specify a binary field without size:
foo(<<A:8,Rest/binary>>) ->
The size of the tail must be evenly divisible by 8.
To match out the rest of a bitstring, specify a field without size:
214 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
6.4 Bit Syntax
foo(<<A:8,Rest/bitstring>>) ->
There are no restrictions on the number of bits in the tail.
6.4.7 Appending to a Binary
Appending to a binary in an efficient way can be done as follows:
triples_to_bin(T) ->
triples_to_bin(T, <<>>).
triples_to_bin([{X,Y,Z} | T], Acc) ->
triples_to_bin(T, <<Acc/binary,X:32,Y:32,Z:32>>);
triples_to_bin([], Acc) ->
Acc.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 215
7.1 Introduction
7 Efficiency Guide
7.1 Introduction
7.1.1 Purpose
"Premature optimization is the root of all evil" (D.E. Knuth)
Efficient code can be well-structured and clean, based on a sound overall architecture and sound algorithms. Efficient
code can be highly implementation-code that bypasses documented interfaces and takes advantage of obscure quirks
in the current implementation.
Ideally, your code only contains the first type of efficient code. If that turns out to be too slow, profile the application
to find out where the performance bottlenecks are and optimize only the bottlenecks. Let other code stay as clean as
possible.
This Efficiency Guide cannot really teach you how to write efficient code. It can give you a few pointers about what to
avoid and what to use, and some understanding of how certain language features are implemented. This guide does not
include general tips about optimization that works in any language, such as moving common calculations out of loops.
7.1.2 Prerequisites
It is assumed that you are familiar with the Erlang programming language and the OTP concepts.
7.2 The Seven Myths of Erlang Performance
Some truths seem to live on well beyond their best-before date, perhaps because "information" spreads faster from
person-to-person than a single release note that says, for example, that body-recursive calls have become faster.
This section tries to kill the old truths (or semi-truths) that have become myths.
7.2.1 Myth: Tail-Recursive Functions are Much Faster Than Recursive
Functions
According to the myth, using a tail-recursive function that builds a list in reverse followed by a call to
lists:reverse/1 is faster than a body-recursive function that builds the list in correct order; the reason being
that body-recursive functions use more memory than tail-recursive functions.
That was true to some extent before R12B. It was even more true before R7B. Today, not so much. A body-recursive
function generally uses the same amount of memory as a tail-recursive function. It is generally not possible to predict
whether the tail-recursive or the body-recursive version will be faster. Therefore, use the version that makes your code
cleaner (hint: it is usually the body-recursive version).
For a more thorough discussion about tail and body recursion, see Erlang's Tail Recursion is Not a Silver Bullet.
216 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.2 The Seven Myths of Erlang Performance
Note:
A tail-recursive function that does not need to reverse the list at the end is faster than a body-recursive function,
as are tail-recursive functions that do not construct any terms at all (for example, a function that sums all integers
in a list).
7.2.2 Myth: Operator "++" is Always Bad
The ++ operator has, somewhat undeservedly, got a bad reputation. It probably has something to do with code like
the following, which is the most inefficient way there is to reverse a list:
DO NOT
naive_reverse([H|T]) ->
naive_reverse(T)++[H];
naive_reverse([]) ->
[].
As the ++ operator copies its left operand, the result is copied repeatedly, leading to quadratic complexity.
But using ++ as follows is not bad:
OK
naive_but_ok_reverse([H|T], Acc) ->
naive_but_ok_reverse(T, [H]++Acc);
naive_but_ok_reverse([], Acc) ->
Acc.
Each list element is copied only once. The growing result Acc is the right operand for the ++ operator, and it is not
copied.
Experienced Erlang programmers would write as follows:
DO
vanilla_reverse([H|T], Acc) ->
vanilla_reverse(T, [H|Acc]);
vanilla_reverse([], Acc) ->
Acc.
This is slightly more efficient because here you do not build a list element only to copy it directly. (Or it would be
more efficient if the compiler did not automatically rewrite [H]++Acc to [H|Acc].)
7.2.3 Myth: Strings are Slow
String handling can be slow if done improperly. In Erlang, you need to think a little more about how the strings are
used and choose an appropriate representation. If you use regular expressions, use the re module in STDLIB instead
of the obsolete regexp module.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 217
7.3 Common Caveats
7.2.4 Myth: Repairing a Dets File is Very Slow
The repair time is still proportional to the number of records in the file, but Dets repairs used to be much slower in
the past. Dets has been massively rewritten and improved.
7.2.5 Myth: BEAM is a Stack-Based Byte-Code Virtual Machine (and
Therefore Slow)
BEAM is a register-based virtual machine. It has 1024 virtual registers that are used for holding temporary values and
for passing arguments when calling functions. Variables that need to survive a function call are saved to the stack.
BEAM is a threaded-code interpreter. Each instruction is word pointing directly to executable C-code, making
instruction dispatching very fast.
7.2.6 Myth: Use "_" to Speed Up Your Program When a Variable is Not
Used
That was once true, but from R6B the BEAM compiler can see that a variable is not used.
Similarly, trivial transformations on the source-code level such as converting a case statement to clauses at the toplevel of the function seldom makes any difference to the generated code.
7.2.7 Myth: A NIF Always Speeds Up Your Program
Rewriting Erlang code to a NIF to make it faster should be seen as a last resort. It is only guaranteed to be dangerous,
but not guaranteed to speed up the program.
Doing too much work in each NIF call will degrade responsiveness of the VM. Doing too little work may mean that
the gain of the faster processing in the NIF is eaten up by the overhead of calling the NIF and checking the arguments.
Be sure to read about Long-running NIFs before writing a NIF.
7.3 Common Caveats
This section lists a few modules and BIFs to watch out for, not only from a performance point of view.
7.3.1 Timer Module
Creating timers using erlang:send_after/3 and erlang:start_timer/3 , is much more efficient than using the timers
provided by the timer module in STDLIB. The timer module uses a separate process to manage the timers. That
process can easily become overloaded if many processes create and cancel timers frequently (especially when using
the SMP emulator).
The functions in the timer module that do not manage timers (such as timer:tc/3 or timer:sleep/1), do
not call the timer-server process and are therefore harmless.
7.3.2 list_to_atom/1
Atoms are not garbage-collected. Once an atom is created, it is never removed. The emulator terminates if the limit
for the number of atoms (1,048,576 by default) is reached.
Therefore, converting arbitrary input strings to atoms can be dangerous in a system that runs continuously. If only
certain well-defined atoms are allowed as input, list_to_existing_atom/1 can be used to to guard against a denial-ofservice attack. (All atoms that are allowed must have been created earlier, for example, by simply using all of them
in a module and loading that module.)
218 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.3 Common Caveats
Using list_to_atom/1 to construct an atom that is passed to apply/3 as follows, is quite expensive and not
recommended in time-critical code:
apply(list_to_atom("some_prefix"++Var), foo, Args)
7.3.3 length/1
The time for calculating the length of a list is proportional to the length of the list, as opposed to tuple_size/1,
byte_size/1, and bit_size/1, which all execute in constant time.
Normally, there is no need to worry about the speed of length/1, because it is efficiently implemented in C. In
time-critical code, you might want to avoid it if the input list could potentially be very long.
Some uses of length/1 can be replaced by matching. For example, the following code:
foo(L) when length(L) >= 3 ->
...
can be rewritten to:
foo([_,_,_|_]=L) ->
...
One slight difference is that length(L) fails if L is an improper list, while the pattern in the second code fragment
accepts an improper list.
7.3.4 setelement/3
setelement/3 copies the tuple it modifies. Therefore, updating a tuple in a loop using setelement/3 creates a new
copy of the tuple every time.
There is one exception to the rule that the tuple is copied. If the compiler clearly can see that destructively updating
the tuple would give the same result as if the tuple was copied, the call to setelement/3 is replaced with a special
destructive setelement instruction. In the following code sequence, the first setelement/3 call copies the tuple
and modifies the ninth element:
multiple_setelement(T0) ->
T1 = setelement(9, T0, bar),
T2 = setelement(7, T1, foobar),
setelement(5, T2, new_value).
The two following setelement/3 calls modify the tuple in place.
For the optimization to be applied, all the followings conditions must be true:
•
•
•
•
The indices must be integer literals, not variables or expressions.
The indices must be given in descending order.
There must be no calls to another function in between the calls to setelement/3.
The tuple returned from one setelement/3 call must only be used in the subsequent call to
setelement/3.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 219
7.3 Common Caveats
If the code cannot be structured as in the multiple_setelement/1 example, the best way to modify multiple
elements in a large tuple is to convert the tuple to a list, modify the list, and convert it back to a tuple.
7.3.5 size/1
size/1 returns the size for both tuples and binaries.
Using the BIFs tuple_size/1 and byte_size/1 gives the compiler and the runtime system more opportunities
for optimization. Another advantage is that the BIFs give Dialyzer more type information.
7.3.6 split_binary/2
It is usually more efficient to split a binary using matching instead of calling the split_binary/2 function.
Furthermore, mixing bit syntax matching and split_binary/2 can prevent some optimizations of bit syntax
matching.
DO
<<Bin1:Num/binary,Bin2/binary>> = Bin,
DO NOT
{Bin1,Bin2} = split_binary(Bin, Num)
7.3.7 Operator "--"
The "--" operator has a complexity proportional to the product of the length of its operands. This means that the
operator is very slow if both of its operands are long lists:
DO NOT
HugeList1 -- HugeList2
Instead use the ordsets module in STDLIB:
DO
HugeSet1 = ordsets:from_list(HugeList1),
HugeSet2 = ordsets:from_list(HugeList2),
ordsets:subtract(HugeSet1, HugeSet2)
Obviously, that code does not work if the original order of the list is important. If the order of the list must be preserved,
do as follows:
DO
Set = gb_sets:from_list(HugeList2),
[E || E <- HugeList1, not gb_sets:is_element(E, Set)]
220 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.4 Constructing and Matching Binaries
Note:
This code behaves differently from "--" if the lists contain duplicate elements (one occurrence of an element in
HugeList2 removes all occurrences in HugeList1.)
Also, this code compares lists elements using the "==" operator, while "--" uses the "=:=" operator. If that
difference is important, sets can be used instead of gb_sets, but sets:from_list/1 is much slower than
gb_sets:from_list/1 for long lists.
Using the "--" operator to delete an element from a list is not a performance problem:
OK
HugeList1 -- [Element]
7.4 Constructing and Matching Binaries
Binaries can be efficiently built in the following way:
DO
my_list_to_binary(List) ->
my_list_to_binary(List, <<>>).
my_list_to_binary([H|T], Acc) ->
my_list_to_binary(T, <<Acc/binary,H>>);
my_list_to_binary([], Acc) ->
Acc.
Binaries can be efficiently matched like this:
DO
my_binary_to_list(<<H,T/binary>>) ->
[H|my_binary_to_list(T)];
my_binary_to_list(<<>>) -> [].
7.4.1 How Binaries are Implemented
Internally, binaries and bitstrings are implemented in the same way. In this section, they are called binaries because
that is what they are called in the emulator source code.
Four types of binary objects are available internally:
•
Two are containers for binary data and are called:
•
• Refc binaries (short for reference-counted binaries)
• Heap binaries
Two are merely references to a part of a binary and are called:
•
•
sub binaries
match contexts
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 221
7.4 Constructing and Matching Binaries
Refc Binaries
Refc binaries consist of two parts:
•
•
An object stored on the process heap, called a ProcBin
The binary object itself, stored outside all process heaps
The binary object can be referenced by any number of ProcBins from any number of processes. The object contains a
reference counter to keep track of the number of references, so that it can be removed when the last reference disappears.
All ProcBin objects in a process are part of a linked list, so that the garbage collector can keep track of them and
decrement the reference counters in the binary when a ProcBin disappears.
Heap Binaries
Heap binaries are small binaries, up to 64 bytes, and are stored directly on the process heap. They are copied when
the process is garbage-collected and when they are sent as a message. They do not require any special handling by
the garbage collector.
Sub Binaries
The reference objects sub binaries and match contexts can reference part of a refc binary or heap binary.
A sub binary is created by split_binary/2 and when a binary is matched out in a binary pattern. A sub binary is
a reference into a part of another binary (refc or heap binary, but never into another sub binary). Therefore, matching
out a binary is relatively cheap because the actual binary data is never copied.
Match Context
A match context is similar to a sub binary, but is optimized for binary matching. For example, it contains a direct
pointer to the binary data. For each field that is matched out of a binary, the position in the match context is incremented.
The compiler tries to avoid generating code that creates a sub binary, only to shortly afterwards create a new match
context and discard the sub binary. Instead of creating a sub binary, the match context is kept.
The compiler can only do this optimization if it knows that the match context will not be shared. If it would be shared,
the functional properties (also called referential transparency) of Erlang would break.
7.4.2 Constructing Binaries
Appending to a binary or bitstring is specially optimized by the runtime system:
<<Binary/binary, ...>>
<<Binary/bitstring, ...>>
As the runtime system handles the optimization (instead of the compiler), there are very few circumstances in which
the optimization does not work.
To explain how it works, let us examine the following code line by line:
Bin0 = <<0>>,
Bin1 = <<Bin0/binary,1,2,3>>,
Bin2 = <<Bin1/binary,4,5,6>>,
Bin3 = <<Bin2/binary,7,8,9>>,
Bin4 = <<Bin1/binary,17>>,
{Bin4,Bin3}
•
%%
%%
%%
%%
%%
%%
1
2
3
4
5 !!!
6
Line 1 (marked with the %% 1 comment), assigns a heap binary to the Bin0 variable.
222 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.4 Constructing and Matching Binaries
•
•
•
•
Line 2 is an append operation. As Bin0 has not been involved in an append operation, a new refc binary is
created and the contents of Bin0 is copied into it. The ProcBin part of the refc binary has its size set to the size
of the data stored in the binary, while the binary object has extra space allocated. The size of the binary object is
either twice the size of Bin1 or 256, whichever is larger. In this case it is 256.
Line 3 is more interesting. Bin1 has been used in an append operation, and it has 252 bytes of unused storage
at the end, so the 3 new bytes are stored there.
Line 4. The same applies here. There are 249 bytes left, so there is no problem storing another 3 bytes.
Line 5. Here, something interesting happens. Notice that the result is not appended to the previous
result in Bin3, but to Bin1. It is expected that Bin4 will be assigned the value <<0,1,2,3,17>>.
It is also expected that Bin3 will retain its value (<<0,1,2,3,4,5,6,7,8,9>>). Clearly, the
runtime system cannot write byte 17 into the binary, because that would change the value of Bin3 to
<<0,1,2,3,4,17,6,7,8,9>>.
The runtime system sees that Bin1 is the result from a previous append operation (not from the latest append
operation), so it copies the contents of Bin1 to a new binary, reserve extra storage, and so on. (Here is not explained
how the runtime system can know that it is not allowed to write into Bin1; it is left as an exercise to the curious reader
to figure out how it is done by reading the emulator sources, primarily erl_bits.c.)
Circumstances That Force Copying
The optimization of the binary append operation requires that there is a single ProcBin and a single reference to the
ProcBin for the binary. The reason is that the binary object can be moved (reallocated) during an append operation,
and when that happens, the pointer in the ProcBin must be updated. If there would be more than one ProcBin pointing
to the binary object, it would not be possible to find and update all of them.
Therefore, certain operations on a binary mark it so that any future append operation will be forced to copy the binary.
In most cases, the binary object will be shrunk at the same time to reclaim the extra space allocated for growing.
When appending to a binary as follows, only the binary returned from the latest append operation will support further
cheap append operations:
Bin = <<Bin0,...>>
In the code fragment in the beginning of this section, appending to Bin will be cheap, while appending to Bin0 will
force the creation of a new binary and copying of the contents of Bin0.
If a binary is sent as a message to a process or port, the binary will be shrunk and any further append operation will copy
the binary data into a new binary. For example, in the following code fragment Bin1 will be copied in the third line:
Bin1 = <<Bin0,...>>,
PortOrPid ! Bin1,
Bin = <<Bin1,...>> %% Bin1 will be COPIED
The same happens if you insert a binary into an Ets table, send it to a port using erlang:port_command/2, or
pass it to enif_inspect_binary in a NIF.
Matching a binary will also cause it to shrink and the next append operation will copy the binary data:
Bin1 = <<Bin0,...>>,
<<X,Y,Z,T/binary>> = Bin1,
Bin = <<Bin1,...>> %% Bin1 will be COPIED
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 223
7.4 Constructing and Matching Binaries
The reason is that a match context contains a direct pointer to the binary data.
If a process simply keeps binaries (either in "loop data" or in the process dictionary), the garbage collector can
eventually shrink the binaries. If only one such binary is kept, it will not be shrunk. If the process later appends to a
binary that has been shrunk, the binary object will be reallocated to make place for the data to be appended.
7.4.3 Matching Binaries
Let us revisit the example in the beginning of the previous section:
DO
my_binary_to_list(<<H,T/binary>>) ->
[H|my_binary_to_list(T)];
my_binary_to_list(<<>>) -> [].
The first time my_binary_to_list/1 is called, a match context is created. The match context points to the first
byte of the binary. 1 byte is matched out and the match context is updated to point to the second byte in the binary.
At this point it would make sense to create a sub binary, but in this particular example the compiler sees that there
will soon be a call to a function (in this case, to my_binary_to_list/1 itself) that immediately will create a new
match context and discard the sub binary.
Therefore my_binary_to_list/1 calls itself with the match context instead of with a sub binary. The instruction
that initializes the matching operation basically does nothing when it sees that it was passed a match context instead
of a binary.
When the end of the binary is reached and the second clause matches, the match context will simply be discarded
(removed in the next garbage collection, as there is no longer any reference to it).
To summarize, my_binary_to_list/1 only needs to create one match context and no sub binaries.
Notice that the match context in my_binary_to_list/1 was discarded when the entire binary had been traversed.
What happens if the iteration stops before it has reached the end of the binary? Will the optimization still work?
after_zero(<<0,T/binary>>) ->
T;
after_zero(<<_,T/binary>>) ->
after_zero(T);
after_zero(<<>>) ->
<<>>.
Yes, it will. The compiler will remove the building of the sub binary in the second clause:
...
after_zero(<<_,T/binary>>) ->
after_zero(T);
...
But it will generate code that builds a sub binary in the first clause:
after_zero(<<0,T/binary>>) ->
T;
224 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.4 Constructing and Matching Binaries
...
Therefore, after_zero/1 builds one match context and one sub binary (assuming it is passed a binary that contains
a zero byte).
Code like the following will also be optimized:
all_but_zeroes_to_list(Buffer, Acc, 0) ->
{lists:reverse(Acc),Buffer};
all_but_zeroes_to_list(<<0,T/binary>>, Acc, Remaining) ->
all_but_zeroes_to_list(T, Acc, Remaining-1);
all_but_zeroes_to_list(<<Byte,T/binary>>, Acc, Remaining) ->
all_but_zeroes_to_list(T, [Byte|Acc], Remaining-1).
The compiler removes building of sub binaries in the second and third clauses, and it adds an instruction to the first
clause that converts Buffer from a match context to a sub binary (or do nothing if Buffer is a binary already).
Before you begin to think that the compiler can optimize any binary patterns, the following function cannot be
optimized by the compiler (currently, at least):
non_opt_eq([H|T1], <<H,T2/binary>>) ->
non_opt_eq(T1, T2);
non_opt_eq([_|_], <<_,_/binary>>) ->
false;
non_opt_eq([], <<>>) ->
true.
It was mentioned earlier that the compiler can only delay creation of sub binaries if it knows that the binary will not
be shared. In this case, the compiler cannot know.
Soon it is shown how to rewrite non_opt_eq/2 so that the delayed sub binary optimization can be applied, and
more importantly, it is shown how you can find out whether your code can be optimized.
Option bin_opt_info
Use the bin_opt_info option to have the compiler print a lot of information about binary optimizations. It can be
given either to the compiler or erlc:
erlc +bin_opt_info Mod.erl
or passed through an environment variable:
export ERL_COMPILER_OPTIONS=bin_opt_info
Notice that the bin_opt_info is not meant to be a permanent option added to your Makefiles, because all
messages that it generates cannot be eliminated. Therefore, passing the option through the environment is in most
cases the most practical approach.
The warnings look as follows:
./efficiency_guide.erl:60: Warning: NOT OPTIMIZED: sub binary is used or returned
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 225
7.4 Constructing and Matching Binaries
./efficiency_guide.erl:62: Warning: OPTIMIZED: creation of sub binary delayed
To make it clearer exactly what code the warnings refer to, the warnings in the following examples are inserted as
comments after the clause they refer to, for example:
after_zero(<<0,T/binary>>) ->
%% NOT OPTIMIZED: sub binary is used or returned
T;
after_zero(<<_,T/binary>>) ->
%% OPTIMIZED: creation of sub binary delayed
after_zero(T);
after_zero(<<>>) ->
<<>>.
The warning for the first clause says that the creation of a sub binary cannot be delayed, because it will be returned.
The warning for the second clause says that a sub binary will not be created (yet).
Let us revisit the earlier example of the code that could not be optimized and find out why:
non_opt_eq([H|T1], <<H,T2/binary>>) ->
%% INFO: matching anything else but a plain variable to
%%
the left of binary pattern will prevent delayed
%%
sub binary optimization;
%%
SUGGEST changing argument order
%% NOT OPTIMIZED: called function non_opt_eq/2 does not
%%
begin with a suitable binary matching instruction
non_opt_eq(T1, T2);
non_opt_eq([_|_], <<_,_/binary>>) ->
false;
non_opt_eq([], <<>>) ->
true.
The compiler emitted two warnings. The INFO warning refers to the function non_opt_eq/2 as a callee, indicating
that any function that call non_opt_eq/2 cannot make delayed sub binary optimization. There is also a suggestion
to change argument order. The second warning (that happens to refer to the same line) refers to the construction of
the sub binary itself.
Soon another example will show the difference between the INFO and NOT OPTIMIZED warnings somewhat clearer,
but let us first follow the suggestion to change argument order:
opt_eq(<<H,T1/binary>>, [H|T2]) ->
%% OPTIMIZED: creation of sub binary delayed
opt_eq(T1, T2);
opt_eq(<<_,_/binary>>, [_|_]) ->
false;
opt_eq(<<>>, []) ->
true.
The compiler gives a warning for the following code fragment:
match_body([0|_], <<H,_/binary>>) ->
%% INFO: matching anything else but a plain variable to
%%
the left of binary pattern will prevent delayed
%%
sub binary optimization;
226 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.5 List Handling
%%
...
SUGGEST changing argument order
done;
The warning means that if there is a call to match_body/2 (from another clause in match_body/2 or another
function), the delayed sub binary optimization will not be possible. More warnings will occur for any place where a
sub binary is matched out at the end of and passed as the second argument to match_body/2, for example:
match_head(List, <<_:10,Data/binary>>) ->
%% NOT OPTIMIZED: called function match_body/2 does not
%%
begin with a suitable binary matching instruction
match_body(List, Data).
Unused Variables
The compiler figures out if a variable is unused. The same code is generated for each of the following functions:
count1(<<_,T/binary>>, Count) -> count1(T, Count+1);
count1(<<>>, Count) -> Count.
count2(<<H,T/binary>>, Count) -> count2(T, Count+1);
count2(<<>>, Count) -> Count.
count3(<<_H,T/binary>>, Count) -> count3(T, Count+1);
count3(<<>>, Count) -> Count.
In each iteration, the first 8 bits in the binary will be skipped, not matched out.
7.4.4 Historical Note
Binary handling was significantly improved in R12B. Because code that was efficient in R11B might not be efficient
in R12B, and vice versa, earlier revisions of this Efficiency Guide contained some information about binary handling
in R11B.
7.5 List Handling
7.5.1 Creating a List
Lists can only be built starting from the end and attaching list elements at the beginning. If you use the "++" operator
as follows, a new list is created that is a copy of the elements in List1, followed by List2:
List1 ++ List2
Looking at how lists:append/1 or ++ would be implemented in plain Erlang, clearly the first list is copied:
append([H|T], Tail) ->
[H|append(T, Tail)];
append([], Tail) ->
Tail.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 227
7.5 List Handling
When recursing and building a list, it is important to ensure that you attach the new elements to the beginning of the
list. In this way, you will build one list, not hundreds or thousands of copies of the growing result list.
Let us first see how it is not to be done:
DO NOT
bad_fib(N) ->
bad_fib(N, 0, 1, []).
bad_fib(0, _Current, _Next, Fibs) ->
Fibs;
bad_fib(N, Current, Next, Fibs) ->
bad_fib(N - 1, Next, Current + Next, Fibs ++ [Current]).
Here more than one list is built. In each iteration step a new list is created that is one element longer than the new
previous list.
To avoid copying the result in each iteration, build the list in reverse order and reverse the list when you are done:
DO
tail_recursive_fib(N) ->
tail_recursive_fib(N, 0, 1, []).
tail_recursive_fib(0, _Current, _Next, Fibs) ->
lists:reverse(Fibs);
tail_recursive_fib(N, Current, Next, Fibs) ->
tail_recursive_fib(N - 1, Next, Current + Next, [Current|Fibs]).
7.5.2 List Comprehensions
Lists comprehensions still have a reputation for being slow. They used to be implemented using funs, which used
to be slow.
A list comprehension:
[Expr(E) || E <- List]
is basically translated to a local function:
'lc^0'([E|Tail], Expr) ->
[Expr(E)|'lc^0'(Tail, Expr)];
'lc^0'([], _Expr) -> [].
If the result of the list comprehension will obviously not be used, a list will not be constructed. For example, in this
code:
[io:put_chars(E) || E <- List],
ok.
or in this code:
228 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.5 List Handling
...
case Var of
... ->
[io:put_chars(E) || E <- List];
... ->
end,
some_function(...),
...
the value is not assigned to a variable, not passed to another function, and not returned. This means that there is no
need to construct a list and the compiler will simplify the code for the list comprehension to:
'lc^0'([E|Tail], Expr) ->
Expr(E),
'lc^0'(Tail, Expr);
'lc^0'([], _Expr) -> [].
The compiler also understands that assigning to '_' means that the value will not used. Therefore, the code in the
following example will also be optimized:
_ = [io:put_chars(E) || E <- List],
ok.
7.5.3 Deep and Flat Lists
lists:flatten/1 builds an entirely new list. It is therefore expensive, and even more expensive than the ++ operator
(which copies its left argument, but not its right argument).
In the following situations, you can easily avoid calling lists:flatten/1:
•
•
•
When sending data to a port. Ports understand deep lists so there is no reason to flatten the list before sending it
to the port.
When calling BIFs that accept deep lists, such as list_to_binary/1 or iolist_to_binary/1.
When you know that your list is only one level deep, you can use lists:append/1.
Port Example
DO
...
port_command(Port, DeepList)
...
DO NOT
...
port_command(Port, lists:flatten(DeepList))
...
A common way to send a zero-terminated string to a port is the following:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 229
7.5 List Handling
DO NOT
...
TerminatedStr = String ++ [0], % String="foo" => [$f, $o, $o, 0]
port_command(Port, TerminatedStr)
...
Instead:
DO
...
TerminatedStr = [String, 0], % String="foo" => [[$f, $o, $o], 0]
port_command(Port, TerminatedStr)
...
Append Example
DO
> lists:append([[1], [2], [3]]).
[1,2,3]
>
DO NOT
> lists:flatten([[1], [2], [3]]).
[1,2,3]
>
7.5.4 Recursive List Functions
In section about myths, the following myth was exposed: Tail-Recursive Functions are Much Faster Than Recursive
Functions.
There is usually not much difference between a body-recursive list function and tail-recursive function that reverses the
list at the end. Therefore, concentrate on writing beautiful code and forget about the performance of your list functions.
In the time-critical parts of your code (and only there), measure before rewriting your code.
Note:
This section is about list functions that construct lists. A tail-recursive function that does not construct a list runs
in constant space, while the corresponding body-recursive function uses stack space proportional to the length
of the list.
For example, a function that sums a list of integers, is not to be written as follows:
DO NOT
230 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.6 Functions
recursive_sum([H|T]) -> H+recursive_sum(T);
recursive_sum([])
-> 0.
Instead:
DO
sum(L) -> sum(L, 0).
sum([H|T], Sum) -> sum(T, Sum + H);
sum([], Sum)
-> Sum.
7.6 Functions
7.6.1 Pattern Matching
Pattern matching in function head as well as in case and receive clauses are optimized by the compiler. With a
few exceptions, there is nothing to gain by rearranging clauses.
One exception is pattern matching of binaries. The compiler does not rearrange clauses that match binaries. Placing
the clause that matches against the empty binary last is usually slightly faster than placing it first.
The following is a rather unnatural example to show another exception:
DO NOT
atom_map1(one) -> 1;
atom_map1(two) -> 2;
atom_map1(three) -> 3;
atom_map1(Int) when is_integer(Int) -> Int;
atom_map1(four) -> 4;
atom_map1(five) -> 5;
atom_map1(six) -> 6.
The problem is the clause with the variable Int. As a variable can match anything, including the atoms four, five,
and six, which the following clauses also match, the compiler must generate suboptimal code that executes as follows:
•
First, the input value is compared to one, two, and three (using a single instruction that does a binary
search; thus, quite efficient even if there are many values) to select which one of the first three clauses to
execute (if any).
If none of the first three clauses match, the fourth clause match as a variable always matches.
If the guard test is_integer(Int) succeeds, the fourth clause is executed.
If the guard test fails, the input value is compared to four, five, and six, and the appropriate clause is
selected. (There is a function_clause exception if none of the values matched.)
•
•
•
Rewriting to either:
DO
atom_map2(one) -> 1;
atom_map2(two) -> 2;
atom_map2(three) -> 3;
atom_map2(four) -> 4;
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 231
7.6 Functions
atom_map2(five) -> 5;
atom_map2(six) -> 6;
atom_map2(Int) when is_integer(Int) -> Int.
or:
DO
atom_map3(Int) when is_integer(Int) -> Int;
atom_map3(one) -> 1;
atom_map3(two) -> 2;
atom_map3(three) -> 3;
atom_map3(four) -> 4;
atom_map3(five) -> 5;
atom_map3(six) -> 6.
gives slightly more efficient matching code.
Another example:
DO NOT
map_pairs1(_Map, [], Ys) ->
Ys;
map_pairs1(_Map, Xs, [] ) ->
Xs;
map_pairs1(Map, [X|Xs], [Y|Ys]) ->
[Map(X, Y)|map_pairs1(Map, Xs, Ys)].
The first argument is not a problem. It is variable, but it is a variable in all clauses. The problem is the variable in the
second argument, Xs, in the middle clause. Because the variable can match anything, the compiler is not allowed to
rearrange the clauses, but must generate code that matches them in the order written.
If the function is rewritten as follows, the compiler is free to rearrange the clauses:
DO
map_pairs2(_Map, [], Ys) ->
Ys;
map_pairs2(_Map, [_|_]=Xs, [] ) ->
Xs;
map_pairs2(Map, [X|Xs], [Y|Ys]) ->
[Map(X, Y)|map_pairs2(Map, Xs, Ys)].
The compiler will generate code similar to this:
DO NOT (already done by the compiler)
explicit_map_pairs(Map, Xs0, Ys0) ->
case Xs0 of
[X|Xs] ->
case Ys0 of
[Y|Ys] ->
[Map(X, Y)|explicit_map_pairs(Map, Xs, Ys)];
[] ->
Xs0
232 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.6 Functions
end;
[] ->
Ys0
end.
This is slightly faster for probably the most common case that the input lists are not empty or very short. (Another
advantage is that Dialyzer can deduce a better type for the Xs variable.)
7.6.2 Function Calls
This is an intentionally rough guide to the relative costs of different calls. It is based on benchmark figures run on
Solaris/Sparc:
•
•
Calls to local or external functions (foo(), m:foo()) are the fastest calls.
Calling or applying a fun (Fun(), apply(Fun, [])) is about three times as expensive as calling a local
function.
Applying an exported function (Mod:Name(), apply(Mod, Name, [])) is about twice as expensive as
calling a fun or about six times as expensive as calling a local function.
•
Notes and Implementation Details
Calling and applying a fun does not involve any hash-table lookup. A fun contains an (indirect) pointer to the function
that implements the fun.
apply/3 must look up the code for the function to execute in a hash table. It is therefore always slower than a direct
call or a fun call.
It no longer matters (from a performance point of view) whether you write:
Module:Function(Arg1, Arg2)
or:
apply(Module, Function, [Arg1,Arg2])
The compiler internally rewrites the latter code into the former.
The following code is slightly slower because the shape of the list of arguments is unknown at compile time.
apply(Module, Function, Arguments)
7.6.3 Memory Usage in Recursion
When writing recursive functions, it is preferable to make them tail-recursive so that they can execute in constant
memory space:
DO
list_length(List) ->
list_length(List, 0).
list_length([], AccLen) ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 233
7.7 Tables and Databases
AccLen; % Base case
list_length([_|Tail], AccLen) ->
list_length(Tail, AccLen + 1). % Tail-recursive
DO NOT
list_length([]) ->
0. % Base case
list_length([_ | Tail]) ->
list_length(Tail) + 1. % Not tail-recursive
7.7 Tables and Databases
7.7.1 Ets, Dets, and Mnesia
Every example using Ets has a corresponding example in Mnesia. In general, all Ets examples also apply to Dets tables.
Select/Match Operations
Select/match operations on Ets and Mnesia tables can become very expensive operations. They usually need to scan
the complete table. Try to structure the data to minimize the need for select/match operations. However, if you require
a select/match operation, it is still more efficient than using tab2list. Examples of this and of how to avoid select/
match are provided in the following sections. The functions ets:select/2 and mnesia:select/3 are to be
preferred over ets:match/2, ets:match_object/2, and mnesia:match_object/3.
In some circumstances, the select/match operations do not need to scan the complete table. For example, if part of the
key is bound when searching an ordered_set table, or if it is a Mnesia table and there is a secondary index on the
field that is selected/matched. If the key is fully bound, there is no point in doing a select/match, unless you have a
bag table and are only interested in a subset of the elements with the specific key.
When creating a record to be used in a select/match operation, you want most of the fields to have the value "_". The
easiest and fastest way to do that is as follows:
#person{age = 42, _ = '_'}.
Deleting an Element
The delete operation is considered successful if the element was not present in the table. Hence all attempts to
check that the element is present in the Ets/Mnesia table before deletion are unnecessary. Here follows an example
for Ets tables:
DO
...
ets:delete(Tab, Key),
...
DO NOT
...
234 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.7 Tables and Databases
case ets:lookup(Tab, Key) of
[] ->
ok;
[_|_] ->
ets:delete(Tab, Key)
end,
...
Fetching Data
Do not fetch data that you already have.
Consider that you have a module that handles the abstract data type Person. You export the interface
function print_person/1, which uses the internal functions print_name/1, print_age/1, and
print_occupation/1.
Note:
If the function print_name/1, and so on, had been interface functions, the situation would have been different,
as you do not want the user of the interface to know about the internal data representation.
DO
%%% Interface function
print_person(PersonId) ->
%% Look up the person in the named table person,
case ets:lookup(person, PersonId) of
[Person] ->
print_name(Person),
print_age(Person),
print_occupation(Person);
[] ->
io:format("No person with ID = ~p~n", [PersonID])
end.
%%% Internal functions
print_name(Person) ->
io:format("No person ~p~n", [Person#person.name]).
print_age(Person) ->
io:format("No person ~p~n", [Person#person.age]).
print_occupation(Person) ->
io:format("No person ~p~n", [Person#person.occupation]).
DO NOT
%%% Interface function
print_person(PersonId) ->
%% Look up the person in the named table person,
case ets:lookup(person, PersonId) of
[Person] ->
print_name(PersonID),
print_age(PersonID),
print_occupation(PersonID);
[] ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 235
7.7 Tables and Databases
end.
io:format("No person with ID = ~p~n", [PersonID])
%%% Internal functionss
print_name(PersonID) ->
[Person] = ets:lookup(person, PersonId),
io:format("No person ~p~n", [Person#person.name]).
print_age(PersonID) ->
[Person] = ets:lookup(person, PersonId),
io:format("No person ~p~n", [Person#person.age]).
print_occupation(PersonID) ->
[Person] = ets:lookup(person, PersonId),
io:format("No person ~p~n", [Person#person.occupation]).
Non-Persistent Database Storage
For non-persistent database storage, prefer Ets tables over Mnesia local_content tables. Even the Mnesia
dirty_write operations carry a fixed overhead compared to Ets writes. Mnesia must check if the table is replicated
or has indices, this involves at least one Ets lookup for each dirty_write. Thus, Ets writes is always faster than
Mnesia writes.
tab2list
Assuming an Ets table that uses idno as key and contains the following:
[#person{idno
#person{idno
#person{idno
#person{idno
=
=
=
=
1,
2,
3,
4,
name
name
name
name
=
=
=
=
"Adam",
"Bryan",
"Bryan",
"Carl",
age
age
age
age
=
=
=
=
31,
31,
35,
25,
occupation
occupation
occupation
occupation
=
=
=
=
"mailman"},
"cashier"},
"banker"},
"mailman"}]
If you must return all data stored in the Ets table, you can use ets:tab2list/1. However, usually you are only
interested in a subset of the information in which case ets:tab2list/1 is expensive. If you only want to extract
one field from each record, for example, the age of every person, then:
DO
...
ets:select(Tab,[{ #person{idno='_',
name='_',
age='$1',
occupation = '_'},
[],
['$1']}]),
...
DO NOT
...
TabList = ets:tab2list(Tab),
lists:map(fun(X) -> X#person.age end, TabList),
...
If you are only interested in the age of all persons named "Bryan", then:
236 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.7 Tables and Databases
DO
...
ets:select(Tab,[{ #person{idno='_',
name="Bryan",
age='$1',
occupation = '_'},
[],
['$1']}]),
...
DO NOT
...
TabList = ets:tab2list(Tab),
lists:foldl(fun(X, Acc) -> case X#person.name of
"Bryan" ->
[X#person.age|Acc];
_ ->
Acc
end
end, [], TabList),
...
REALLY DO NOT
...
TabList = ets:tab2list(Tab),
BryanList = lists:filter(fun(X) -> X#person.name == "Bryan" end,
TabList),
lists:map(fun(X) -> X#person.age end, BryanList),
...
If you need all information stored in the Ets table about persons named "Bryan", then:
DO
...
ets:select(Tab, [{#person{idno='_',
name="Bryan",
age='_',
occupation = '_'}, [], ['$_']}]),
...
DO NOT
...
TabList = ets:tab2list(Tab),
lists:filter(fun(X) -> X#person.name == "Bryan" end, TabList),
...
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 237
7.7 Tables and Databases
Ordered_set Tables
If the data in the table is to be accessed so that the order of the keys in the table is significant, the table type
ordered_set can be used instead of the more usual set table type. An ordered_set is always traversed in
Erlang term order regarding the key field so that the return values from functions such as select, match_object,
and foldl are ordered by the key values. Traversing an ordered_set with the first and next operations also
returns the keys ordered.
Note:
An ordered_set only guarantees that objects are processed in key order. Results from functions such as
ets:select/2 appear in key order even if the key is not included in the result.
7.7.2 Ets-Specific
Using Keys of Ets Table
An Ets table is a single-key table (either a hash table or a tree ordered by the key) and is to be used as one. In other
words, use the key to look up things whenever possible. A lookup by a known key in a set Ets table is constant and
for an ordered_set Ets table it is O(logN). A key lookup is always preferable to a call where the whole table has
to be scanned. In the previous examples, the field idno is the key of the table and all lookups where only the name
is known result in a complete scan of the (possibly large) table for a matching result.
A simple solution would be to use the name field as the key instead of the idno field, but that would cause problems
if the names were not unique. A more general solution would be to create a second table with name as key and idno
as data, that is, to index (invert) the table regarding the name field. Clearly, the second table would have to be kept
consistent with the master table. Mnesia can do this for you, but a home brew index table can be very efficient compared
to the overhead involved in using Mnesia.
An index table for the table in the previous examples would have to be a bag (as keys would appear more than once)
and can have the following contents:
[#index_entry{name="Adam", idno=1},
#index_entry{name="Bryan", idno=2},
#index_entry{name="Bryan", idno=3},
#index_entry{name="Carl", idno=4}]
Given this index table, a lookup of the age fields for all persons named "Bryan" can be done as follows:
...
MatchingIDs = ets:lookup(IndexTable,"Bryan"),
lists:map(fun(#index_entry{idno = ID}) ->
[#person{age = Age}] = ets:lookup(PersonTable, ID),
Age
end,
MatchingIDs),
...
Notice that this code never uses ets:match/2 but instead uses the ets:lookup/2 call. The lists:map/2
call is only used to traverse the idnos matching the name "Bryan" in the table; thus the number of lookups in the
master table is minimized.
238 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.7 Tables and Databases
Keeping an index table introduces some overhead when inserting records in the table. The number of operations gained
from the table must therefore be compared against the number of operations inserting objects in the table. However,
notice that the gain is significant when the key can be used to lookup elements.
7.7.3 Mnesia-Specific
Secondary Index
If you frequently do a lookup on a field that is not the key of the table, you lose performance using "mnesia:select/
match_object" as this function traverses the whole table. You can create a secondary index instead and use
"mnesia:index_read" to get faster access, however this requires more memory.
Example
-record(person, {idno, name, age, occupation}).
...
{atomic, ok} =
mnesia:create_table(person, [{index,[#person.age]},
{attributes,
record_info(fields, person)}]),
{atomic, ok} = mnesia:add_table_index(person, age),
...
PersonsAge42 =
mnesia:dirty_index_read(person, 42, #person.age),
...
Transactions
Using transactions is a way to guarantee that the distributed Mnesia database remains consistent, even when many
different processes update it in parallel. However, if you have real-time requirements it is recommended to use dirty
operations instead of transactions. When using dirty operations, you lose the consistency guarantee; this is usually
solved by only letting one process update the table. Other processes must send update requests to that process.
Example
...
% Using transaction
Fun = fun() ->
[mnesia:read({Table, Key}),
mnesia:read({Table2, Key2})]
end,
{atomic, [Result1, Result2]}
...
= mnesia:transaction(Fun),
% Same thing using dirty operations
...
Result1 = mnesia:dirty_read({Table, Key}),
Result2 = mnesia:dirty_read({Table2, Key2}),
...
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 239
7.8 Processes
7.8 Processes
7.8.1 Creating an Erlang Process
An Erlang process is lightweight compared to threads and processes in operating systems.
A newly spawned Erlang process uses 309 words of memory in the non-SMP emulator without HiPE support. (SMP
support and HiPE support both add to this size.) The size can be found as follows:
Erlang (BEAM) emulator version 5.6 [async-threads:0] [kernel-poll:false]
Eshell V5.6 (abort with ^G)
1> Fun = fun() -> receive after infinity -> ok end end.
#Fun<...>
2> {_,Bytes} = process_info(spawn(Fun), memory).
{memory,1232}
3> Bytes div erlang:system_info(wordsize).
309
The size includes 233 words for the heap area (which includes the stack). The garbage collector increases the heap
as needed.
The main (outer) loop for a process must be tail-recursive. Otherwise, the stack grows until the process terminates.
DO NOT
loop() ->
receive
{sys, Msg} ->
handle_sys_msg(Msg),
loop();
{From, Msg} ->
Reply = handle_msg(Msg),
From ! Reply,
loop()
end,
io:format("Message is processed~n", []).
The call to io:format/2 will never be executed, but a return address will still be pushed to the stack each time
loop/0 is called recursively. The correct tail-recursive version of the function looks as follows:
DO
loop() ->
receive
{sys, Msg} ->
handle_sys_msg(Msg),
loop();
{From, Msg} ->
Reply = handle_msg(Msg),
From ! Reply,
loop()
end.
240 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.8 Processes
Initial Heap Size
The default initial heap size of 233 words is quite conservative to support Erlang systems with hundreds of thousands
or even millions of processes. The garbage collector grows and shrinks the heap as needed.
In a system that use comparatively few processes, performance might be improved by increasing the minimum heap
size using either the +h option for erl or on a process-per-process basis using the min_heap_size option for
spawn_opt/4.
The gain is twofold:
•
Although the garbage collector grows the heap, it grows it step-by-step, which is more costly than directly
establishing a larger heap when the process is spawned.
The garbage collector can also shrink the heap if it is much larger than the amount of data stored on it; setting
the minimum heap size prevents that.
•
Warning:
The emulator probably uses more memory, and because garbage collections occur less frequently, huge binaries
can be kept much longer.
In systems with many processes, computation tasks that run for a short time can be spawned off into a new process with
a higher minimum heap size. When the process is done, it sends the result of the computation to another process and
terminates. If the minimum heap size is calculated properly, the process might not have to do any garbage collections
at all. This optimization is not to be attempted without proper measurements.
7.8.2 Process Messages
All data in messages between Erlang processes is copied, except for refc binaries on the same Erlang node.
When a message is sent to a process on another Erlang node, it is first encoded to the Erlang External Format before
being sent through a TCP/IP socket. The receiving Erlang node decodes the message and distributes it to the correct
process.
Constant Pool
Constant Erlang terms (also called literals) are kept in constant pools; each loaded module has its own pool. The
following function does not build the tuple every time it is called (only to have it discarded the next time the garbage
collector was run), but the tuple is located in the module's constant pool:
DO
days_in_month(M) ->
element(M, {31,28,31,30,31,30,31,31,30,31,30,31}).
But if a constant is sent to another process (or stored in an Ets table), it is copied. The reason is that the runtime system
must be able to keep track of all references to constants to unload code containing constants properly. (When the code
is unloaded, the constants are copied to the heap of the processes that refer to them.) The copying of constants might
be eliminated in a future Erlang/OTP release.
Loss of Sharing
Shared subterms are not preserved in the following cases:
•
When a term is sent to another process
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 241
7.8 Processes
•
•
When a term is passed as the initial process arguments in the spawn call
When a term is stored in an Ets table
That is an optimization. Most applications do not send messages with shared subterms.
The following example shows how a shared subterm can be created:
kilo_byte() ->
kilo_byte(10, [42]).
kilo_byte(0, Acc) ->
Acc;
kilo_byte(N, Acc) ->
kilo_byte(N-1, [Acc|Acc]).
kilo_byte/1 creates a deep list. If list_to_binary/1 is called, the deep list can be converted to a binary
of 1024 bytes:
1> byte_size(list_to_binary(efficiency_guide:kilo_byte())).
1024
Using the erts_debug:size/1 BIF, it can be seen that the deep list only requires 22 words of heap space:
2> erts_debug:size(efficiency_guide:kilo_byte()).
22
Using the erts_debug:flat_size/1 BIF, the size of the deep list can be calculated if sharing is ignored. It
becomes the size of the list when it has been sent to another process or stored in an Ets table:
3> erts_debug:flat_size(efficiency_guide:kilo_byte()).
4094
It can be verified that sharing will be lost if the data is inserted into an Ets table:
4> T = ets:new(tab, []).
#Ref<0.1662103692.2407923716.214181>
5> ets:insert(T, {key,efficiency_guide:kilo_byte()}).
true
6> erts_debug:size(element(2, hd(ets:lookup(T, key)))).
4094
7> erts_debug:flat_size(element(2, hd(ets:lookup(T, key)))).
4094
When the data has passed through an Ets table, erts_debug:size/1 and erts_debug:flat_size/1 return
the same value. Sharing has been lost.
In a future Erlang/OTP release, it might be implemented a way to (optionally) preserve sharing.
242 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.9 Drivers
7.8.3 SMP Emulator
The SMP emulator (introduced in R11B) takes advantage of a multi-core or multi-CPU computer by running several
Erlang scheduler threads (typically, the same as the number of cores). Each scheduler thread schedules Erlang processes
in the same way as the Erlang scheduler in the non-SMP emulator.
To gain performance by using the SMP emulator, your application must have more than one runnable Erlang
process most of the time. Otherwise, the Erlang emulator can still only run one Erlang process at the time, but you
must still pay the overhead for locking. Although Erlang/OTP tries to reduce the locking overhead as much as possible,
it will never become exactly zero.
Benchmarks that appear to be concurrent are often sequential. The estone benchmark, for example, is entirely
sequential. So is the most common implementation of the "ring benchmark"; usually one process is active, while the
others wait in a receive statement.
7.9 Drivers
This section provides a brief overview on how to write efficient drivers.
It is assumed that you have a good understanding of drivers.
7.9.1 Drivers and Concurrency
The runtime system always takes a lock before running any code in a driver.
By default, that lock is at the driver level, that is, if several ports have been opened to the same driver, only code for
one port at the same time can be running.
A driver can be configured to have one lock for each port instead.
If a driver is used in a functional way (that is, holds no state, but only does some heavy calculation and returns a
result), several ports with registered names can be opened beforehand, and the port to be used can be chosen based
on the scheduler ID as follows:
-define(PORT_NAMES(),
{some_driver_01, some_driver_02,
some_driver_05, some_driver_06,
some_driver_09, some_driver_10,
some_driver_13, some_driver_14,
some_driver_03,
some_driver_07,
some_driver_11,
some_driver_15,
some_driver_04,
some_driver_08,
some_driver_12,
some_driver_16}).
client_port() ->
element(erlang:system_info(scheduler_id) rem tuple_size(?PORT_NAMES()) + 1,
?PORT_NAMES()).
As long as there are no more than 16 schedulers, there will never be any lock contention on the port lock for the driver.
7.9.2 Avoiding Copying Binaries When Calling a Driver
There are basically two ways to avoid copying a binary that is sent to a driver:
•
If the Data argument for port_control/3 is a binary, the driver will be passed a pointer to the contents of the
binary and the binary will not be copied. If the Data argument is an iolist (list of binaries and lists), all binaries
in the iolist will be copied.
Therefore, if you want to send both a pre-existing binary and some extra data to a driver without copying the
binary, you must call port_control/3 twice; once with the binary and once with the extra data. However,
that will only work if there is only one process communicating with the port (because otherwise another process
can call the driver in-between the calls).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 243
7.10 Advanced
•
Implement an outputv callback (instead of an output callback) in the driver. If a driver has an outputv
callback, refc binaries passed in an iolist in the Data argument for port_command/2 will be passed as references
to the driver.
7.9.3 Returning Small Binaries from a Driver
The runtime system can represent binaries up to 64 bytes as heap binaries. They are always copied when sent in
messages, but they require less memory if they are not sent to another process and garbage collection is cheaper.
If you know that the binaries you return are always small, you are advised to use driver API calls that do not require
a pre-allocated binary, for example, driver_output() or erl_drv_output_term(), using the ERL_DRV_BUF2BINARY
format, to allow the runtime to construct a heap binary.
7.9.4 Returning Large Binaries without Copying from a Driver
To avoid copying data when a large binary is sent or returned from the driver to an Erlang process, the driver must
first allocate the binary and then send it to an Erlang process in some way.
Use driver_alloc_binary() to allocate a binary.
There are several ways to send a binary created with driver_alloc_binary():
•
•
•
From the control callback, a binary can be returned if set_port_control_flags() has been called with the flag
value PORT_CONTROL_FLAG_BINARY.
A single binary can be sent with driver_output_binary().
Using erl_drv_output_term() or erl_drv_send_term(), a binary can be included in an Erlang term.
7.10 Advanced
7.10.1 Memory
A good start when programming efficiently is to know how much memory different data types and operations require.
It is implementation-dependent how much memory the Erlang data types and other items consume, but the following
table shows some figures for the erts-8.0 system in OTP 19.0.
The unit of measurement is memory words. There exists both a 32-bit and a 64-bit implementation. A word is therefore
4 bytes or 8 bytes, respectively.
Data Type
Memory Size
Small integer
1 word.
On 32-bit architectures: -134217729 < i < 134217728
(28 bits).
On 64-bit architectures: -576460752303423489 < i <
576460752303423488 (60 bits).
Large integer
3..N words.
Atom
1 word.
An atom refers into an atom table, which also consumes
memory. The atom text is stored once for each unique
atom in this table. The atom table is not garbagecollected.
Float
On 32-bit architectures: 4 words.
244 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.10 Advanced
On 64-bit architectures: 3 words.
Binary
3..6 words + data (can be shared).
List
1 word + 1 word per element + the size of each element.
String (is the same as a list of integers)
1 word + 2 words per character.
Tuple
2 words + the size of each element.
Small Map
5 words + the size of all keys and values.
Large Map (> 32 keys)
N x F words + the size of all keys and values.
N is the number of keys in the Map.
F is a sparsity factor that can vary between 1.6 and 1.8
due to the probabilistic nature of the internal HAMT
data structure.
Pid
1 word for a process identifier from the current local
node + 5 words for a process identifier from another
node.
A process identifier refers into a process table and a
node table, which also consumes memory.
Port
1 word for a port identifier from the current local node +
5 words for a port identifier from another node.
A port identifier refers into a port table and a node table,
which also consumes memory.
Reference
On 32-bit architectures: 5 words for a reference from
the current local node + 7 words for a reference from
another node.
On 64-bit architectures: 4 words for a reference from
the current local node + 6 words for a reference from
another node.
A reference refers into a node table, which also
consumes memory.
Fun
9..13 words + the size of environment.
A fun refers into a fun table, which also consumes
memory.
Ets table
Initially 768 words + the size of each element (6 words
+ the size of Erlang data). The table grows when
necessary.
Erlang process
338 words when spawned, including a heap of 233
words.
Table 10.1: Memory Size of Different Data Types
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 245
7.10 Advanced
7.10.2 System Limits
The Erlang language specification puts no limits on the number of processes, length of atoms, and so on. However,
for performance and memory saving reasons, there will always be limits in a practical implementation of the Erlang
language and execution environment.
Processes
The maximum number of simultaneously alive Erlang
processes is by default 32,768. This limit can be
configured at startup. For more information, see the
+P command-line flag in the erl(1) manual page in
ERTS.
Known nodes
A remote node Y must be known to node X if there
exists any pids, ports, references, or funs (Erlang data
types) from Y on X, or if X and Y are connected. The
maximum number of remote nodes simultaneously/ever
known to a node is limited by the maximum number of
atoms available for node names. All data concerning
remote nodes, except for the node name atom, are
garbage-collected.
Connected nodes
The maximum number of simultaneously connected
nodes is limited by either the maximum number of
simultaneously known remote nodes, the maximum
number of (Erlang) ports available, or the maximum
number of sockets available.
Characters in an atom
255.
Atoms
By default, the maximum number of atoms is 1,048,576.
This limit can be raised or lowered using the +t option.
Ets tables
Default is 1400. It can be changed with the environment
variable ERL_MAX_ETS_TABLES.
Elements in a tuple
The maximum number of elements in a tuple is
67,108,863 (26-bit unsigned integer). Clearly, other
factors such as the available memory can make it
difficult to create a tuple of that size.
Size of binary
In the 32-bit implementation of Erlang, 536,870,911
bytes is the largest binary that can be constructed
or matched using the bit syntax. In the 64bit implementation, the maximum size is
2,305,843,009,213,693,951 bytes. If the limit
is exceeded, bit syntax construction fails with a
system_limit exception, while any attempt to
match a binary that is too large fails. This limit is
enforced starting in R11B-4.
In earlier Erlang/OTP releases, operations on too large
binaries in general either fail or give incorrect results.
In future releases, other operations that create binaries
246 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.10 Advanced
(such as list_to_binary/1) will probably also
enforce the same limit.
Total amount of data allocated by an Erlang node
The Erlang runtime system can use the complete 32-bit
(or 64-bit) address space, but the operating system often
limits a single process to use less than that.
Length of a node name
An Erlang node name has the form host@shortname
or host@longname. The node name is used as an atom
within the system, so the maximum size of 255 holds
also for the node name.
Open ports
The maximum number of simultaneously open Erlang
ports is often by default 16,384. This limit can be
configured at startup. For more information, see the
+Q command-line flag in the erl(1) manual page in
ERTS.
Open files and sockets
The maximum number of simultaneously open files
and sockets depends on the maximum number of Erlang
ports available, as well as on operating system-specific
settings and limits.
Number of arguments to a function or fun
255
Unique References on a Runtime System Instance
Each scheduler thread has its own set of references,
and all other threads have a shared set of references.
Each set of references consist of 2## - 1 unique
references. That is the total amount of unique references
that can be produced on a runtime system instance
is (NoSchedulers + 1) * (2## - 1). If
a scheduler thread create a new reference each nano
second, references will at earliest be reused after more
than 584 years. That is, for the foreseeable future they
are unique enough.
Unique Integers on a Runtime System Instance
There are two types of unique integers both created
using the erlang:unique_integer() BIF. Unique integers
created:
with the monotonic modifier
consist of a set of 2## - 1 unique integers.
without the monotonic modifier
consist of a set of 2## - 1 unique integers
per scheduler thread and a set of 2## - 1
unique integers shared by other threads. That is
the total amount of unique integers without the
monotonic modifier is (NoSchedulers +
1) * (2## - 1)
If a unique integer is created each nano second, unique
integers will at earliest be reused after more than 584
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 247
7.11 Profiling
years. That is, for the foreseeable future they are unique
enough.
Table 10.2: System Limits
7.11 Profiling
7.11.1 Do Not Guess About Performance - Profile
Even experienced software developers often guess wrong about where the performance bottlenecks are in their
programs. Therefore, profile your program to see where the performance bottlenecks are and concentrate on optimizing
them.
Erlang/OTP contains several tools to help finding bottlenecks:
•
•
•
•
fprof provides the most detailed information about where the program time is spent, but it significantly slows
down the program it profiles.
eprof provides time information of each function used in the program. No call graph is produced, but eprof
has considerable less impact on the program it profiles.
If the program is too large to be profiled by fprof or eprof, the cover and cprof tools can be used to locate
code parts that are to be more thoroughly profiled using fprof or eprof.
cover provides execution counts per line per process, with less overhead than fprof. Execution counts can,
with some caution, be used to locate potential performance bottlenecks.
cprof is the most lightweight tool, but it only provides execution counts on a function basis (for all processes,
not per process).
The tools are further described in Tools.
7.11.2 Large Systems
For a large system, it can be interesting to run profiling on a simulated and limited scenario to start with. But bottlenecks
have a tendency to appear or cause problems only when many things are going on at the same time, and when many
nodes are involved. Therefore, it is also desirable to run profiling in a system test plant on a real target system.
For a large system, you do not want to run the profiling tools on the whole system. Instead you want to concentrate
on central processes and modules, which contribute for a big part of the execution.
7.11.3 What to Look For
When analyzing the result file from the profiling activity, look for functions that are called many times and have a
long "own" execution time (time excluding calls to other functions). Functions that are called a lot of times can also
be interesting, as even small things can add up to quite a bit if repeated often. Also ask yourself what you can do to
reduce this time. The following are appropriate types of questions to ask yourself:
•
•
•
•
•
•
Is it possible to reduce the number of times the function is called?
Can any test be run less often if the order of tests is changed?
Can any redundant tests be removed?
Does any calculated expression give the same result each time?
Are there other ways to do this that are equivalent and more efficient?
Can another internal data representation be used to make things more efficient?
These questions are not always trivial to answer. Some benchmarks might be needed to back up your theory and to
avoid making things slower if your theory is wrong. For details, see Benchmarking.
248 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
7.11 Profiling
7.11.4 Tools
fprof
fprof measures the execution time for each function, both own time, that is, how much time a function has used
for its own execution, and accumulated time, that is, including called functions. The values are displayed per process.
You also get to know how many times each function has been called.
fprof is based on trace to file to minimize runtime performance impact. Using fprof is just a matter of calling a
few library functions, see the fprof manual page in Tools .fprof was introduced in R8.
eprof
eprof is based on the Erlang trace_info BIFs. eprof shows how much time has been used by each process,
and in which function calls this time has been spent. Time is shown as percentage of total time and absolute time. For
more information, see the eprof manual page in Tools.
cover
The primary use of cover is coverage analysis to verify test cases, making sure that all relevant code is covered.
cover counts how many times each executable line of code is executed when a program is run, on a per module basis.
Clearly, this information can be used to determine what code is run very frequently and can therefore be subject for
optimization. Using cover is just a matter of calling a few library functions, see the cover manual page in Tools.
cprof
cprof is something in between fprof and cover regarding features. It counts how many times each function is
called when the program is run, on a per module basis. cprof has a low performance degradation effect (compared
with fprof) and does not need to recompile any modules to profile (compared with cover). For more information,
see the cprof manual page in Tools.
Tool Summary
Effects on
Program
Execution
Time
Records
Number of
Calls
Records
Execution
Time
Records
Called by
Records
Garbage
Collection
Significant
slowdown
Yes
Total and
own
Yes
Yes
Only total
No
No
Tool
Results
Size of
Result
fprof
Per process
to screen/
file
Large
eprof
Per process/
function to Medium
screen/file
Small
slowdown
Yes
cover
Per module
to screen/
file
Small
Moderate
slowdown
Yes, per line No
No
No
cprof
Per module
to caller
Small
Small
slowdown
Yes
No
No
No
Table 11.1: Tool Summary
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 249
7.12 Retired Myths
7.11.5 Benchmarking
The main purpose of benchmarking is to find out which implementation of a given algorithm or function is the fastest.
Benchmarking is far from an exact science. Today's operating systems generally run background tasks that are difficult
to turn off. Caches and multiple CPU cores does not facilitate benchmarking. It would be best to run UNIX computers
in single-user mode when benchmarking, but that is inconvenient to say the least for casual testing.
Benchmarks can measure wall-clock time or CPU time.
•
•
timer:tc/3 measures wall-clock time. The advantage with wall-clock time is that I/O, swapping, and other
activities in the operating system kernel are included in the measurements. The disadvantage is that the
measurements vary a lot. Usually it is best to run the benchmark several times and note the shortest time, which
is to be the minimum time that is possible to achieve under the best of circumstances.
statistics/1 with argument runtime measures CPU time spent in the Erlang virtual machine. The advantage
with CPU time is that the results are more consistent from run to run. The disadvantage is that the time spent
in the operating system kernel (such as swapping and I/O) is not included. Therefore, measuring CPU time is
misleading if any I/O (file or socket) is involved.
It is probably a good idea to do both wall-clock measurements and CPU time measurements.
Some final advice:
•
•
•
The granularity of both measurement types can be high. Therefore, ensure that each individual measurement
lasts for at least several seconds.
To make the test fair, each new test run is to run in its own, newly created Erlang process. Otherwise, if all tests
run in the same process, the later tests start out with larger heap sizes and therefore probably do fewer garbage
collections. Also consider restarting the Erlang emulator between each test.
Do not assume that the fastest implementation of a given algorithm on computer architecture X is also the
fastest on computer architecture Y.
7.12 Retired Myths
We belive that the truth finally has caught with the following, retired myths.
7.12.1 Myth: Funs are Slow
Funs used to be very slow, slower than apply/3. Originally, funs were implemented using nothing more than
compiler trickery, ordinary tuples, apply/3, and a great deal of ingenuity.
But that is history. Funs was given its own data type in R6B and was further optimized in R7B. Now the cost for a fun
call falls roughly between the cost for a call to a local function and apply/3.
7.12.2 Myth: List Comprehensions are Slow
List comprehensions used to be implemented using funs, and in the old days funs were indeed slow.
Nowadays, the compiler rewrites list comprehensions into an ordinary recursive function. Using a tail-recursive
function with a reverse at the end would be still faster. Or would it? That leads us to the myth that tail-recursive
functions are faster than body-recursive functions.
250 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.1 Introduction
8 Interoperability Tutorial
8.1 Introduction
This section informs on interoperability, that is, information exchange, between Erlang and other programming
languages. The included examples mainly treat interoperability between Erlang and C.
8.1.1 Purpose
The purpose of this tutorial is to describe different interoperability mechanisms that can be used when integrating a
program written in Erlang with a program written in another programming language, from the Erlang programmer's
perspective.
8.1.2 Prerequisites
It is assumed that you are a skilled Erlang programmer, familiar with concepts such as Erlang data types, processes,
messages, and error handling.
To illustrate the interoperability principles, C programs running in a UNIX environment have been used. It is assumed
that you have enough knowledge to apply these principles to the relevant programming languages and platforms.
Note:
For readability, the example code is kept as simple as possible. For example, it does not include error handling,
which might be vital in a real-life system.
8.2 Overview
8.2.1 Built-In Mechanisms
Two interoperability mechanisms are built into the Erlang runtime system, distributed Erlang and ports. A variation
of ports is linked-in drivers.
Distributed Erlang
An Erlang runtime system is made a distributed Erlang node by giving it a name. A distributed Erlang node can connect
to, and monitor, other nodes. It can also spawn processes at other nodes. Message passing and error handling between
processes at different nodes are transparent. A number of useful STDLIB modules are available in a distributed Erlang
system. For example, global, which provides global name registration. The distribution mechanism is implemented
using TCP/IP sockets.
When to use: Distributed Erlang is primarily used for Erlang-Erlang communication. It can also be used for
communication between Erlang and C, if the C program is implemented as a C node, see C and Java Libraries.
Where to read more: Distributed Erlang and some distributed programming techniques are described in the Erlang
book.
For more information, see Distributed Programming.
Relevant manual pages are the following:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 251
8.2 Overview
•
•
•
•
•
•
•
erlang manual page in ERTS (describes the BIFs)
global manual page in Kernel
net_adm manual page in Kernel
pg2 manual page in Kernel
rpc manual page in Kernel
pool manual page in STDLIB
slave manual page in STDLIB
Ports and Linked-In Drivers
Ports provide the basic mechanism for communication with the external world, from Erlang's point of view. The ports
provide a byte-oriented interface to an external program. When a port is created, Erlang can communicate with it by
sending and receiving lists of bytes (not Erlang terms). This means that the programmer might have to invent a suitable
encoding and decoding scheme.
The implementation of the port mechanism depends on the platform. For UNIX, pipes are used and the external
program is assumed to read from standard input and write to standard output. The external program can be written
in any programming language as long as it can handle the interprocess communication mechanism with which the
port is implemented.
The external program resides in another OS process than the Erlang runtime system. In some cases this is not
acceptable. Consider, for example, drivers with very hard time requirements. It is therefore possible to write a program
in C according to certain principles, and dynamically link it to the Erlang runtime system. This is called a linkedin driver.
When to use: Ports can be used for all kinds of interoperability situations where the Erlang program and the other
program runs on the same machine. Programming is fairly straight-forward.
Linked-in drivers involves writing certain call-back functions in C. This requires very good skills as the code is linked
to the Erlang runtime system.
Warning:
A faulty linked-in driver causes the entire Erlang runtime system to leak memory, hang, or crash.
Where to read more: Ports are described in section "Miscellaneous Items" of the Erlang book. Linked-in drivers are
described in Appendix E.
The BIF open_port/2 is documented in the erlang manual page in ERTS.
For linked-in drivers, the programmer needs to read the erl_ddll manual page in Kernel.
Examples: Port example in Ports.
8.2.2 C and Java Libraries
Erl_Interface
The program at the other side of a port is often a C program. To help the C programmer, the Erl_Interface library has
been developed, including the following five parts:
•
•
•
erl_marshal, erl_eterm, erl_format, and erl_malloc: Handling of the Erlang external term
format
erl_connect: Communication with distributed Erlang, see C nodes below
erl_error: Error print routines
252 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.2 Overview
•
•
erl_global: Access globally registered names
Registry: Store and backup of key-value pairs
The Erlang external term format is a representation of an Erlang term as a sequence of bytes, that is, a binary.
Conversion between the two representations is done using the following BIFs:
Binary = term_to_binary(Term)
Term = binary_to_term(Binary)
A port can be set to use binaries instead of lists of bytes. It is then not necessary to invent any encoding/decoding
scheme. Erl_Interface functions are used for unpacking the binary and convert it into a struct similar to an Erlang term.
Such a struct can be manipulated in different ways, be converted to the Erlang external format, and sent to Erlang.
When to use: In C code, in conjunction with Erlang binaries.
Where to read more: See the Erlang Interface User's Guide, Command Reference, and Library Reference. In Erlang/
OTP R5B, and earlier versions, the information is part of the Kernel application.
Examples: Erl_Interface example in Erl_Interface.
C Nodes
A C program that uses the Erl_Interface functions for setting up a connection to, and communicating with, a distributed
Erlang node is called a C node, or a hidden node. The main advantage with a C node is that the communication from
the Erlang programmer's perspective is extremely easy, as the C program behaves as a distributed Erlang node.
When to use: C nodes can typically be used on device processors (as opposed to control processors) where C is a
better choice than Erlang due to memory limitations or application characteristics, or both.
Where to read more: See the erl_connect part of the Erl_Interface documentation. The programmer also needs
to be familiar with TCP/IP sockets, see Sockets in Standard Protocols and Distributed Erlang in Built-In Mechanisms.
Example: C node example in C Nodes.
Jinterface
In Erlang/OTP R6B, a library similar to Erl_Interface for Java was added called jinterface. It provides a tool for Java
programs to communicate with Erlang nodes.
8.2.3 Standard Protocols
Sometimes communication between an Erlang program and another program using a standard protocol is desirable.
Erlang/OTP currently supports TCP/IP and UDP sockets: as follows:
•
•
•
SNMP
HTTP
IIOP (CORBA)
Using one of the latter three requires good knowledge about the protocol and is not covered by this tutorial. See the
SNMP, Inets, and Orber applications, respectively.
Sockets
Simply put, connection-oriented socket communication (TCP/IP) consists of an initiator socket ("server") started at a
certain host with a certain port number. A connector socket ("client"), which is aware of the initiator host name and
port number, can connect to it and data can be sent between them.
Connection-less socket communication (UDP) consists of an initiator socket at a certain host with a certain port number
and a connector socket sending data to it.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 253
8.3 Problem Example
For a detailed description of the socket concept, refer to a suitable book about network programming. A suggestion
is UNIX Network Programming, Volume 1: Networking APIs - Sockets and XTI by W. Richard Stevens, ISBN:
013490012X.
In Erlang/OTP, access to TCP/IP and UDP sockets is provided by the modules gen_tcp and gen_udp in Kernel.
Both are easy to use and do not require detailed knowledge about the socket concept.
When to use: For programs running on the same or on another machine than the Erlang program.
Where to read more: See the gen_tcp and the gen_udp manual pages in Kernel.
8.2.4 IC
IC (Erlang IDL Compiler) is an interface generator that, given an IDL interface specification, automatically generates
stub code in Erlang, C, or Java. See the IC User's Guide and IC Reference Manual.
For details, see the ic manual page in IC.
8.2.5 Old Applications
Two old applications are of interest regarding interoperability. Both have been replaced by IC and are mentioned here
for reference only:
•
IG - Removed from Erlang/OTP R6B.
•
IG (Interface Generator) automatically generated code for port or socket communication between an Erlang
program and a C program, given a C header file with certain keywords.
Jive - Removed from Erlang/OTP R7B.
Jive provided a simple interface between an Erlang program and a Java program.
8.3 Problem Example
8.3.1 Description
A common interoperability situation is when you want to incorporate a piece of code, solving a complex problem,
in your Erlang program. Suppose for example, that you have the following C functions that you would like to call
from Erlang:
/* complex.c */
int foo(int x) {
return x+1;
}
int bar(int y) {
return y*2;
}
The functions are deliberately kept as simple as possible, for readability reasons.
From an Erlang perspective, it is preferable to be able to call foo and bar without having to bother about that they
are C functions:
% Erlang code
...
254 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.4 Ports
Res = complex:foo(X),
...
Here, the communication with C is hidden in the implementation of complex.erl. In the following sections, it is
shown how this module can be implemented using the different interoperability mechanisms.
8.4 Ports
This section outlines an example of how to solve the example problem in the previous section by using a port.
The scenario is illustrated in the following figure:
Figure 4.1: Port Communication
8.4.1 Erlang Program
All communication between Erlang and C must be established by creating the port. The Erlang process that creates
a port is said to be the connected process of the port. All communication to and from the port must go through the
connected process. If the connected process terminates, the port also terminates (and the external program, if it is
written properly).
The port is created using the BIF open_port/2 with {spawn,ExtPrg} as the first argument. The string ExtPrg
is the name of the external program, including any command line arguments. The second argument is a list of options, in
this case only {packet,2}. This option says that a 2 byte length indicator is to be used to simplify the communication
between C and Erlang. The Erlang port automatically adds the length indicator, but this must be done explicitly in
the external C program.
The process is also set to trap exits, which enables detection of failure of the external program:
-module(complex1).
-export([start/1, init/1]).
start(ExtPrg) ->
spawn(?MODULE, init, [ExtPrg]).
init(ExtPrg) ->
register(complex, self()),
process_flag(trap_exit, true),
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 255
8.4 Ports
Port = open_port({spawn, ExtPrg}, [{packet, 2}]),
loop(Port).
Now complex1:foo/1 and complex1:bar/1 can be implemented. Both send a message to the complex
process and receive the following replies:
foo(X) ->
call_port({foo, X}).
bar(Y) ->
call_port({bar, Y}).
call_port(Msg) ->
complex ! {call, self(), Msg},
receive
{complex, Result} ->
Result
end.
The complex process does the following:
•
•
•
•
•
Encodes the message into a sequence of bytes.
Sends it to the port.
Waits for a reply.
Decodes the reply.
Sends it back to the caller:
loop(Port) ->
receive
{call, Caller, Msg} ->
Port ! {self(), {command, encode(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, decode(Data)}
end,
loop(Port)
end.
Assuming that both the arguments and the results from the C functions are less than 256, a simple encoding/decoding
scheme is employed. In this scheme, foo is represented by byte 1, bar is represented by 2, and the argument/result
is represented by a single byte as well:
encode({foo, X}) -> [1, X];
encode({bar, Y}) -> [2, Y].
decode([Int]) -> Int.
The resulting Erlang program, including functionality for stopping the port and detecting port failures, is as follows:
-module(complex1).
-export([start/1, stop/0, init/1]).
-export([foo/1, bar/1]).
256 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.4 Ports
start(ExtPrg) ->
spawn(?MODULE, init, [ExtPrg]).
stop() ->
complex ! stop.
foo(X) ->
call_port({foo, X}).
bar(Y) ->
call_port({bar, Y}).
call_port(Msg) ->
complex ! {call, self(), Msg},
receive
{complex, Result} ->
Result
end.
init(ExtPrg) ->
register(complex, self()),
process_flag(trap_exit, true),
Port = open_port({spawn, ExtPrg}, [{packet, 2}]),
loop(Port).
loop(Port) ->
receive
{call, Caller, Msg} ->
Port ! {self(), {command, encode(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, decode(Data)}
end,
loop(Port);
stop ->
Port ! {self(), close},
receive
{Port, closed} ->
exit(normal)
end;
{'EXIT', Port, Reason} ->
exit(port_terminated)
end.
encode({foo, X}) -> [1, X];
encode({bar, Y}) -> [2, Y].
decode([Int]) -> Int.
8.4.2 C Program
On the C side, it is necessary to write functions for receiving and sending data with 2 byte length indicators from/to
Erlang. By default, the C program is to read from standard input (file descriptor 0) and write to standard output (file
descriptor 1). Examples of such functions, read_cmd/1 and write_cmd/2, follows:
/* erl_comm.c */
typedef unsigned char byte;
read_cmd(byte *buf)
{
int len;
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 257
8.4 Ports
}
if (read_exact(buf, 2) != 2)
return(-1);
len = (buf[0] << 8) | buf[1];
return read_exact(buf, len);
write_cmd(byte *buf, int len)
{
byte li;
li = (len >> 8) & 0xff;
write_exact(&li, 1);
li = len & 0xff;
write_exact(&li, 1);
}
return write_exact(buf, len);
read_exact(byte *buf, int len)
{
int i, got=0;
do {
if ((i = read(0, buf+got, len-got)) <= 0)
return(i);
got += i;
} while (got<len);
}
return(len);
write_exact(byte *buf, int len)
{
int i, wrote = 0;
do {
if ((i = write(1, buf+wrote, len-wrote)) <= 0)
return (i);
wrote += i;
} while (wrote<len);
}
return (len);
Notice that stdin and stdout are for buffered input/output and must not be used for the communication with Erlang.
In the main function, the C program is to listen for a message from Erlang and, according to the selected encoding/
decoding scheme, use the first byte to determine which function to call and the second byte as argument to the function.
The result of calling the function is then to be sent back to Erlang:
/* port.c */
typedef unsigned char byte;
int main() {
int fn, arg, res;
byte buf[100];
while (read_cmd(buf) > 0) {
258 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.5 Erl_Interface
fn = buf[0];
arg = buf[1];
if (fn == 1) {
res = foo(arg);
} else if (fn == 2) {
res = bar(arg);
}
}
}
buf[0] = res;
write_cmd(buf, 1);
Notice that the C program is in a while-loop, checking for the return value of read_cmd/1. This is because the
C program must detect when the port closes and terminates.
8.4.3 Running the Example
Step 1. Compile the C code:
unix> gcc -o extprg complex.c erl_comm.c port.c
Step 2. Start Erlang and compile the Erlang code:
unix> erl
Erlang (BEAM) emulator version 4.9.1.2
Eshell V4.9.1.2 (abort with ^G)
1> c(complex1).
{ok,complex1}
Step 3. Run the example:
2> complex1:start("extprg").
<0.34.0>
3> complex1:foo(3).
4
4> complex1:bar(5).
10
5> complex1:stop().
stop
8.5 Erl_Interface
This section outlines an example of how to solve the example problem in Problem Example by using a port and
Erl_Interface. It is necessary to read the port example in Ports before reading this section.
8.5.1 Erlang Program
The following example shows an Erlang program communicating with a C program over a plain port with home made
encoding:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 259
8.5 Erl_Interface
-module(complex1).
-export([start/1, stop/0, init/1]).
-export([foo/1, bar/1]).
start(ExtPrg) ->
spawn(?MODULE, init, [ExtPrg]).
stop() ->
complex ! stop.
foo(X) ->
call_port({foo, X}).
bar(Y) ->
call_port({bar, Y}).
call_port(Msg) ->
complex ! {call, self(), Msg},
receive
{complex, Result} ->
Result
end.
init(ExtPrg) ->
register(complex, self()),
process_flag(trap_exit, true),
Port = open_port({spawn, ExtPrg}, [{packet, 2}]),
loop(Port).
loop(Port) ->
receive
{call, Caller, Msg} ->
Port ! {self(), {command, encode(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, decode(Data)}
end,
loop(Port);
stop ->
Port ! {self(), close},
receive
{Port, closed} ->
exit(normal)
end;
{'EXIT', Port, Reason} ->
exit(port_terminated)
end.
encode({foo, X}) -> [1, X];
encode({bar, Y}) -> [2, Y].
decode([Int]) -> Int.
There are two differences when using Erl_Interface on the C side compared to the example in Ports, using only the
plain port:
•
•
As Erl_Interface operates on the Erlang external term format, the port must be set to use binaries.
Instead of inventing an encoding/decoding scheme, the term_to_binary/1 and binary_to_term/1
BIFs are to be used.
That is:
260 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.5 Erl_Interface
open_port({spawn, ExtPrg}, [{packet, 2}])
is replaced with:
open_port({spawn, ExtPrg}, [{packet, 2}, binary])
And:
Port ! {self(), {command, encode(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, decode(Data)}
end
is replaced with:
Port ! {self(), {command, term_to_binary(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, binary_to_term(Data)}
end
The resulting Erlang program is as follows:
-module(complex2).
-export([start/1, stop/0, init/1]).
-export([foo/1, bar/1]).
start(ExtPrg) ->
spawn(?MODULE, init, [ExtPrg]).
stop() ->
complex ! stop.
foo(X) ->
call_port({foo, X}).
bar(Y) ->
call_port({bar, Y}).
call_port(Msg) ->
complex ! {call, self(), Msg},
receive
{complex, Result} ->
Result
end.
init(ExtPrg) ->
register(complex, self()),
process_flag(trap_exit, true),
Port = open_port({spawn, ExtPrg}, [{packet, 2}, binary]),
loop(Port).
loop(Port) ->
receive
{call, Caller, Msg} ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 261
8.5 Erl_Interface
Port ! {self(), {command, term_to_binary(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, binary_to_term(Data)}
end,
loop(Port);
stop ->
Port ! {self(), close},
receive
{Port, closed} ->
exit(normal)
end;
{'EXIT', Port, Reason} ->
exit(port_terminated)
end.
Notice that calling complex2:foo/1 and complex2:bar/1 results in the tuple {foo,X} or {bar,Y} being
sent to the complex process, which codes them as binaries and sends them to the port. This means that the C program
must be able to handle these two tuples.
8.5.2 C Program
The following example shows a C program communicating with an Erlang program over a plain port with home made
encoding:
/* port.c */
typedef unsigned char byte;
int main() {
int fn, arg, res;
byte buf[100];
while (read_cmd(buf) > 0) {
fn = buf[0];
arg = buf[1];
if (fn == 1) {
res = foo(arg);
} else if (fn == 2) {
res = bar(arg);
}
}
}
buf[0] = res;
write_cmd(buf, 1);
Compared to the C program in Ports, using only the plain port, the while-loop must be rewritten. Messages coming
from the port is on the Erlang external term format. They must be converted into an ETERM struct, which is a C struct
similar to an Erlang term. The result of calling foo() or bar() must be converted to the Erlang external term format
before being sent back to the port. But before calling any other Erl_Interface function, the memory handling must
be initiated:
erl_init(NULL, 0);
262 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.5 Erl_Interface
The following functions, read_cmd() and write_cmd(), from the erl_comm.c example in Ports can still be
used for reading from and writing to the port:
/* erl_comm.c */
typedef unsigned char byte;
read_cmd(byte *buf)
{
int len;
}
if (read_exact(buf, 2) != 2)
return(-1);
len = (buf[0] << 8) | buf[1];
return read_exact(buf, len);
write_cmd(byte *buf, int len)
{
byte li;
li = (len >> 8) & 0xff;
write_exact(&li, 1);
li = len & 0xff;
write_exact(&li, 1);
}
return write_exact(buf, len);
read_exact(byte *buf, int len)
{
int i, got=0;
do {
if ((i = read(0, buf+got, len-got)) <= 0)
return(i);
got += i;
} while (got<len);
}
return(len);
write_exact(byte *buf, int len)
{
int i, wrote = 0;
do {
if ((i = write(1, buf+wrote, len-wrote)) <= 0)
return (i);
wrote += i;
} while (wrote<len);
}
return (len);
The function erl_decode() from erl_marshal converts the binary into an ETERM struct:
int main() {
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 263
8.5 Erl_Interface
ETERM *tuplep;
while (read_cmd(buf) > 0) {
tuplep = erl_decode(buf);
Here, tuplep points to an ETERM struct representing a tuple with two elements; the function name (atom) and the
argument (integer). Using the function erl_element() from erl_eterm, these elements can be extracted, but
they must also be declared as pointers to an ETERM struct:
fnp = erl_element(1, tuplep);
argp = erl_element(2, tuplep);
The macros ERL_ATOM_PTR and ERL_INT_VALUE from erl_eterm can be used to obtain the actual values of
the atom and the integer. The atom value is represented as a string. By comparing this value with the strings "foo"
and "bar", it can be decided which function to call:
if (strncmp(ERL_ATOM_PTR(fnp), "foo", 3) == 0) {
res = foo(ERL_INT_VALUE(argp));
} else if (strncmp(ERL_ATOM_PTR(fnp), "bar", 3) == 0) {
res = bar(ERL_INT_VALUE(argp));
}
Now an ETERM struct that represents the integer result can be constructed using the function erl_mk_int() from
erl_eterm. The function erl_format() from the module erl_format can also be used:
intp = erl_mk_int(res);
The resulting ETERM struct is converted into the Erlang external term format using the function erl_encode()
from erl_marshal and sent to Erlang using write_cmd():
erl_encode(intp, buf);
write_cmd(buf, erl_eterm_len(intp));
Finally, the memory allocated by the ETERM creating functions must be freed:
erl_free_compound(tuplep);
erl_free_term(fnp);
erl_free_term(argp);
erl_free_term(intp);
The resulting C program is as follows:
/* ei.c */
#include "erl_interface.h"
#include "ei.h"
264 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.5 Erl_Interface
typedef unsigned char byte;
int main() {
ETERM *tuplep, *intp;
ETERM *fnp, *argp;
int res;
byte buf[100];
long allocated, freed;
erl_init(NULL, 0);
while (read_cmd(buf) > 0) {
tuplep = erl_decode(buf);
fnp = erl_element(1, tuplep);
argp = erl_element(2, tuplep);
if (strncmp(ERL_ATOM_PTR(fnp), "foo", 3) == 0) {
res = foo(ERL_INT_VALUE(argp));
} else if (strncmp(ERL_ATOM_PTR(fnp), "bar", 3) == 0) {
res = bar(ERL_INT_VALUE(argp));
}
intp = erl_mk_int(res);
erl_encode(intp, buf);
write_cmd(buf, erl_term_len(intp));
}
}
erl_free_compound(tuplep);
erl_free_term(fnp);
erl_free_term(argp);
erl_free_term(intp);
8.5.3 Running the Example
Step 1. Compile the C code. This provides the paths to the include files erl_interface.h and ei.h, and also
to the libraries erl_interface and ei:
unix> gcc -o extprg -I/usr/local/otp/lib/erl_interface-3.9.2/include \\
-L/usr/local/otp/lib/erl_interface-3.9.2/lib \\
complex.c erl_comm.c ei.c -lerl_interface -lei -lpthread
In Erlang/OTP R5B and later versions of OTP, the include and lib directories are situated under OTPROOT/lib/
erl_interface-VSN, where OTPROOT is the root directory of the OTP installation (/usr/local/otp in the
recent example) and VSN is the version of the Erl_interface application (3.2.1 in the recent example).
In R4B and earlier versions of OTP, include and lib are situated under OTPROOT/usr.
Step 2. Start Erlang and compile the Erlang code:
unix> erl
Erlang (BEAM) emulator version 4.9.1.2
Eshell V4.9.1.2 (abort with ^G)
1> c(complex2).
{ok,complex2}
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 265
8.6 Port Drivers
Step 3. Run the example:
2> complex2:start("./extprg").
<0.34.0>
3> complex2:foo(3).
4
4> complex2:bar(5).
10
5> complex2:bar(352).
704
6> complex2:stop().
stop
8.6 Port Drivers
This section outlines an example of how to solve the example problem in Problem Example by using a linked-in port
driver.
A port driver is a linked-in driver that is accessible as a port from an Erlang program. It is a shared library (SO in
UNIX, DLL in Windows), with special entry points. The Erlang runtime system calls these entry points when the
driver is started and when data is sent to the port. The port driver can also send data to Erlang.
As a port driver is dynamically linked into the emulator process, this is the fastest way of calling C-code from Erlang.
Calling functions in the port driver requires no context switches. But it is also the least safe way, because a crash in
the port driver brings the emulator down too.
The scenario is illustrated in the following figure:
Figure 6.1: Port Driver Communication
266 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.6 Port Drivers
8.6.1 Erlang Program
Like a port program, the port communicates with an Erlang process. All communication goes through one Erlang
process that is the connected process of the port driver. Terminating this process closes the port driver.
Before the port is created, the driver must be loaded. This is done with the function erl_dll:load_driver/1,
with the name of the shared library as argument.
The port is then created using the BIF open_port/2, with the tuple {spawn, DriverName} as the first
argument. The string SharedLib is the name of the port driver. The second argument is a list of options, none in
this case:
-module(complex5).
-export([start/1, init/1]).
start(SharedLib) ->
case erl_ddll:load_driver(".", SharedLib) of
ok -> ok;
{error, already_loaded} -> ok;
_ -> exit({error, could_not_load_driver})
end,
spawn(?MODULE, init, [SharedLib]).
init(SharedLib) ->
register(complex, self()),
Port = open_port({spawn, SharedLib}, []),
loop(Port).
Now complex5:foo/1 and complex5:bar/1 can be implemented. Both send a message to the complex
process and receive the following reply:
foo(X) ->
call_port({foo, X}).
bar(Y) ->
call_port({bar, Y}).
call_port(Msg) ->
complex ! {call, self(), Msg},
receive
{complex, Result} ->
Result
end.
The complex process performs the following:
•
•
•
•
•
Encodes the message into a sequence of bytes.
Sends it to the port.
Waits for a reply.
Decodes the reply.
Sends it back to the caller:
loop(Port) ->
receive
{call, Caller, Msg} ->
Port ! {self(), {command, encode(Msg)}},
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 267
8.6 Port Drivers
end.
receive
{Port, {data, Data}} ->
Caller ! {complex, decode(Data)}
end,
loop(Port)
Assuming that both the arguments and the results from the C functions are less than 256, a simple encoding/decoding
scheme is employed. In this scheme, foo is represented by byte 1, bar is represented by 2, and the argument/result
is represented by a single byte as well:
encode({foo, X}) -> [1, X];
encode({bar, Y}) -> [2, Y].
decode([Int]) -> Int.
The resulting Erlang program, including functions for stopping the port and detecting port failures, is as follows:
-module(complex5).
-export([start/1, stop/0, init/1]).
-export([foo/1, bar/1]).
start(SharedLib) ->
case erl_ddll:load_driver(".", SharedLib) of
ok -> ok;
{error, already_loaded} -> ok;
_ -> exit({error, could_not_load_driver})
end,
spawn(?MODULE, init, [SharedLib]).
init(SharedLib) ->
register(complex, self()),
Port = open_port({spawn, SharedLib}, []),
loop(Port).
stop() ->
complex ! stop.
foo(X) ->
call_port({foo, X}).
bar(Y) ->
call_port({bar, Y}).
call_port(Msg) ->
complex ! {call, self(), Msg},
receive
{complex, Result} ->
Result
end.
loop(Port) ->
receive
{call, Caller, Msg} ->
Port ! {self(), {command, encode(Msg)}},
receive
{Port, {data, Data}} ->
Caller ! {complex, decode(Data)}
end,
268 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.6 Port Drivers
loop(Port);
stop ->
Port ! {self(), close},
receive
{Port, closed} ->
exit(normal)
end;
{'EXIT', Port, Reason} ->
io:format("~p ~n", [Reason]),
exit(port_terminated)
end.
encode({foo, X}) -> [1, X];
encode({bar, Y}) -> [2, Y].
decode([Int]) -> Int.
8.6.2 C Driver
The C driver is a module that is compiled and linked into a shared library. It uses a driver structure and includes the
header file erl_driver.h.
The driver structure is filled with the driver name and function pointers. It is returned from the special entry point,
declared with the macro DRIVER_INIT(<driver_name>).
The functions for receiving and sending data are combined into a function, pointed out by the driver structure. The
data sent into the port is given as arguments, and the replied data is sent with the C-function driver_output.
As the driver is a shared module, not a program, no main function is present. All function pointers are not used in this
example, and the corresponding fields in the driver_entry structure are set to NULL.
All functions in the driver takes a handle (returned from start) that is just passed along by the Erlang process. This
must in some way refer to the port driver instance.
The example_drv_start, is the only function that is called with a handle to the port instance, so this must be
saved. It is customary to use an allocated driver-defined structure for this one, and to pass a pointer back as a reference.
It is not a good idea to use a global variable as the port driver can be spawned by multiple Erlang processes. This
driver-structure is to be instantiated multiple times:
/* port_driver.c */
#include <stdio.h>
#include "erl_driver.h"
typedef struct {
ErlDrvPort port;
} example_data;
static ErlDrvData example_drv_start(ErlDrvPort port, char *buff)
{
example_data* d = (example_data*)driver_alloc(sizeof(example_data));
d->port = port;
return (ErlDrvData)d;
}
static void example_drv_stop(ErlDrvData handle)
{
driver_free((char*)handle);
}
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 269
8.6 Port Drivers
static void example_drv_output(ErlDrvData handle, char *buff,
ErlDrvSizeT bufflen)
{
example_data* d = (example_data*)handle;
char fn = buff[0], arg = buff[1], res;
if (fn == 1) {
res = foo(arg);
} else if (fn == 2) {
res = bar(arg);
}
driver_output(d->port, &res, 1);
}
ErlDrvEntry example_driver_entry = {
NULL,
/* F_PTR init, called when driver is loaded */
example_drv_start, /* L_PTR start, called when port is opened */
example_drv_stop, /* F_PTR stop, called when port is closed */
example_drv_output, /* F_PTR output, called when erlang has sent */
NULL,
/* F_PTR ready_input, called when input descriptor ready */
NULL,
/* F_PTR ready_output, called when output descriptor ready */
"example_drv", /* char *driver_name, the argument to open_port */
NULL,
/* F_PTR finish, called when unloaded */
NULL,
/* void *handle, Reserved by VM */
NULL,
/* F_PTR control, port_command callback */
NULL,
/* F_PTR timeout, reserved */
NULL,
/* F_PTR outputv, reserved */
NULL,
/* F_PTR ready_async, only for async drivers */
NULL,
/* F_PTR flush, called when port is about
to be closed, but there is data in driver
queue */
NULL,
/* F_PTR call, much like control, sync call
to driver */
NULL,
/* F_PTR event, called when an event selected
by driver_event() occurs. */
ERL_DRV_EXTENDED_MARKER,
/* int extended marker, Should always be
set to indicate driver versioning */
ERL_DRV_EXTENDED_MAJOR_VERSION, /* int major_version, should always be
set to this value */
ERL_DRV_EXTENDED_MINOR_VERSION, /* int minor_version, should always be
set to this value */
0,
/* int driver_flags, see documentation */
NULL,
/* void *handle2, reserved for VM use */
NULL,
/* F_PTR process_exit, called when a
monitored process dies */
NULL
/* F_PTR stop_select, called to close an
event object */
};
DRIVER_INIT(example_drv) /* must match name in driver_entry */
{
return &example_driver_entry;
}
8.6.3 Running the Example
Step 1. Compile the C code:
unix> gcc -o example_drv.so -fpic -shared complex.c port_driver.c
windows> cl -LD -MD -Fe example_drv.dll complex.c port_driver.c
270 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.7 C Nodes
Step 2. Start Erlang and compile the Erlang code:
> erl
Erlang (BEAM) emulator version 5.1
Eshell V5.1 (abort with ^G)
1> c(complex5).
{ok,complex5}
Step 3. Run the example:
2> complex5:start("example_drv").
<0.34.0>
3> complex5:foo(3).
4
4> complex5:bar(5).
10
5> complex5:stop().
stop
8.7 C Nodes
This section outlines an example of how to solve the example problem in Problem Example by using a C node. Notice
that a C node is not typically used for solving simple problems like this, a port is sufficient.
8.7.1 Erlang Program
From Erlang's point of view, the C node is treated like a normal Erlang node. Thus, calling the functions foo and bar
only involves sending a message to the C node asking for the function to be called, and receiving the result. Sending a
message requires a recipient, that is, a process that can be defined using either a pid or a tuple, consisting of a registered
name and a node name. In this case, a tuple is the only alternative as no pid is known:
{RegName, Node} ! Msg
The node name Node is to be the name of the C node. If short node names are used, the plain name of the node is cN,
where N is an integer. If long node names are used, there is no such restriction. An example of a C node name using
short node names is thus c1@idril, an example using long node names is cnode@idril.ericsson.se.
The registered name, RegName, can be any atom. The name can be ignored by the C code, or, for example, be used
to distinguish between different types of messages. An example of Erlang code using short node names follows:
-module(complex3).
-export([foo/1, bar/1]).
foo(X) ->
call_cnode({foo, X}).
bar(Y) ->
call_cnode({bar, Y}).
call_cnode(Msg) ->
{any, c1@idril} ! {call, self(), Msg},
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 271
8.7 C Nodes
receive
{cnode, Result} ->
Result
end.
When using long node names, the code is slightly different as shown in the following example:
-module(complex4).
-export([foo/1, bar/1]).
foo(X) ->
call_cnode({foo, X}).
bar(Y) ->
call_cnode({bar, Y}).
call_cnode(Msg) ->
{any, 'cnode@idril.du.uab.ericsson.se'} ! {call, self(), Msg},
receive
{cnode, Result} ->
Result
end.
8.7.2 C Program
Setting Up Communication
Before calling any other function in Erl_Interface, the memory handling must be initiated:
erl_init(NULL, 0);
Now the C node can be initiated. If short node names are used, this is done by calling erl_connect_init():
erl_connect_init(1, "secretcookie", 0);
Here:
•
The first argument is the integer used to construct the node name.
•
•
In the example, the plain node name is c1.
The second argument is a string defining the magic cookie.
The third argument is an integer that is used to identify a particular instance of a C node.
If long node node names are used, initiation is done by calling erl_connect_xinit():
erl_connect_xinit("idril", "cnode", "cnode@idril.ericsson.se",
&addr, "secretcookie", 0);
Here:
•
•
The first argument is the host name.
The second argument is the plain node name.
272 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.7 C Nodes
•
•
•
•
The third argument is the full node name.
The fourth argument is a pointer to an in_addr struct with the IP address of the host.
The fifth argument is the magic cookie.
The sixth argument is the instance number.
The C node can act as a server or a client when setting up the Erlang-C communication. If it acts as a client, it connects
to an Erlang node by calling erl_connect(), which returns an open file descriptor at success:
fd = erl_connect("e1@idril");
If the C node acts as a server, it must first create a socket (call bind() and listen()) listening to a certain port
number port. It then publishes its name and port number with epmd, the Erlang port mapper daemon. For details,
see the epmd manual page in ERTS:
erl_publish(port);
Now the C node server can accept connections from Erlang nodes:
fd = erl_accept(listen, &conn);
The second argument to erl_accept is a struct ErlConnect which contains useful information when a connection
has been established, for example, the name of the Erlang node.
Sending and Receiving Messages
The C node can receive a message from Erlang by calling erl_receive msg(). This function reads data from the
open file descriptor fd into a buffer and puts the result in an ErlMessage struct emsg. ErlMessage has a field
type defining what kind of data is received. In this case, the type of interest is ERL_REG_SEND which indicates that
Erlang sent a message to a registered process at the C node. The actual message, an ETERM, is in the msg field.
It is also necessary to take care of the types ERL_ERROR (an error occurred) and ERL_TICK (alive check from other
node, is to be ignored). Other possible types indicate process events such as link, unlink, and exit:
while (loop) {
got = erl_receive_msg(fd, buf, BUFSIZE, &emsg);
if (got == ERL_TICK) {
/* ignore */
} else if (got == ERL_ERROR) {
loop = 0; /* exit while loop */
} else {
if (emsg.type == ERL_REG_SEND) {
As the message is an ETERM struct, Erl_Interface functions can be used to manipulate it. In this case, the message
becomes a 3-tuple, because that is how the Erlang code is written. The second element will be the pid of the caller and
the third element will be the tuple {Function,Arg} determining which function to call, and with which argument.
The result of calling the function is made into an ETERM struct as well and sent back to Erlang using erl_send(),
which takes the open file descriptor, a pid, and a term as arguments:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 273
8.7 C Nodes
fromp = erl_element(2, emsg.msg);
tuplep = erl_element(3, emsg.msg);
fnp = erl_element(1, tuplep);
argp = erl_element(2, tuplep);
if (strncmp(ERL_ATOM_PTR(fnp), "foo", 3) == 0) {
res = foo(ERL_INT_VALUE(argp));
} else if (strncmp(ERL_ATOM_PTR(fnp), "bar", 3) == 0) {
res = bar(ERL_INT_VALUE(argp));
}
resp = erl_format("{cnode, ~i}", res);
erl_send(fd, fromp, resp);
Finally, the memory allocated by the ETERM creating functions (including erl_receive_msg() must be freed:
erl_free_term(emsg.from); erl_free_term(emsg.msg);
erl_free_term(fromp); erl_free_term(tuplep);
erl_free_term(fnp); erl_free_term(argp);
erl_free_term(resp);
The following examples show the resulting C programs. First a C node server using short node names:
/* cnode_s.c */
#include
#include
#include
#include
<stdio.h>
<sys/types.h>
<sys/socket.h>
<netinet/in.h>
#include "erl_interface.h"
#include "ei.h"
#define BUFSIZE 1000
int main(int argc, char **argv) {
int port;
int listen;
int fd;
ErlConnect conn;
int loop = 1;
int got;
unsigned char buf[BUFSIZE];
ErlMessage emsg;
/*
/*
/*
/*
Listen port number */
Listen socket */
fd to Erlang node */
Connection data */
/*
/*
/*
/*
Loop flag */
Result of receive */
Buffer for incoming message */
Incoming message */
ETERM *fromp, *tuplep, *fnp, *argp, *resp;
int res;
port = atoi(argv[1]);
erl_init(NULL, 0);
if (erl_connect_init(1, "secretcookie", 0) == -1)
erl_err_quit("erl_connect_init");
/* Make a listen socket */
if ((listen = my_listen(port)) <= 0)
erl_err_quit("my_listen");
274 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.7 C Nodes
if (erl_publish(port) == -1)
erl_err_quit("erl_publish");
if ((fd = erl_accept(listen, &conn)) == ERL_ERROR)
erl_err_quit("erl_accept");
fprintf(stderr, "Connected to %s\n\r", conn.nodename);
while (loop) {
got = erl_receive_msg(fd, buf, BUFSIZE, &emsg);
if (got == ERL_TICK) {
/* ignore */
} else if (got == ERL_ERROR) {
loop = 0;
} else {
if (emsg.type == ERL_REG_SEND) {
fromp = erl_element(2, emsg.msg);
tuplep = erl_element(3, emsg.msg);
fnp = erl_element(1, tuplep);
argp = erl_element(2, tuplep);
if (strncmp(ERL_ATOM_PTR(fnp), "foo", 3) == 0) {
res = foo(ERL_INT_VALUE(argp));
} else if (strncmp(ERL_ATOM_PTR(fnp), "bar", 3) == 0) {
res = bar(ERL_INT_VALUE(argp));
}
resp = erl_format("{cnode, ~i}", res);
erl_send(fd, fromp, resp);
erl_free_term(emsg.from); erl_free_term(emsg.msg);
erl_free_term(fromp); erl_free_term(tuplep);
erl_free_term(fnp); erl_free_term(argp);
erl_free_term(resp);
}
}
} /* while */
}
int my_listen(int port) {
int listen_fd;
struct sockaddr_in addr;
int on = 1;
if ((listen_fd = socket(AF_INET, SOCK_STREAM, 0)) < 0)
return (-1);
setsockopt(listen_fd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on));
memset((void*) &addr, 0, (size_t) sizeof(addr));
addr.sin_family = AF_INET;
addr.sin_port = htons(port);
addr.sin_addr.s_addr = htonl(INADDR_ANY);
if (bind(listen_fd, (struct sockaddr*) &addr, sizeof(addr)) < 0)
return (-1);
}
listen(listen_fd, 5);
return listen_fd;
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 275
8.7 C Nodes
A C node server using long node names:
/* cnode_s2.c */
#include
#include
#include
#include
<stdio.h>
<sys/types.h>
<sys/socket.h>
<netinet/in.h>
#include "erl_interface.h"
#include "ei.h"
#define BUFSIZE 1000
int main(int argc, char **argv) {
struct in_addr addr;
int port;
int listen;
int fd;
ErlConnect conn;
int loop = 1;
int got;
unsigned char buf[BUFSIZE];
ErlMessage emsg;
/*
/*
/*
/*
/*
32-bit IP number of host */
Listen port number */
Listen socket */
fd to Erlang node */
Connection data */
/*
/*
/*
/*
Loop flag */
Result of receive */
Buffer for incoming message */
Incoming message */
ETERM *fromp, *tuplep, *fnp, *argp, *resp;
int res;
port = atoi(argv[1]);
erl_init(NULL, 0);
addr.s_addr = inet_addr("134.138.177.89");
if (erl_connect_xinit("idril", "cnode", "cnode@idril.du.uab.ericsson.se",
&addr, "secretcookie", 0) == -1)
erl_err_quit("erl_connect_xinit");
/* Make a listen socket */
if ((listen = my_listen(port)) <= 0)
erl_err_quit("my_listen");
if (erl_publish(port) == -1)
erl_err_quit("erl_publish");
if ((fd = erl_accept(listen, &conn)) == ERL_ERROR)
erl_err_quit("erl_accept");
fprintf(stderr, "Connected to %s\n\r", conn.nodename);
while (loop) {
got = erl_receive_msg(fd, buf, BUFSIZE, &emsg);
if (got == ERL_TICK) {
/* ignore */
} else if (got == ERL_ERROR) {
loop = 0;
} else {
if (emsg.type == ERL_REG_SEND) {
fromp = erl_element(2, emsg.msg);
tuplep = erl_element(3, emsg.msg);
fnp = erl_element(1, tuplep);
276 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.7 C Nodes
argp = erl_element(2, tuplep);
if (strncmp(ERL_ATOM_PTR(fnp), "foo", 3) == 0) {
res = foo(ERL_INT_VALUE(argp));
} else if (strncmp(ERL_ATOM_PTR(fnp), "bar", 3) == 0) {
res = bar(ERL_INT_VALUE(argp));
}
resp = erl_format("{cnode, ~i}", res);
erl_send(fd, fromp, resp);
erl_free_term(emsg.from); erl_free_term(emsg.msg);
erl_free_term(fromp); erl_free_term(tuplep);
erl_free_term(fnp); erl_free_term(argp);
erl_free_term(resp);
}
}
}
}
int my_listen(int port) {
int listen_fd;
struct sockaddr_in addr;
int on = 1;
if ((listen_fd = socket(AF_INET, SOCK_STREAM, 0)) < 0)
return (-1);
setsockopt(listen_fd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on));
memset((void*) &addr, 0, (size_t) sizeof(addr));
addr.sin_family = AF_INET;
addr.sin_port = htons(port);
addr.sin_addr.s_addr = htonl(INADDR_ANY);
if (bind(listen_fd, (struct sockaddr*) &addr, sizeof(addr)) < 0)
return (-1);
}
listen(listen_fd, 5);
return listen_fd;
Finally, the code for the C node client:
/* cnode_c.c */
#include
#include
#include
#include
<stdio.h>
<sys/types.h>
<sys/socket.h>
<netinet/in.h>
#include "erl_interface.h"
#include "ei.h"
#define BUFSIZE 1000
int main(int argc, char **argv) {
int fd;
int loop = 1;
/* fd to Erlang node */
/* Loop flag */
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 277
8.7 C Nodes
int got;
unsigned char buf[BUFSIZE];
ErlMessage emsg;
/* Result of receive */
/* Buffer for incoming message */
/* Incoming message */
ETERM *fromp, *tuplep, *fnp, *argp, *resp;
int res;
erl_init(NULL, 0);
if (erl_connect_init(1, "secretcookie", 0) == -1)
erl_err_quit("erl_connect_init");
if ((fd = erl_connect("e1@idril")) < 0)
erl_err_quit("erl_connect");
fprintf(stderr, "Connected to ei@idril\n\r");
while (loop) {
got = erl_receive_msg(fd, buf, BUFSIZE, &emsg);
if (got == ERL_TICK) {
/* ignore */
} else if (got == ERL_ERROR) {
loop = 0;
} else {
if (emsg.type == ERL_REG_SEND) {
fromp = erl_element(2, emsg.msg);
tuplep = erl_element(3, emsg.msg);
fnp = erl_element(1, tuplep);
argp = erl_element(2, tuplep);
if (strncmp(ERL_ATOM_PTR(fnp), "foo", 3) == 0) {
res = foo(ERL_INT_VALUE(argp));
} else if (strncmp(ERL_ATOM_PTR(fnp), "bar", 3) == 0) {
res = bar(ERL_INT_VALUE(argp));
}
resp = erl_format("{cnode, ~i}", res);
erl_send(fd, fromp, resp);
erl_free_term(emsg.from); erl_free_term(emsg.msg);
erl_free_term(fromp); erl_free_term(tuplep);
erl_free_term(fnp); erl_free_term(argp);
erl_free_term(resp);
}
}
}
}
8.7.3 Running the Example
Step 1. Compile the C code. This provides the paths to the Erl_Interface include files and libraries, and to the socket
and nsl libraries:
> gcc -o cserver \\
-I/usr/local/otp/lib/erl_interface-3.2.1/include \\
-L/usr/local/otp/lib/erl_interface-3.2.1/lib \\
complex.c cnode_s.c \\
-lerl_interface -lei -lsocket -lnsl
unix> gcc -o cserver2 \\
-I/usr/local/otp/lib/erl_interface-3.2.1/include \\
278 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.7 C Nodes
-L/usr/local/otp/lib/erl_interface-3.2.1/lib \\
complex.c cnode_s2.c \\
-lerl_interface -lei -lsocket -lnsl
unix> gcc -o cclient \\
-I/usr/local/otp/lib/erl_interface-3.2.1/include \\
-L/usr/local/otp/lib/erl_interface-3.2.1/lib \\
complex.c cnode_c.c \\
-lerl_interface -lei -lsocket -lnsl
In Erlang/OTP R5B and later versions of OTP, the include and lib directories are situated under OTPROOT/lib/
erl_interface-VSN, where OTPROOT is the root directory of the OTP installation (/usr/local/otp in the
recent example) and VSN is the version of the Erl_Interface application (3.2.1 in the recent example).
In R4B and earlier versions of OTP, include and lib are situated under OTPROOT/usr.
Step 2. Compile the Erlang code:
unix> erl -compile complex3 complex4
Step 3. Run the C node server example with short node names.
Do as follows:
•
•
•
Start the C program cserver and Erlang in different windows.
cserver takes a port number as argument and must be started before trying to call the Erlang functions.
The Erlang node is to be given the short name e1 and must be set to use the same magic cookie as the C node,
secretcookie:
unix> cserver 3456
unix> erl -sname e1 -setcookie secretcookie
Erlang (BEAM) emulator version 4.9.1.2
Eshell V4.9.1.2 (abort with ^G)
(e1@idril)1> complex3:foo(3).
4
(e1@idril)2> complex3:bar(5).
10
Step 4. Run the C node client example. Terminate cserver, but not Erlang, and start cclient. The Erlang node
must be started before the C node client:
unix> cclient
(e1@idril)3> complex3:foo(3).
4
(e1@idril)4> complex3:bar(5).
10
Step 5. Run the C node server example with long node names:
unix> cserver2 3456
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 279
8.8 NIFs
unix> erl -name e1 -setcookie secretcookie
Erlang (BEAM) emulator version 4.9.1.2
Eshell V4.9.1.2 (abort with ^G)
(e1@idril.du.uab.ericsson.se)1> complex4:foo(3).
4
(e1@idril.du.uab.ericsson.se)2> complex4:bar(5).
10
8.8 NIFs
This section outlines an example of how to solve the example problem in Problem Example by using Native
Implemented Functions (NIFs).
NIFs were introduced in Erlang/OTP R13B03 as an experimental feature. It is a simpler and more efficient way of
calling C-code than using port drivers. NIFs are most suitable for synchronous functions, such as foo and bar in the
example, that do some relatively short calculations without side effects and return the result.
A NIF is a function that is implemented in C instead of Erlang. NIFs appear as any other functions to the callers. They
belong to a module and are called like any other Erlang functions. The NIFs of a module are compiled and linked into
a dynamic loadable, shared library (SO in UNIX, DLL in Windows). The NIF library must be loaded in runtime by
the Erlang code of the module.
As a NIF library is dynamically linked into the emulator process, this is the fastest way of calling C-code from Erlang
(alongside port drivers). Calling NIFs requires no context switches. But it is also the least safe, because a crash in a
NIF brings the emulator down too.
8.8.1 Erlang Program
Even if all functions of a module are NIFs, an Erlang module is still needed for two reasons:
•
•
The NIF library must be explicitly loaded by Erlang code in the same module.
All NIFs of a module must have an Erlang implementation as well.
Normally these are minimal stub implementations that throw an exception. But they can also be used as fallback
implementations for functions that do not have native implemenations on some architectures.
NIF libraries are loaded by calling erlang:load_nif/2, with the name of the shared library as argument. The
second argument can be any term that will be passed on to the library and used for initialization:
-module(complex6).
-export([foo/1, bar/1]).
-on_load(init/0).
init() ->
ok = erlang:load_nif("./complex6_nif", 0).
foo(_X) ->
exit(nif_library_not_loaded).
bar(_Y) ->
exit(nif_library_not_loaded).
Here, the directive on_load is used to get function init to be automatically called when the module is loaded. If
init returns anything other than ok, such when the loading of the NIF library fails in this example, the module is
unloaded and calls to functions within it, fail.
280 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
8.8 NIFs
Loading the NIF library overrides the stub implementations and cause calls to foo and bar to be dispatched to the
NIF implementations instead.
8.8.2 NIF Library Code
The NIFs of the module are compiled and linked into a shared library. Each NIF is implemented as a normal C function.
The macro ERL_NIF_INIT together with an array of structures defines the names, arity, and function pointers of
all the NIFs in the module. The header file erl_nif.h must be included. As the library is a shared module, not a
program, no main function is to be present.
The function arguments passed to a NIF appears in an array argv, with argc as the length of the array, and thus the
arity of the function. The Nth argument of the function can be accessed as argv[N-1]. NIFs also take an environment
argument that serves as an opaque handle that is needed to be passed on to most API functions. The environment
contains information about the calling Erlang process:
#include <erl_nif.h>
extern int foo(int x);
extern int bar(int y);
static ERL_NIF_TERM foo_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
{
int x, ret;
if (!enif_get_int(env, argv[0], &x)) {
return enif_make_badarg(env);
}
ret = foo(x);
return enif_make_int(env, ret);
}
static ERL_NIF_TERM bar_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
{
int y, ret;
if (!enif_get_int(env, argv[0], &y)) {
return enif_make_badarg(env);
}
ret = bar(y);
return enif_make_int(env, ret);
}
static ErlNifFunc nif_funcs[] = {
{"foo", 1, foo_nif},
{"bar", 1, bar_nif}
};
ERL_NIF_INIT(complex6, nif_funcs, NULL, NULL, NULL, NULL)
Here,ERL_NIF_INIT has the following arguments:
•
•
•
The first argument must be the name of the Erlang module as a C-identifier. It will be stringified by the macro.
The second argument is the array of ErlNifFunc structures containing name, arity, and function pointer of
each NIF.
The remaining arguments are pointers to callback functions that can be used to initialize the library. They are
not used in this simple example, hence they are all set to NULL.
Function arguments and return values are represented as values of type ERL_NIF_TERM. Here, functions like
enif_get_int and enif_make_int are used to convert between Erlang term and C-type. If the function
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 281
8.8 NIFs
argument argv[0] is not an integer, enif_get_int returns false, in which case it returns by throwing a badargexception with enif_make_badarg.
8.8.3 Running the Example
Step 1. Compile the C code:
unix> gcc -o complex6_nif.so -fpic -shared complex.c complex6_nif.c
windows> cl -LD -MD -Fe complex6_nif.dll complex.c complex6_nif.c
Step 2: Start Erlang and compile the Erlang code:
> erl
Erlang R13B04 (erts-5.7.5) [64-bit] [smp:4:4] [rq:4] [async-threads:0] [kernel-poll:false]
Eshell V5.7.5 (abort with ^G)
1> c(complex6).
{ok,complex6}
Step 3: Run the example:
3>
4
4>
10
5>
**
complex6:foo(3).
complex6:bar(5).
complex6:foo("not an integer").
exception error: bad argument
in function complex6:foo/1
called as comlpex6:foo("not an integer")
282 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.1 Overview
9 OTP Design Principles
9.1 Overview
The OTP design principles define how to structure Erlang code in terms of processes, modules, and directories.
9.1.1 Supervision Trees
A basic concept in Erlang/OTP is the supervision tree. This is a process structuring model based on the idea of
workers and supervisors:
•
•
•
Workers are processes that perform computations, that is, they do the actual work.
Supervisors are processes that monitor the behaviour of workers. A supervisor can restart a worker if something
goes wrong.
The supervision tree is a hierarchical arrangement of code into supervisors and workers, which makes it
possible to design and program fault-tolerant software.
In the following figure, square boxes represents supervisors and circles represent workers:
Figure 1.1: Supervision Tree
9.1.2 Behaviours
In a supervision tree, many of the processes have similar structures, they follow similar patterns. For example, the
supervisors are similar in structure. The only difference between them is which child processes they supervise. Many
of the workers are servers in a server-client relation, finite-state machines, or event handlers such as error loggers.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 283
9.1 Overview
Behaviours are formalizations of these common patterns. The idea is to divide the code for a process in a generic part
(a behaviour module) and a specific part (a callback module).
The behaviour module is part of Erlang/OTP. To implement a process such as a supervisor, the user only has to
implement the callback module which is to export a pre-defined set of functions, the callback functions.
The following example illustrate how code can be divided into a generic and a specific part. Consider the following
code (written in plain Erlang) for a simple server, which keeps track of a number of "channels". Other processes can
allocate and free the channels by calling the functions alloc/0 and free/1, respectively.
-module(ch1).
-export([start/0]).
-export([alloc/0, free/1]).
-export([init/0]).
start() ->
spawn(ch1, init, []).
alloc() ->
ch1 ! {self(), alloc},
receive
{ch1, Res} ->
Res
end.
free(Ch) ->
ch1 ! {free, Ch},
ok.
init() ->
register(ch1, self()),
Chs = channels(),
loop(Chs).
loop(Chs) ->
receive
{From, alloc} ->
{Ch, Chs2} = alloc(Chs),
From ! {ch1, Ch},
loop(Chs2);
{free, Ch} ->
Chs2 = free(Ch, Chs),
loop(Chs2)
end.
The code for the server can be rewritten into a generic part server.erl:
-module(server).
-export([start/1]).
-export([call/2, cast/2]).
-export([init/1]).
start(Mod) ->
spawn(server, init, [Mod]).
call(Name, Req) ->
Name ! {call, self(), Req},
receive
{Name, Res} ->
Res
284 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.1 Overview
end.
cast(Name, Req) ->
Name ! {cast, Req},
ok.
init(Mod) ->
register(Mod, self()),
State = Mod:init(),
loop(Mod, State).
loop(Mod, State) ->
receive
{call, From, Req} ->
{Res, State2} = Mod:handle_call(Req, State),
From ! {Mod, Res},
loop(Mod, State2);
{cast, Req} ->
State2 = Mod:handle_cast(Req, State),
loop(Mod, State2)
end.
And a callback module ch2.erl:
-module(ch2).
-export([start/0]).
-export([alloc/0, free/1]).
-export([init/0, handle_call/2, handle_cast/2]).
start() ->
server:start(ch2).
alloc() ->
server:call(ch2, alloc).
free(Ch) ->
server:cast(ch2, {free, Ch}).
init() ->
channels().
handle_call(alloc, Chs) ->
alloc(Chs). % => {Ch,Chs2}
handle_cast({free, Ch}, Chs) ->
free(Ch, Chs). % => Chs2
Notice the following:
•
•
•
•
The code in server can be reused to build many different servers.
The server name, in this example the atom ch2, is hidden from the users of the client functions. This means
that the name can be changed without affecting them.
The protocol (messages sent to and received from the server) is also hidden. This is good programming practice
and allows one to change the protocol without changing the code using the interface functions.
The functionality of server can be extended without having to change ch2 or any other callback module.
In ch1.erl and ch2.erl above, the implementation of channels/0, alloc/1, and free/2 has been
intentionally left out, as it is not relevant to the example. For completeness, one way to write these functions are
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 285
9.1 Overview
given below. This is an example only, a realistic implementation must be able to handle situations like running out
of channels to allocate, and so on.
channels() ->
{_Allocated = [], _Free = lists:seq(1,100)}.
alloc({Allocated, [H|T] = _Free}) ->
{H, {[H|Allocated], T}}.
free(Ch, {Alloc, Free} = Channels) ->
case lists:member(Ch, Alloc) of
true ->
{lists:delete(Ch, Alloc), [Ch|Free]};
false ->
Channels
end.
Code written without using behaviours can be more efficient, but the increased efficiency is at the expense of generality.
The ability to manage all applications in the system in a consistent manner is important.
Using behaviours also makes it easier to read and understand code written by other programmers. Improvised
programming structures, while possibly more efficient, are always more difficult to understand.
The server module corresponds, greatly simplified, to the Erlang/OTP behaviour gen_server.
The standard Erlang/OTP behaviours are:
•
gen_server
•
For implementing the server of a client-server relation
gen_fsm
•
For implementing finite-state machines (Old)
gen_statem
•
For implementing state machines (New)
gen_event
•
For implementing event handling functionality
supervisor
For implementing a supervisor in a supervision tree
The compiler understands the module attribute -behaviour(Behaviour) and issues warnings about missing
callback functions, for example:
-module(chs3).
-behaviour(gen_server).
...
3> c(chs3).
./chs3.erl:10: Warning: undefined call-back function handle_call/3
{ok,chs3}
9.1.3 Applications
Erlang/OTP comes with a number of components, each implementing some specific functionality. Components are
with Erlang/OTP terminology called applications. Examples of Erlang/OTP applications are Mnesia, which has
286 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.2 gen_server Behaviour
everything needed for programming database services, and Debugger, which is used to debug Erlang programs. The
minimal system based on Erlang/OTP consists of the following two applications:
•
•
Kernel - Functionality necessary to run Erlang
STDLIB - Erlang standard libraries
The application concept applies both to program structure (processes) and directory structure (modules).
The simplest applications do not have any processes, but consist of a collection of functional modules. Such an
application is called a library application. An example of a library application is STDLIB.
An application with processes is easiest implemented as a supervision tree using the standard behaviours.
How to program applications is described in Applications.
9.1.4 Releases
A release is a complete system made out from a subset of Erlang/OTP applications and a set of user-specific
applications.
How to program releases is described in Releases.
How to install a release in a target environment is described in the section about target systems in Section 2 System
Principles.
9.1.5 Release Handling
Release handling is upgrading and downgrading between different versions of a release, in a (possibly) running
system. How to do this is described in Release Handling.
9.2 gen_server Behaviour
This section is to be read with the gen_server(3) manual page in stdblib, where all interface functions and callback
functions are described in detail.
9.2.1 Client-Server Principles
The client-server model is characterized by a central server and an arbitrary number of clients. The client-server model
is used for resource management operations, where several different clients want to share a common resource. The
server is responsible for managing this resource.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 287
9.2 gen_server Behaviour
Figure 2.1: Client-Server Model
9.2.2 Example
An example of a simple server written in plain Erlang is provided in Overview. The server can be reimplemented using
gen_server, resulting in this callback module:
-module(ch3).
-behaviour(gen_server).
-export([start_link/0]).
-export([alloc/0, free/1]).
-export([init/1, handle_call/3, handle_cast/2]).
start_link() ->
gen_server:start_link({local, ch3}, ch3, [], []).
alloc() ->
gen_server:call(ch3, alloc).
free(Ch) ->
gen_server:cast(ch3, {free, Ch}).
init(_Args) ->
{ok, channels()}.
handle_call(alloc, _From, Chs) ->
{Ch, Chs2} = alloc(Chs),
{reply, Ch, Chs2}.
288 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.2 gen_server Behaviour
handle_cast({free, Ch}, Chs) ->
Chs2 = free(Ch, Chs),
{noreply, Chs2}.
The code is explained in the next sections.
9.2.3 Starting a Gen_Server
In the example in the previous section, gen_server is started by calling ch3:start_link():
start_link() ->
gen_server:start_link({local, ch3}, ch3, [], []) => {ok, Pid}
start_link calls function gen_server:start_link/4. This function spawns and links to a new process, a
gen_server.
•
•
•
•
The first argument, {local, ch3}, specifies the name. The gen_server is then locally registered as ch3.
If the name is omitted, the gen_server is not registered. Instead its pid must be used. The name can also be given
as {global, Name}, in which case the gen_server is registered using global:register_name/2.
The second argument, ch3, is the name of the callback module, that is, the module where the callback functions
are located.
The interface functions (start_link, alloc, and free) are then located in the same module as the callback
functions (init, handle_call, and handle_cast). This is normally good programming practice, to have
the code corresponding to one process contained in one module.
The third argument, [], is a term that is passed as is to the callback function init. Here, init does not need
any indata and ignores the argument.
The fourth argument, [], is a list of options. See the gen_server(3) manual page for available options.
If name registration succeeds, the new gen_server process calls the callback function ch3:init([]). init is
expected to return {ok, State}, where State is the internal state of the gen_server. In this case, the state
is the available channels.
init(_Args) ->
{ok, channels()}.
gen_server:start_link is synchronous. It does not return until the gen_server has been initialized and is
ready to receive requests.
gen_server:start_link must be used if the gen_server is part of a supervision tree, that is, started by
a supervisor. There is another function, gen_server:start, to start a standalone gen_server, that is, a
gen_server that is not part of a supervision tree.
9.2.4 Synchronous Requests - Call
The synchronous request alloc() is implemented using gen_server:call/2:
alloc() ->
gen_server:call(ch3, alloc).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 289
9.2 gen_server Behaviour
ch3 is the name of the gen_server and must agree with the name used to start it. alloc is the actual request.
The request is made into a message and sent to the gen_server. When the request is received, the
gen_server calls handle_call(Request, From, State), which is expected to return a tuple
{reply,Reply,State1}. Reply is the reply that is to be sent back to the client, and State1 is a new value
for the state of the gen_server.
handle_call(alloc, _From, Chs) ->
{Ch, Chs2} = alloc(Chs),
{reply, Ch, Chs2}.
In this case, the reply is the allocated channel Ch and the new state is the set of remaining available channels Chs2.
Thus, the call ch3:alloc() returns the allocated channel Ch and the gen_server then waits for new requests,
now with an updated list of available channels.
9.2.5 Asynchronous Requests - Cast
The asynchronous request free(Ch) is implemented using gen_server:cast/2:
free(Ch) ->
gen_server:cast(ch3, {free, Ch}).
ch3 is the name of the gen_server. {free, Ch} is the actual request.
The request is made into a message and sent to the gen_server. cast, and thus free, then returns ok.
When the request is received, the gen_server calls handle_cast(Request, State), which is expected to
return a tuple {noreply,State1}. State1 is a new value for the state of the gen_server.
handle_cast({free, Ch}, Chs) ->
Chs2 = free(Ch, Chs),
{noreply, Chs2}.
In this case, the new state is the updated list of available channels Chs2. The gen_server is now ready for new
requests.
9.2.6 Stopping
In a Supervision Tree
If the gen_server is part of a supervision tree, no stop function is needed. The gen_server is automatically
terminated by its supervisor. Exactly how this is done is defined by a shutdown strategy set in the supervisor.
If it is necessary to clean up before termination, the shutdown strategy must be a time-out value and the gen_server
must be set to trap exit signals in function init. When ordered to shutdown, the gen_server then calls the callback
function terminate(shutdown, State):
init(Args) ->
...,
process_flag(trap_exit, true),
...,
{ok, State}.
290 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
...
terminate(shutdown, State) ->
..code for cleaning up here..
ok.
Standalone Gen_Servers
If the gen_server is not part of a supervision tree, a stop function can be useful, for example:
...
export([stop/0]).
...
stop() ->
gen_server:cast(ch3, stop).
...
handle_cast(stop, State) ->
{stop, normal, State};
handle_cast({free, Ch}, State) ->
....
...
terminate(normal, State) ->
ok.
The callback function handling the stop request returns a tuple {stop,normal,State1}, where normal
specifies that it is a normal termination and State1 is a new value for the state of the gen_server. This causes
the gen_server to call terminate(normal, State1) and then it terminates gracefully.
9.2.7 Handling Other Messages
If the gen_server is to be able to receive other messages than requests, the callback function
handle_info(Info, State) must be implemented to handle them. Examples of other messages are exit
messages, if the gen_server is linked to other processes (than the supervisor) and trapping exit signals.
handle_info({'EXIT', Pid, Reason}, State) ->
..code to handle exits here..
{noreply, State1}.
The code_change method must also be implemented.
code_change(OldVsn, State, Extra) ->
..code to convert state (and more) during code change
{ok, NewState}.
9.3 gen_statem Behavior
This section is to be read with the gen_statem(3) manual page in STDLIB, where all interface functions and
callback functions are described in detail.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 291
9.3 gen_statem Behavior
Note:
This is a new behavior in Erlang/OTP 19.0. It has been thoroughly reviewed, is stable enough to be used by at
least two heavy OTP applications, and is here to stay. Depending on user feedback, we do not expect but can find
it necessary to make minor not backward compatible changes into Erlang/OTP 20.0.
9.3.1 Event-Driven State Machines
Established Automata Theory does not deal much with how a state transition is triggered, but assumes that the output
is a function of the input (and the state) and that they are some kind of values.
For an Event-Driven State Machine, the input is an event that triggers a state transition and the output is actions
executed during the state transition. It can analogously to the mathematical model of a Finite-State Machine be
described as a set of relations of the following form:
State(S) x Event(E) -> Actions(A), State(S')
These relations are interpreted as follows: if we are in state S and event E occurs, we are to perform actions A and
make a transition to state S'. Notice that S' can be equal to S.
As A and S' depend only on S and E, the kind of state machine described here is a Mealy Machine (see, for example,
the corresponding Wikipedia article).
Like most gen_ behaviors, gen_statem keeps a server Data besides the state. Because of this, and as there is
no restriction on the number of states (assuming that there is enough virtual machine memory) or on the number of
distinct input events, a state machine implemented with this behavior is in fact Turing complete. But it feels mostly
like an Event-Driven Mealy Machine.
9.3.2 Callback Modes
The gen_statem behavior supports two callback modes:
•
In mode state_functions, the state transition rules are written as some Erlang functions, which conform
to the following convention:
StateName(EventType, EventContent, Data) ->
... code for actions here ...
{next_state, NewStateName, NewData}.
•
This form is used in most examples here for example in section Example.
In mode handle_event_function, only one Erlang function provides all state transition rules:
handle_event(EventType, EventContent, State, Data) ->
... code for actions here ...
{next_state, NewState, NewData}
See section One Event Handler for an example.
292 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
Both these modes allow other return tuples; see Module:StateName/3 in the gen_statem manual page. These
other return tuples can, for example, stop the machine, execute state transition actions on the machine engine itself,
and send replies.
Choosing the Callback Mode
The two callback modes give different possibilities and restrictions, but one goal remains: you want to handle all
possible combinations of events and states.
This can be done, for example, by focusing on one state at the time and for every state ensure that all events are
handled. Alternatively, you can focus on one event at the time and ensure that it is handled in every state. You can
also use a mix of these strategies.
With state_functions, you are restricted to use atom-only states, and the gen_statem engine branches
depending on state name for you. This encourages the callback module to gather the implementation of all event actions
particular to one state in the same place in the code, hence to focus on one state at the time.
This mode fits well when you have a regular state diagram, like the ones in this chapter, which describes all events
and actions belonging to a state visually around that state, and each state has its unique name.
With handle_event_function, you are free to mix strategies, as all events and states are handled in the same
callback function.
This mode works equally well when you want to focus on one event at the time or on one state at the time, but function
Module:handle_event/4 quickly grows too large to handle without branching to helper functions.
The mode enables the use of non-atom states, for example, complex states or even hierarchical states. If, for
example, a state diagram is largely alike for the client side and the server side of a protocol, you can have a state
{StateName,server} or {StateName,client}, and make StateName determine where in the code to
handle most events in the state. The second element of the tuple is then used to select whether to handle special clientside or server-side events.
9.3.3 State Enter Calls
The gen_statem behavior can regardless of callback mode automatically call the state callback with special
arguments whenever the state changes so you can write state entry actions near the rest of the state transition rules.
It typically looks like this:
StateName(enter, _OldState, Data) ->
... code for state entry actions here ...
{keep_state, NewData};
StateName(EventType, EventContent, Data) ->
... code for actions here ...
{next_state, NewStateName, NewData}.
Depending on how your state machine is specified, this can be a very useful feature, but it forces you to handle the
state enter calls in all states. See also the State Entry Actions chapter.
9.3.4 Actions
In the first section Event-Driven State Machines actions were mentioned as a part of the general state machine
model. These general actions are implemented with the code that callback module gen_statem executes in an eventhandling callback function before returning to the gen_statem engine.
There are more specific state-transition actions that a callback function can order the gen_statem engine to do after
the callback function return. These are ordered by returning a list of actions in the return tuple from the callback
function. These state transition actions affect the gen_statem engine itself and can do the following:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 293
9.3 gen_statem Behavior
•
•
•
•
•
•
•
Postpone the current event, see section Postponing Events
Hibernate the gen_statem, treated in Hibernation
Start a state time-out, read more in section State Time-Outs
Start a generic time-out, read more in section Generic Time-Outs
Start an event time-out, see more in section Event Time-Outs
Reply to a caller, mentioned at the end of section All State Events
Generate the next event to handle, see section Self-Generated Events
For details, see the gen_statem(3) manual page. You can, for example, reply to many callers, generate multiple
next events, and set time-outs to relative or absolute times.
9.3.5 Event Types
Events are categorized in different event types. Events of all types are handled in the same callback function, for a
given state, and the function gets EventType and EventContent as arguments.
The following is a complete list of event types and where they come from:
cast
Generated by gen_statem:cast.
{call,From}
Generated by gen_statem:call, where From is the reply address to use when replying either through the
state transition action {reply,From,Msg} or by calling gen_statem:reply.
info
Generated by any regular process message sent to the gen_statem process.
state_timeout
Generated by state transition action {state_timeout,Time,EventContent} state timer timing out.
{timeout,Name}
Generated by state transition action {{timeout,Name},Time,EventContent} generic timer timing
out.
timeout
Generated by state transition action {timeout,Time,EventContent} (or its short form Time) event
timer timing out.
internal
Generated by state transition action {next_event,internal,EventContent}. All event types above
can also be generated using {next_event,EventType,EventContent}.
9.3.6 Example
This example starts off as equivalent to the example in section gen_fsm Behavior. In later sections, additions and
tweaks are made using features in gen_statem that gen_fsm does not have. The end of this chapter provides the
example again with all the added features.
A door with a code lock can be seen as a state machine. Initially, the door is locked. When someone presses a button, an
event is generated. Depending on what buttons have been pressed before, the sequence so far can be correct, incomplete,
or wrong. If correct, the door is unlocked for 10 seconds (10,000 milliseconds). If incomplete, we wait for another
button to be pressed. If wrong, we start all over, waiting for a new button sequence.
294 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
Figure 3.1: Code Lock State Diagram
This code lock state machine can be implemented using gen_statem with the following callback module:
-module(code_lock).
-behaviour(gen_statem).
-define(NAME, code_lock).
-export([start_link/1]).
-export([button/1]).
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 295
9.3 gen_statem Behavior
-export([init/1,callback_mode/0,terminate/3,code_change/4]).
-export([locked/3,open/3]).
start_link(Code) ->
gen_statem:start_link({local,?NAME}, ?MODULE, Code, []).
button(Digit) ->
gen_statem:cast(?NAME, {button,Digit}).
init(Code) ->
do_lock(),
Data = #{code => Code, remaining => Code},
{ok, locked, Data}.
callback_mode() ->
state_functions.
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] ->
do_unlock(),
{next_state, open, Data#{remaining := Code},
[{state_timeout,10000,lock}]};
[Digit|Rest] -> % Incomplete
{next_state, locked, Data#{remaining := Rest}};
_Wrong ->
{next_state, locked, Data#{remaining := Code}}
end.
open(state_timeout, lock, Data) ->
do_lock(),
{next_state, locked, Data};
open(cast, {button,_}, Data) ->
{next_state, open, Data}.
do_lock() ->
io:format("Lock~n", []).
do_unlock() ->
io:format("Unlock~n", []).
terminate(_Reason, State, _Data) ->
State =/= locked andalso do_lock(),
ok.
code_change(_Vsn, State, Data, _Extra) ->
{ok, State, Data}.
The code is explained in the next sections.
9.3.7 Starting gen_statem
In the example in the previous section, gen_statem is started by calling code_lock:start_link(Code):
start_link(Code) ->
gen_statem:start_link({local,?NAME}, ?MODULE, Code, []).
start_link calls function gen_statem:start_link/4, which spawns and links to a new process, a
gen_statem.
296 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
•
•
•
•
The first argument, {local,?NAME}, specifies the name. In this case, the gen_statem is locally registered
as code_lock through the macro ?NAME.
If the name is omitted, the gen_statem is not registered. Instead its pid must be used. The name can also be
specified as {global,Name}, then the gen_statem is registered using global:register_name/2 in
Kernel.
The second argument, ?MODULE, is the name of the callback module, that is, the module where the callback
functions are located, which is this module.
The interface functions (start_link/1 and button/1) are located in the same module as the callback
functions (init/1, locked/3, and open/3). It is normally good programming practice to have the clientside code and the server-side code contained in one module.
The third argument, Code, is a list of digits, which is the correct unlock code that is passed to callback function
init/1.
The fourth argument, [], is a list of options. For the available options, see gen_statem:start_link/3.
If name registration succeeds, the new gen_statem process calls callback function code_lock:init(Code).
This function is expected to return {ok, State, Data}, where State is the initial state of the gen_statem,
in this case locked; assuming that the door is locked to begin with. Data is the internal server data of the
gen_statem. Here the server data is a map with key code that stores the correct button sequence, and key
remaining that stores the remaining correct button sequence (the same as the code to begin with).
init(Code) ->
do_lock(),
Data = #{code => Code, remaining => Code},
{ok,locked,Data}.
Function gen_statem:start_link is synchronous. It does not return until the gen_statem is initialized and
is ready to receive events.
Function gen_statem:start_link must be used if the gen_statem is part of a supervision tree, that is, started
by a supervisor. Another function, gen_statem:start can be used to start a standalone gen_statem, that is,
a gen_statem that is not part of a supervision tree.
callback_mode() ->
state_functions.
Function Module:callback_mode/0 selects the CallbackMode for the callback module, in this case
state_functions. That is, each state has got its own handler function.
9.3.8 Handling Events
The function notifying the code lock about a button event is implemented using gen_statem:cast/2:
button(Digit) ->
gen_statem:cast(?NAME, {button,Digit}).
The first argument is the name of the gen_statem and must agree with the name used to start it. So, we use the
same macro ?NAME as when starting. {button,Digit} is the event content.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 297
9.3 gen_statem Behavior
The event is made into a message and sent to the gen_statem. When the event is received, the gen_statem calls
StateName(cast, Event, Data), which is expected to return a tuple {next_state, NewStateName,
NewData}, or {next_state, NewStateName, NewData, Actions}. StateName is the name of the
current state and NewStateName is the name of the next state to go to. NewData is a new value for the server data
of the gen_statem, and Actions is a list of actions on the gen_statem engine.
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] -> % Complete
do_unlock(),
{next_state, open, Data#{remaining := Code},
[{state_timeout,10000,lock}]};
[Digit|Rest] -> % Incomplete
{next_state, locked, Data#{remaining := Rest}};
[_|_] -> % Wrong
{next_state, locked, Data#{remaining := Code}}
end.
open(state_timeout, lock, Data) ->
do_lock(),
{next_state, locked, Data};
open(cast, {button,_}, Data) ->
{next_state, open, Data}.
If the door is locked and a button is pressed, the pressed button is compared with the next correct button. Depending on
the result, the door is either unlocked and the gen_statem goes to state open, or the door remains in state locked.
If the pressed button is incorrect, the server data restarts from the start of the code sequence.
If the whole code is correct, the server changes states to open.
In state open, a button event is ignored by staying in the same state. This can also be done by returning
{keep_state, Data} or in this case since Data unchanged even by returning keep_state_and_data.
9.3.9 State Time-Outs
When a correct code has been given, the door is unlocked and the following tuple is returned from locked/2:
{next_state, open, Data#{remaining := Code},
[{state_timeout,10000,lock}]};
10,000 is a time-out value in milliseconds. After this time (10 seconds), a time-out occurs. Then,
StateName(state_timeout, lock, Data) is called. The time-out occurs when the door has been in state
open for 10 seconds. After that the door is locked again:
open(state_timeout, lock, Data) ->
do_lock(),
{next_state, locked, Data};
298 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
The timer for a state time-out is automatically cancelled when the state machine changes states. You can restart a state
time-out by setting it to a new time, which cancels the running timer and starts a new. This implies that you can cancel
a state time-out by restarting it with time infinity.
9.3.10 All State Events
Sometimes events can arrive in any state of the gen_statem. It is convenient to handle these in a common state
handler function that all state functions call for events not specific to the state.
Consider a code_length/0 function that returns the length of the correct code (that should not be sensitive to
reveal). We dispatch all events that are not state-specific to the common function handle_event/3:
...
-export([button/1,code_length/0]).
...
code_length() ->
gen_statem:call(?NAME, code_length).
...
locked(...) -> ... ;
locked(EventType, EventContent, Data) ->
handle_event(EventType, EventContent, Data).
...
open(...) -> ... ;
open(EventType, EventContent, Data) ->
handle_event(EventType, EventContent, Data).
handle_event({call,From}, code_length, #{code := Code} = Data) ->
{keep_state, Data, [{reply,From,length(Code)}]}.
This example uses gen_statem:call/2, which waits for a reply from the server. The reply is sent with a
{reply,From,Reply} tuple in an action list in the {keep_state, ...} tuple that retains the current state.
This return form is convenient when you want to stay in the current state but do not know or care about what it is.
9.3.11 One Event Handler
If mode handle_event_function is used, all events are handled in Module:handle_event/4 and we can
(but do not have to) use an event-centered approach where we first branch depending on event and then depending
on state:
...
-export([handle_event/4]).
...
callback_mode() ->
handle_event_function.
handle_event(cast, {button,Digit}, State, #{code := Code} = Data) ->
case State of
locked ->
case maps:get(remaining, Data) of
[Digit] -> % Complete
do_unlock(),
{next_state, open, Data#{remaining := Code},
[{state_timeout,10000,lock}]};
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 299
9.3 gen_statem Behavior
[Digit|Rest] -> % Incomplete
{keep_state, Data#{remaining := Rest}};
[_|_] -> % Wrong
{keep_state, Data#{remaining := Code}}
end;
open ->
keep_state_and_data
end;
handle_event(state_timeout, lock, open, Data) ->
do_lock(),
{next_state, locked, Data}.
...
9.3.12 Stopping
In a Supervision Tree
If the gen_statem is part of a supervision tree, no stop function is needed. The gen_statem is automatically
terminated by its supervisor. Exactly how this is done is defined by a shutdown strategy set in the supervisor.
If it is necessary to clean up before termination, the shutdown strategy must be a time-out value and the gen_statem
must in function init/1 set itself to trap exit signals by calling process_flag(trap_exit, true):
init(Args) ->
process_flag(trap_exit, true),
do_lock(),
...
When ordered to shut down, the gen_statem then calls callback function terminate(shutdown, State,
Data).
In this example, function terminate/3 locks the door if it is open, so we do not accidentally leave the door open
when the supervision tree terminates:
terminate(_Reason, State, _Data) ->
State =/= locked andalso do_lock(),
ok.
Standalone gen_statem
If the gen_statem is not part of a supervision tree, it can be stopped using gen_statem:stop, preferably through
an API function:
...
-export([start_link/1,stop/0]).
...
stop() ->
gen_statem:stop(?NAME).
300 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
This makes the gen_statem call callback function terminate/3 just like for a supervised server and waits for
the process to terminate.
9.3.13 Event Time-Outs
A time-out feature inherited from gen_statem's predecessor gen_fsm, is an event time-out, that is, if an event
arrives the timer is cancelled. You get either an event or a time-out, but not both.
It is ordered by the state transition action {timeout,Time,EventContent}, or just Time, or even just Time
instead of an action list (the latter is a form inherited from gen_fsm.
This type of time-out is useful to for example act on inactivity. Let us restart the code sequence if no button is pressed
for say 30 seconds:
...
locked(
timeout, _,
#{code := Code, remaining := Remaining} = Data) ->
{next_state, locked, Data#{remaining := Code}};
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
...
[Digit|Rest] -> % Incomplete
{next_state, locked, Data#{remaining := Rest}, 30000};
...
Whenever we receive a button event we start an event time-out of 30 seconds, and if we get an event type timeout
we reset the remaining code sequence.
An event time-out is cancelled by any other event so you either get some other event or the time-out event. It is therefore
not possible nor needed to cancel or restart an event time-out. Whatever event you act on has already cancelled the
event time-out...
9.3.14 Generic Time-Outs
The previous example of state time-outs only work if the state machine stays in the same state during the time-out
time. And event time-outs only work if no disturbing unrelated events occur.
You may want to start a timer in one state and respond to the time-out in another, maybe cancel the time-out without
changing states, or perhaps run multiple time-outs in parallel. All this can be accomplished with generic time-outs.
They may look a little bit like event time-outs but contain a name to allow for any number of them simultaneously
and they are not automatically cancelled.
Here is how to accomplish the state time-out in the previous example by instead using a generic time-out named
open_tm:
...
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] ->
do_unlock(),
{next_state, open, Data#{remaining := Code},
[{{timeout,open_tm},10000,lock}]};
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 301
9.3 gen_statem Behavior
...
open({timeout,open_tm}, lock, Data) ->
do_lock(),
{next_state,locked,Data};
open(cast, {button,_}, Data) ->
{keep_state,Data};
...
Just as state time-outs you can restart or cancel a specific generic time-out by setting it to a new time or infinity.
Another way to handle a late time-out can be to not cancel it, but to ignore it if it arrives in a state where it is known
to be late.
9.3.15 Erlang Timers
The most versatile way to handle time-outs is to use Erlang Timers; see erlang:start_timer3,4. Most timeout tasks can be performed with the time-out features in gen_statem, but an example of one that can not is if you
should need the return value from erlang:cancel_timer(Tref), that is; the remaining time of the timer.
Here is how to accomplish the state time-out in the previous example by instead using an Erlang Timer:
...
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] ->
do_unlock(),
Tref = erlang:start_timer(10000, self(), lock),
{next_state, open, Data#{remaining := Code, timer => Tref}};
...
open(info, {timeout,Tref,lock}, #{timer := Tref} = Data) ->
do_lock(),
{next_state,locked,maps:remove(timer, Data)};
open(cast, {button,_}, Data) ->
{keep_state,Data};
...
Removing the timer key from the map when we change to state locked is not strictly necessary since we can only
get into state open with an updated timer map value. But it can be nice to not have outdated values in the state Data!
If you need to cancel a timer because of some other event, you can use erlang:cancel_timer(Tref). Note
that a time-out message cannot arrive after this, unless you have postponed it before (see the next section), so ensure
that you do not accidentally postpone such messages. Also note that a time-out message may have arrived just before
you cancelling it, so you may have to read out such a message from the process mailbox depending on the return value
from erlang:cancel_timer(Tref).
Another way to handle a late time-out can be to not cancel it, but to ignore it if it arrives in a state where it is known
to be late.
9.3.16 Postponing Events
If you want to ignore a particular event in the current state and handle it in a future state, you can postpone the event.
A postponed event is retried after the state has changed, that is, OldState =/= NewState.
302 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
Postponing is ordered by the state transition action postpone.
In this example, instead of ignoring button events while in the open state, we can postpone them and they are queued
and later handled in the locked state:
...
open(cast, {button,_}, Data) ->
{keep_state,Data,[postpone]};
...
Since a postponed event is only retried after a state change, you have to think about where to keep a state data item.
You can keep it in the server Data or in the State itself, for example by having two more or less identical states to
keep a boolean value, or by using a complex state with callback mode handle_event_function. If a change in
the value changes the set of events that is handled, then the value should be kept in the State. Otherwise no postponed
events will be retried since only the server Data changes.
This is not important if you do not postpone events. But if you later decide to start postponing some events, then the
design flaw of not having separate states when they should be, might become a hard to find bug.
Fuzzy State Diagrams
It is not uncommon that a state diagram does not specify how to handle events that are not illustrated in a particular
state in the diagram. Hopefully this is described in an associated text or from the context.
Possible actions: ignore as in drop the event (maybe log it) or deal with the event in some other state as in postpone it.
Selective Receive
Erlang's selective receive statement is often used to describe simple state machine examples in straightforward Erlang
code. The following is a possible implementation of the first example:
-module(code_lock).
-define(NAME, code_lock_1).
-export([start_link/1,button/1]).
start_link(Code) ->
spawn(
fun () ->
true = register(?NAME, self()),
do_lock(),
locked(Code, Code)
end).
button(Digit) ->
?NAME ! {button,Digit}.
locked(Code, [Digit|Remaining]) ->
receive
{button,Digit} when Remaining =:= [] ->
do_unlock(),
open(Code);
{button,Digit} ->
locked(Code, Remaining);
{button,_} ->
locked(Code, Code)
end.
open(Code) ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 303
9.3 gen_statem Behavior
receive
after 10000 ->
do_lock(),
locked(Code, Code)
end.
do_lock() ->
io:format("Locked~n", []).
do_unlock() ->
io:format("Open~n", []).
The selective receive in this case causes implicitly open to postpone any events to the locked state.
A selective receive cannot be used from a gen_statem behavior as for any gen_* behavior, as the receive statement
is within the gen_* engine itself. It must be there because all sys compatible behaviors must respond to system
messages and therefore do that in their engine receive loop, passing non-system messages to the callback module.
The state transition action postpone is designed to model selective receives. A selective receive implicitly postpones
any not received events, but the postpone state transition action explicitly postpones one received event.
Both mechanisms have the same theoretical time and memory complexity, while the selective receive language
construct has smaller constant factors.
9.3.17 State Entry Actions
Say you have a state machine specification that uses state entry actions. Allthough you can code this using selfgenerated events (described in the next section), especially if just one or a few states has got state entry actions, this
is a perfect use case for the built in state enter calls.
You return a list containing state_enter from your callback_mode/0 function and the gen_statem engine
will call your state callback once with the arguments (enter, OldState, ...) whenever the state changes.
Then you just need to handle these event-like calls in all states.
...
init(Code) ->
process_flag(trap_exit, true),
Data = #{code => Code},
{ok, locked, Data}.
callback_mode() ->
[state_functions,state_enter].
locked(enter, _OldState, Data) ->
do_lock(),
{keep_state,Data#{remaining => Code}};
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] ->
{next_state, open, Data};
...
open(enter, _OldState, _Data) ->
do_unlock(),
{keep_state_and_data, [{state_timeout,10000,lock}]};
open(state_timeout, lock, Data) ->
{next_state, locked, Data};
...
304 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
You can repeat the state entry code by returning one of {repeat_state,
...},
{repeat_state_and_data,_} or repeat_state_and_data that otherwise behaves exactly like their
keep_state siblings. See the type state_callback_result() in the reference manual.
9.3.18 Self-Generated Events
It can sometimes be beneficial to be able to generate events to your own state machine. This can be done with the state
transition action {next_event,EventType,EventContent}.
You can generate events of any existing type, but the internal type can only be generated through action
next_event. Hence, it cannot come from an external source, so you can be certain that an internal event is an
event from your state machine to itself.
One example for this is to pre-process incoming data, for example decrypting chunks or collecting characters up to
a line break. Purists may argue that this should be modelled with a separate state machine that sends pre-processed
events to the main state machine. But to decrease overhead the small pre-processing state machine can be implemented
in the common state event handling of the main state machine using a few state data variables that then sends the preprocessed events as internal events to the main state machine.
The following example uses an input model where you give the lock characters with put_chars(Chars) and then
call enter() to finish the input.
...
-export(put_chars/1, enter/0).
...
put_chars(Chars) when is_binary(Chars) ->
gen_statem:call(?NAME, {chars,Chars}).
enter() ->
gen_statem:call(?NAME, enter).
...
locked(enter, _OldState, Data) ->
do_lock(),
{keep_state,Data#{remaining => Code, buf => []}};
...
handle_event({call,From}, {chars,Chars}, #{buf := Buf} = Data) ->
{keep_state, Data#{buf := [Chars|Buf],
[{reply,From,ok}]};
handle_event({call,From}, enter, #{buf := Buf} = Data) ->
Chars = unicode:characters_to_binary(lists:reverse(Buf)),
try binary_to_integer(Chars) of
Digit ->
{keep_state, Data#{buf := []},
[{reply,From,ok},
{next_event,internal,{button,Chars}}]}
catch
error:badarg ->
{keep_state, Data#{buf := []},
[{reply,From,{error,not_an_integer}}]}
end;
...
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 305
9.3 gen_statem Behavior
If
you
start
this
program
with
code_lock:start([17])
you
can
unlock with code_lock:put_chars(<<"001">>),
code_lock:put_chars(<<"7">>),
code_lock:enter().
9.3.19 Example Revisited
This section includes the example after most of the mentioned modifications and some more using state enter calls,
which deserves a new state diagram:
306 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
Figure 3.2: Code Lock State Diagram
Revisited
Ericsson
AB. All Rights Reserved.: Erlang/OTP System Documentation | 307
9.3 gen_statem Behavior
Notice that this state diagram does not specify how to handle a button event in the state open. So, you need to read
somewhere else that unspecified events must be ignored as in not consumed but handled in some other state. Also, the
state diagram does not show that the code_length/0 call must be handled in every state.
Callback Mode: state_functions
Using state functions:
-module(code_lock).
-behaviour(gen_statem).
-define(NAME, code_lock_2).
-export([start_link/1,stop/0]).
-export([button/1,code_length/0]).
-export([init/1,callback_mode/0,terminate/3,code_change/4]).
-export([locked/3,open/3]).
start_link(Code) ->
gen_statem:start_link({local,?NAME}, ?MODULE, Code, []).
stop() ->
gen_statem:stop(?NAME).
button(Digit) ->
gen_statem:cast(?NAME, {button,Digit}).
code_length() ->
gen_statem:call(?NAME, code_length).
init(Code) ->
process_flag(trap_exit, true),
Data = #{code => Code},
{ok, locked, Data}.
callback_mode() ->
[state_functions,state_enter].
locked(enter, _OldState, #{code := Code} = Data) ->
do_lock(),
{keep_state, Data#{remaining => Code}};
locked(
timeout, _,
#{code := Code, remaining := Remaining} = Data) ->
{keep_state, Data#{remaining := Code}};
locked(
cast, {button,Digit},
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] -> % Complete
{next_state, open, Data};
[Digit|Rest] -> % Incomplete
{keep_state, Data#{remaining := Rest}, 30000};
[_|_] -> % Wrong
{keep_state, Data#{remaining := Code}}
end;
locked(EventType, EventContent, Data) ->
handle_event(EventType, EventContent, Data).
open(enter, _OldState, _Data) ->
do_unlock(),
{keep_state_and_data, [{state_timeout,10000,lock}]};
open(state_timeout, lock, Data) ->
{next_state, locked, Data};
open(cast, {button,_}, _) ->
{keep_state_and_data, [postpone]};
308 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
open(EventType, EventContent, Data) ->
handle_event(EventType, EventContent, Data).
handle_event({call,From}, code_length, #{code := Code}) ->
{keep_state_and_data, [{reply,From,length(Code)}]}.
do_lock() ->
io:format("Locked~n", []).
do_unlock() ->
io:format("Open~n", []).
terminate(_Reason, State, _Data) ->
State =/= locked andalso do_lock(),
ok.
code_change(_Vsn, State, Data, _Extra) ->
{ok,State,Data}.
Callback Mode: handle_event_function
This section describes what to change in the example to use one handle_event/4 function. The previously used
approach to first branch depending on event does not work that well here because of the state enter calls, so this
example first branches depending on state:
...
-export([handle_event/4]).
...
callback_mode() ->
[handle_event_function,state_enter].
%% State: locked
handle_event(
enter, _OldState, locked,
#{code := Code} = Data) ->
do_lock(),
{keep_state, Data#{remaining => Code}};
handle_event(
timeout, _, locked,
#{code := Code, remaining := Remaining} = Data) ->
{keep_state, Data#{remaining := Code}};
handle_event(
cast, {button,Digit}, locked,
#{code := Code, remaining := Remaining} = Data) ->
case Remaining of
[Digit] -> % Complete
{next_state, open, Data};
[Digit|Rest] -> % Incomplete
{keep_state, Data#{remaining := Rest}, 30000};
[_|_] -> % Wrong
{keep_state, Data#{remaining := Code}}
end;
%%
%% State: open
handle_event(enter, _OldState, open, _Data) ->
do_unlock(),
{keep_state_and_data, [{state_timeout,10000,lock}]};
handle_event(state_timeout, lock, open, Data) ->
{next_state, locked, Data};
handle_event(cast, {button,_}, open, _) ->
{keep_state_and_data,[postpone]};
%%
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 309
9.3 gen_statem Behavior
%% Any state
handle_event({call,From}, code_length, _State, #{code := Code}) ->
{keep_state_and_data, [{reply,From,length(Code)}]}.
...
Notice that postponing buttons from the locked state to the open state feels like a strange thing to do for a code
lock, but it at least illustrates event postponing.
9.3.20 Filter the State
The example servers so far in this chapter print the full internal state in the error log, for example, when killed by an exit
signal or because of an internal error. This state contains both the code lock code and which digits that remain to unlock.
This state data can be regarded as sensitive, and maybe not what you want in the error log because of some unpredictable
event.
Another reason to filter the state can be that the state is too large to print, as it fills the error log with uninteresting
details.
To avoid this, you can format the internal state that gets in the error log and gets returned from
sys:get_status/1,2 by implementing function Module:format_status/2, for example like this:
...
-export([init/1,terminate/3,code_change/4,format_status/2]).
...
format_status(Opt, [_PDict,State,Data]) ->
StateData =
{State,
maps:filter(
fun (code, _) -> false;
(remaining, _) -> false;
(_, _) -> true
end,
Data)},
case Opt of
terminate ->
StateData;
normal ->
[{data,[{"State",StateData}]}]
end.
It is not mandatory to implement a Module:format_status/2 function. If you do not, a default implementation
is used that does the same as this example function without filtering the Data term, that is, StateData =
{State,Data}, in this example containing sensitive information.
9.3.21 Complex State
The callback mode handle_event_function enables using a non-atom state as described in section Callback
Modes, for example, a complex state term like a tuple.
One reason to use this is when you have a state item that when changed should cancel the state time-out, or one
that affects the event handling in combination with postponing events. We will complicate the previous example by
introducing a configurable lock button (this is the state item in question), which in the open state immediately locks
the door, and an API function set_lock_button/1 to set the lock button.
310 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.3 gen_statem Behavior
Suppose now that we call set_lock_button while the door is open, and have already postponed a button event
that until now was not the lock button. The sensible thing can be to say that the button was pressed too early so it is not
to be recognized as the lock button. However, then it can be surprising that a button event that now is the lock button
event arrives (as retried postponed) immediately after the state transits to locked.
So we make the button/1 function synchronous by using gen_statem:call and still postpone its events in the
open state. Then a call to button/1 during the open state does not return until the state transits to locked, as
it is there the event is handled and the reply is sent.
If a process now calls set_lock_button/1 to change the lock button while another process hangs in button/1
with the new lock button, it can be expected that the hanging lock button call immediately takes effect and locks the
lock. Therefore, we make the current lock button a part of the state, so that when we change the lock button, the state
changes and all postponed events are retried.
We define the state as {StateName,LockButton}, where StateName is as before and LockButton is the
current lock button:
-module(code_lock).
-behaviour(gen_statem).
-define(NAME, code_lock_3).
-export([start_link/2,stop/0]).
-export([button/1,code_length/0,set_lock_button/1]).
-export([init/1,callback_mode/0,terminate/3,code_change/4,format_status/2]).
-export([handle_event/4]).
start_link(Code, LockButton) ->
gen_statem:start_link(
{local,?NAME}, ?MODULE, {Code,LockButton}, []).
stop() ->
gen_statem:stop(?NAME).
button(Digit) ->
gen_statem:call(?NAME, {button,Digit}).
code_length() ->
gen_statem:call(?NAME, code_length).
set_lock_button(LockButton) ->
gen_statem:call(?NAME, {set_lock_button,LockButton}).
init({Code,LockButton}) ->
process_flag(trap_exit, true),
Data = #{code => Code, remaining => undefined},
{ok, {locked,LockButton}, Data}.
callback_mode() ->
[handle_event_function,state_enter].
handle_event(
{call,From}, {set_lock_button,NewLockButton},
{StateName,OldLockButton}, Data) ->
{next_state, {StateName,NewLockButton}, Data,
[{reply,From,OldLockButton}]};
handle_event(
{call,From}, code_length,
{_StateName,_LockButton}, #{code := Code}) ->
{keep_state_and_data,
[{reply,From,length(Code)}]};
%%
%% State: locked
handle_event(
EventType, EventContent,
{locked,LockButton}, #{code := Code, remaining := Remaining} = Data) ->
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 311
9.3 gen_statem Behavior
case {EventType, EventContent} of
{enter, _OldState} ->
do_lock(),
{keep_state, Data#{remaining := Code}};
{timeout, _} ->
{keep_state, Data#{remaining := Code}};
{{call,From}, {button,Digit}} ->
case Remaining of
[Digit] -> % Complete
{next_state, {open,LockButton}, Data,
[{reply,From,ok}]};
[Digit|Rest] -> % Incomplete
{keep_state, Data#{remaining := Rest, 30000},
[{reply,From,ok}]};
[_|_] -> % Wrong
{keep_state, Data#{remaining := Code},
[{reply,From,ok}]}
end
end;
%%
%% State: open
handle_event(
EventType, EventContent,
{open,LockButton}, Data) ->
case {EventType, EventContent} of
{enter, _OldState} ->
do_unlock(),
{keep_state_and_data, [{state_timeout,10000,lock}]};
{state_timeout, lock} ->
{next_state, {locked,LockButton}, Data};
{{call,From}, {button,Digit}} ->
if
Digit =:= LockButton ->
{next_state, {locked,LockButton}, Data,
[{reply,From,locked}]};
true ->
{keep_state_and_data,
[postpone]}
end
end.
do_lock() ->
io:format("Locked~n", []).
do_unlock() ->
io:format("Open~n", []).
terminate(_Reason, State, _Data) ->
State =/= locked andalso do_lock(),
ok.
code_change(_Vsn, State, Data, _Extra) ->
{ok,State,Data}.
format_status(Opt, [_PDict,State,Data]) ->
StateData =
{State,
maps:filter(
fun (code, _) -> false;
(remaining, _) -> false;
(_, _) -> true
end,
Data)},
case Opt of
terminate ->
StateData;
normal ->
[{data,[{"State",StateData}]}]
312 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.4 gen_event Behaviour
end.
It can be an ill-fitting model for a physical code lock that the button/1 call can hang until the lock is locked. But
for an API in general it is not that strange.
9.3.22 Hibernation
If you have many servers in one node and they have some state(s) in their lifetime in which the servers can be expected
to idle for a while, and the amount of heap memory all these servers need is a problem, then the memory footprint of
a server can be mimimized by hibernating it through proc_lib:hibernate/3.
Note:
It is rather costly to hibernate a process; see erlang:hibernate/3. It is not something you want to do after
every event.
We can in this example hibernate in the {open,_} state, because what normally occurs in that state is that the state
time-out after a while triggers a transition to {locked,_}:
...
%% State: open
handle_event(
EventType, EventContent,
{open,LockButton}, Data) ->
case {EventType, EventContent} of
{enter, _OldState} ->
do_unlock(),
{keep_state_and_data,
[{state_timeout,10000,lock},hibernate]};
...
The atom hibernate in the action list on the last line when entering the {open,_} state is the only change. If any
event arrives in the {open,_}, state, we do not bother to rehibernate, so the server stays awake after any event.
To change that we would need to insert action hibernate in more places. For example, for the state-independent
set_lock_button and code_length operations that then would have to be aware of using hibernate while
in the {open,_} state, which would clutter the code.
Another not uncommon scenario is to use the event time-out to triger hibernation after a certain time of inactivity.
This server probably does not use heap memory worth hibernating for. To gain anything from hibernation, your server
would have to produce some garbage during callback execution, for which this example server can serve as a bad
example.
9.4 gen_event Behaviour
This section is to be read with the gen_event(3) manual page in STDLIB, where all interface functions and callback
functions are described in detail.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 313
9.4 gen_event Behaviour
9.4.1 Event Handling Principles
In OTP, an event manager is a named object to which events can be sent. An event can be, for example, an error,
an alarm, or some information that is to be logged.
In the event manager, zero, one, or many event handlers are installed. When the event manager is notified about an
event, the event is processed by all the installed event handlers. For example, an event manager for handling errors
can by default have a handler installed, which writes error messages to the terminal. If the error messages during a
certain period is to be saved to a file as well, the user adds another event handler that does this. When logging to the
file is no longer necessary, this event handler is deleted.
An event manager is implemented as a process and each event handler is implemented as a callback module.
The event manager essentially maintains a list of {Module, State} pairs, where each Module is an event handler,
and State is the internal state of that event handler.
9.4.2 Example
The callback module for the event handler writing error messages to the terminal can look as follows:
-module(terminal_logger).
-behaviour(gen_event).
-export([init/1, handle_event/2, terminate/2]).
init(_Args) ->
{ok, []}.
handle_event(ErrorMsg, State) ->
io:format("***Error*** ~p~n", [ErrorMsg]),
{ok, State}.
terminate(_Args, _State) ->
ok.
The callback module for the event handler writing error messages to a file can look as follows:
-module(file_logger).
-behaviour(gen_event).
-export([init/1, handle_event/2, terminate/2]).
init(File) ->
{ok, Fd} = file:open(File, read),
{ok, Fd}.
handle_event(ErrorMsg, Fd) ->
io:format(Fd, "***Error*** ~p~n", [ErrorMsg]),
{ok, Fd}.
terminate(_Args, Fd) ->
file:close(Fd).
The code is explained in the next sections.
9.4.3 Starting an Event Manager
To start an event manager for handling errors, as described in the previous example, call the following function:
314 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.4 gen_event Behaviour
gen_event:start_link({local, error_man})
This function spawns and links to a new process, an event manager.
The argument, {local,
error_man.
error_man} specifies the name. The event manager is then locally registered as
If the name is omitted, the event manager is not registered. Instead its pid must be used. The name can also be given
as {global, Name}, in which case the event manager is registered using global:register_name/2.
gen_event:start_link must be used if the event manager is part of a supervision tree, that is, started by a
supervisor. There is another function, gen_event:start, to start a standalone event manager, that is, an event
manager that is not part of a supervision tree.
9.4.4 Adding an Event Handler
The following example shows how to start an event manager and add an event handler to it by using the shell:
1> gen_event:start({local, error_man}).
{ok,<0.31.0>}
2> gen_event:add_handler(error_man, terminal_logger, []).
ok
This function sends a message to the event manager registered as error_man, telling it to add the event handler
terminal_logger. The event manager calls the callback function terminal_logger:init([]), where the
argument [] is the third argument to add_handler. init is expected to return {ok, State}, where State
is the internal state of the event handler.
init(_Args) ->
{ok, []}.
Here, init does not need any input data and ignores its argument. For terminal_logger, the internal state is not
used. For file_logger, the internal state is used to save the open file descriptor.
init(File) ->
{ok, Fd} = file:open(File, read),
{ok, Fd}.
9.4.5 Notifying about Events
3> gen_event:notify(error_man, no_reply).
***Error*** no_reply
ok
error_man is the name of the event manager and no_reply is the event.
The event is made into a message and sent to the event manager. When the event is received, the event manager calls
handle_event(Event, State) for each installed event handler, in the same order as they were added. The
function is expected to return a tuple {ok,State1}, where State1 is a new value for the state of the event handler.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 315
9.4 gen_event Behaviour
In terminal_logger:
handle_event(ErrorMsg, State) ->
io:format("***Error*** ~p~n", [ErrorMsg]),
{ok, State}.
In file_logger:
handle_event(ErrorMsg, Fd) ->
io:format(Fd, "***Error*** ~p~n", [ErrorMsg]),
{ok, Fd}.
9.4.6 Deleting an Event Handler
4> gen_event:delete_handler(error_man, terminal_logger, []).
ok
This function sends a message to the event manager registered as error_man, telling it to delete the event handler
terminal_logger. The event manager calls the callback function terminal_logger:terminate([],
State), where the argument [] is the third argument to delete_handler. terminate is to be the opposite of
init and do any necessary cleaning up. Its return value is ignored.
For terminal_logger, no cleaning up is necessary:
terminate(_Args, _State) ->
ok.
For file_logger, the file descriptor opened in init must be closed:
terminate(_Args, Fd) ->
file:close(Fd).
9.4.7 Stopping
When an event manager is stopped, it gives each of the installed event handlers the chance to clean up by calling
terminate/2, the same way as when deleting a handler.
In a Supervision Tree
If the event manager is part of a supervision tree, no stop function is needed. The event manager is automatically
terminated by its supervisor. Exactly how this is done is defined by a shutdown strategy set in the supervisor.
Standalone Event Managers
An event manager can also be stopped by calling:
> gen_event:stop(error_man).
316 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.5 Supervisor Behaviour
ok
9.4.8 Handling Other Messages
If the gen_event is to be able to receive other messages than events, the callback function handle_info(Info,
StateName, StateData) must be implemented to handle them. Examples of other messages are exit messages,
if the gen_event is linked to other processes (than the supervisor) and trapping exit signals.
handle_info({'EXIT', Pid, Reason}, State) ->
..code to handle exits here..
{ok, NewState}.
The code_change method must also be implemented.
code_change(OldVsn, State, Extra) ->
..code to convert state (and more) during code change
{ok, NewState}
9.5 Supervisor Behaviour
This section should be read with the supervisor(3) manual page in STDLIB, where all details about the supervisor
behaviour is given.
9.5.1 Supervision Principles
A supervisor is responsible for starting, stopping, and monitoring its child processes. The basic idea of a supervisor is
that it is to keep its child processes alive by restarting them when necessary.
Which child processes to start and monitor is specified by a list of child specifications. The child processes are started
in the order specified by this list, and terminated in the reversed order.
9.5.2 Example
The callback module for a supervisor starting the server from gen_server Behaviour can look as follows:
-module(ch_sup).
-behaviour(supervisor).
-export([start_link/0]).
-export([init/1]).
start_link() ->
supervisor:start_link(ch_sup, []).
init(_Args) ->
SupFlags = #{strategy => one_for_one, intensity => 1, period => 5},
ChildSpecs = [#{id => ch3,
start => {ch3, start_link, []},
restart => permanent,
shutdown => brutal_kill,
type => worker,
modules => [cg3]}],
{ok, {SupFlags, ChildSpecs}}.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 317
9.5 Supervisor Behaviour
The SupFlags variable in the return value from init/1 represents the supervisor flags.
The ChildSpecs variable in the return value from init/1 is a list of child specifications.
9.5.3 Supervisor Flags
This is the type definition for the supervisor flags:
sup_flags() = #{strategy => strategy(),
% optional
intensity => non_neg_integer(), % optional
period => pos_integer()}
% optional
strategy() = one_for_all
| one_for_one
| rest_for_one
| simple_one_for_one
•
•
strategy specifies the restart strategy.
intensity and period specify the maximum restart intensity.
9.5.4 Restart Strategy
The restart strategy is specified by the strategy key in the supervisor flags map returned by the callback function
init:
SupFlags = #{strategy => Strategy, ...}
The strategy key is optional in this map. If it is not given, it defaults to one_for_one.
one_for_one
If a child process terminates, only that process is restarted.
Figure 5.1: One_For_One Supervision
318 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.5 Supervisor Behaviour
one_for_all
If a child process terminates, all other child processes are terminated, and then all child processes, including the
terminated one, are restarted.
Figure 5.2: One_For_All Supervision
rest_for_one
If a child process terminates, the rest of the child processes (that is, the child processes after the terminated process in
start order) are terminated. Then the terminated child process and the rest of the child processes are restarted.
simple_one_for_one
See simple-one-for-one supervisors.
9.5.5 Maximum Restart Intensity
The supervisors have a built-in mechanism to limit the number of restarts which can occur in a given time interval.
This is specified by the two keys intensity and period in the supervisor flags map returned by the callback
function init:
SupFlags = #{intensity => MaxR, period => MaxT, ...}
If more than MaxR number of restarts occur in the last MaxT seconds, the supervisor terminates all the child processes
and then itself. The termination reason for the supervisor itself in that case will be shutdown.
When the supervisor terminates, then the next higher-level supervisor takes some action. It either restarts the terminated
supervisor or terminates itself.
The intention of the restart mechanism is to prevent a situation where a process repeatedly dies for the same reason,
only to be restarted again.
The keys intensity and period are optional in the supervisor flags map. If they are not given, they default to
1 and 5, respectively.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 319
9.5 Supervisor Behaviour
Tuning the intensity and period
The default values are 1 restart per 5 seconds. This was chosen to be safe for most systems, even with deep supervision
hierarchies, but you will probably want to tune the settings for your particular use case.
First, the intensity decides how big bursts of restarts you want to tolerate. For example, you might want to accept a
burst of at most 5 or 10 attempts, even within the same second, if it results in a successful restart.
Second, you need to consider the sustained failure rate, if crashes keep happening but not often enough to make the
supervisor give up. If you set intensity to 10 and set the period as low as 1, the supervisor will allow child processes
to keep restarting up to 10 times per second, forever, filling your logs with crash reports until someone intervenes
manually.
You should therefore set the period to be long enough that you can accept that the supervisor keeps going at that rate.
For example, if you have picked an intensity value of 5, then setting the period to 30 seconds will give you at most
one restart per 6 seconds for any longer period of time, which means that your logs won't fill up too quickly, and you
will have a chance to observe the failures and apply a fix.
These choices depend a lot on your problem domain. If you don't have real time monitoring and ability to fix problems
quickly, for example in an embedded system, you might want to accept at most one restart per minute before the
supervisor should give up and escalate to the next level to try to clear the error automatically. On the other hand, if
it is more important that you keep trying even at a high failure rate, you might want a sustained rate of as much as
1-2 restarts per second.
Avoiding common mistakes:
•
•
•
Do not forget to consider the burst rate. If you set intensity to 1 and period to 6, it gives the same sustained error
rate as 5/30 or 10/60, but will not allow even 2 restart attempts in quick succession. This is probably not what
you wanted.
Do not set the period to a very high value if you want to tolerate bursts. If you set intensity to 5 and period to
3600 (one hour), the supervisor will allow a short burst of 5 restarts, but then gives up if it sees another single
restart almost an hour later. You probably want to regard those crashes as separate incidents, so setting the period
to 5 or 10 minutes will be more reasonable.
If your application has multiple levels of supervision, then do not simply set the restart intensities to the same
values on all levels. Keep in mind that the total number of restarts (before the top level supervisor gives up and
terminates the application) will be the product of the intensity values of all the supervisors above the failing child
process.
For example, if the top level allows 10 restarts, and the next level also allows 10, a crashing child below that level
will be restarted 100 times, which is probably excessive. Allowing at most 3 restarts for the top level supervisor
might be a better choice in this case.
9.5.6 Child Specification
The type definition for a child specification is as follows:
child_spec() = #{id => child_id(),
% mandatory
start => mfargs(),
% mandatory
restart => restart(),
% optional
shutdown => shutdown(), % optional
type => worker(),
% optional
modules => modules()}
% optional
child_id() = term()
mfargs() = {M :: module(), F :: atom(), A :: [term()]}
modules() = [module()] | dynamic
restart() = permanent | transient | temporary
shutdown() = brutal_kill | timeout()
320 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.5 Supervisor Behaviour
worker() = worker | supervisor
•
id is used to identify the child specification internally by the supervisor.
The id key is mandatory.
•
Note that this identifier occasionally has been called "name". As far as possible, the terms "identifier" or "id"
are now used but in order to keep backwards compatibility, some occurences of "name" can still be found, for
example in error messages.
start defines the function call used to start the child process. It is a module-function-arguments tuple used as
apply(M, F, A).
It is to be (or result in) a call to any of the following:
•
•
•
•
•
•
The start key is mandatory.
restart defines when a terminated child process is to be restarted.
•
•
•
•
supervisor:start_link
gen_server:start_link
gen_statem:start_link
gen_event:start_link
A function compliant with these functions. For details, see the supervisor(3) manual page.
A permanent child process is always restarted.
A temporary child process is never restarted (not even when the supervisor restart strategy is
rest_for_one or one_for_all and a sibling death causes the temporary process to be terminated).
A transient child process is restarted only if it terminates abnormally, that is, with an exit reason other
than normal, shutdown, or {shutdown,Term}.
The restart key is optional. If it is not given, the default value permanent will be used.
shutdown defines how a child process is to be terminated.
•
•
•
brutal_kill means that the child process is unconditionally terminated using exit(Child,
kill).
An integer time-out value means that the supervisor tells the child process to terminate by calling
exit(Child, shutdown) and then waits for an exit signal back. If no exit signal is received within
the specified time, the child process is unconditionally terminated using exit(Child, kill).
If the child process is another supervisor, it is to be set to infinity to give the subtree enough time
to shut down. It is also allowed to set it to infinity, if the child process is a worker. See the warning
below:
Warning:
Be careful when setting the shutdown time to infinity when the child process is a worker. Because, in
this situation, the termination of the supervision tree depends on the child process; it must be implemented
in a safe way and its cleanup procedure must always return.
•
The shutdown key is optional. If it is not given, and the child is of type worker, the default value 5000 will
be used; if the child is of type supervisor, the default value infinity will be used.
type specifies if the child process is a supervisor or a worker.
The type key is optional. If it is not given, the default value worker will be used.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 321
9.5 Supervisor Behaviour
•
modules are to be a list with one element [Module], where Module is the name of the callback module, if
the child process is a supervisor, gen_server, gen_statem. If the child process is a gen_event, the value shall be
dynamic.
This information is used by the release handler during upgrades and downgrades, see Release Handling.
The modules key is optional. If it is not given, it defaults to [M], where M comes from the child's start {M,F,A}.
Example: The child specification to start the server ch3 in the previous example look as follows:
#{id => ch3,
start => {ch3, start_link, []},
restart => permanent,
shutdown => brutal_kill,
type => worker,
modules => [ch3]}
or simplified, relying on the default values:
#{id => ch3,
start => {ch3, start_link, []}
shutdown => brutal_kill}
Example: A child specification to start the event manager from the chapter about gen_event:
#{id => error_man,
start => {gen_event, start_link, [{local, error_man}]},
modules => dynamic}
Both server and event manager are registered processes which can be expected to be always accessible. Thus they
are specified to be permanent.
ch3 does not need to do any cleaning up before termination. Thus, no shutdown time is needed, but brutal_kill
is sufficient. error_man can need some time for the event handlers to clean up, thus the shutdown time is set to
5000 ms (which is the default value).
Example: A child specification to start another supervisor:
#{id => sup,
start => {sup, start_link, []},
restart => transient,
type => supervisor} % will cause default shutdown=>infinity
9.5.7 Starting a Supervisor
In the previous example, the supervisor is started by calling ch_sup:start_link():
start_link() ->
supervisor:start_link(ch_sup, []).
322 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.5 Supervisor Behaviour
ch_sup:start_link calls function supervisor:start_link/2, which spawns and links to a new process,
a supervisor.
•
•
The first argument, ch_sup, is the name of the callback module, that is, the module where the init callback
function is located.
The second argument, [], is a term that is passed as is to the callback function init. Here, init does not
need any indata and ignores the argument.
In this case, the supervisor is not registered. Instead its pid must be used. A name can be
specified by calling supervisor:start_link({local,
Name},
Module,
Args) or
supervisor:start_link({global, Name}, Module, Args).
The new supervisor process calls the callback function ch_sup:init([]). init shall return {ok,
{SupFlags, ChildSpecs}}:
init(_Args) ->
SupFlags = #{},
ChildSpecs = [#{id => ch3,
start => {ch3, start_link, []},
shutdown => brutal_kill}],
{ok, {SupFlags, ChildSpecs}}.
The supervisor then starts all its child processes according to the child specifications in the start specification. In this
case there is one child process, ch3.
supervisor:start_link is synchronous. It does not return until all child processes have been started.
9.5.8 Adding a Child Process
In addition to the static supervision tree, dynamic child processes can be added to an existing supervisor with the
following call:
supervisor:start_child(Sup, ChildSpec)
Sup is the pid, or name, of the supervisor. ChildSpec is a child specification.
Child processes added using start_child/2 behave in the same way as the other child processes, with the an
important exception: if a supervisor dies and is recreated, then all child processes that were dynamically added to the
supervisor are lost.
9.5.9 Stopping a Child Process
Any child process, static or dynamic, can be stopped in accordance with the shutdown specification:
supervisor:terminate_child(Sup, Id)
The child specification for a stopped child process is deleted with the following call:
supervisor:delete_child(Sup, Id)
Sup is the pid, or name, of the supervisor. Id is the value associated with the id key in the child specification.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 323
9.5 Supervisor Behaviour
As with dynamically added child processes, the effects of deleting a static child process is lost if the supervisor itself
restarts.
9.5.10 Simplified one_for_one Supervisors
A supervisor with restart strategy simple_one_for_one is a simplified one_for_one supervisor, where all
child processes are dynamically added instances of the same process.
The following is an example of a callback module for a simple_one_for_one supervisor:
-module(simple_sup).
-behaviour(supervisor).
-export([start_link/0]).
-export([init/1]).
start_link() ->
supervisor:start_link(simple_sup, []).
init(_Args) ->
SupFlags = #{strategy => simple_one_for_one,
intensity => 0,
period => 1},
ChildSpecs = [#{id => call,
start => {call, start_link, []},
shutdown => brutal_kill}],
{ok, {SupFlags, ChildSpecs}}.
When started, the supervisor does not start any child processes. Instead, all child processes are added dynamically
by calling:
supervisor:start_child(Sup, List)
Sup is the pid, or name, of the supervisor. List is an arbitrary list of terms, which are added to the list of arguments
specified in the child specification. If the start function is specified as {M, F, A}, the child process is started by
calling apply(M, F, A++List).
For example, adding a child to simple_sup above:
supervisor:start_child(Pid, [id1])
The result is that the child process is started by calling apply(call, start_link, []++[id1]), or actually:
call:start_link(id1)
A child under a simple_one_for_one supervisor can be terminated with the following:
supervisor:terminate_child(Sup, Pid)
Sup is the pid, or name, of the supervisor and Pid is the pid of the child.
324 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.6 sys and proc_lib
Because a simple_one_for_one supervisor can have many children, it shuts them all down asynchronously. This
means that the children will do their cleanup in parallel and therefore the order in which they are stopped is not defined.
9.5.11 Stopping
Since the supervisor is part of a supervision tree, it is automatically terminated by its supervisor. When asked to shut
down, it terminates all child processes in reversed start order according to the respective shutdown specifications, and
then terminates itself.
9.6 sys and proc_lib
The sys module has functions for simple debugging of processes implemented using behaviours. It also has functions
that, together with functions in the proc_lib module, can be used to implement a special process that complies to
the OTP design principles without using a standard behaviour. These functions can also be used to implement userdefined (non-standard) behaviours.
Both sys and proc_lib belong to the STDLIB application.
9.6.1 Simple Debugging
The sys module has functions for simple debugging of processes implemented using behaviours. The code_lock
example from gen_statem Behaviour is used to illustrate this:
Erlang/OTP 20 [DEVELOPMENT] [erts-9.0] [source-5ace45e] [64-bit] [smp:8:8] [ds:8:8:10] [async-threads:10] [hi
Eshell V9.0 (abort with ^G)
1> code_lock:start_link([1,2,3,4]).
Lock
{ok,<0.63.0>}
2> sys:statistics(code_lock, true).
ok
3> sys:trace(code_lock, true).
ok
4> code_lock:button(1).
*DBG* code_lock receive cast {button,1} in state locked
ok
*DBG* code_lock consume cast {button,1} in state locked
5> code_lock:button(2).
*DBG* code_lock receive cast {button,2} in state locked
ok
*DBG* code_lock consume cast {button,2} in state locked
6> code_lock:button(3).
*DBG* code_lock receive cast {button,3} in state locked
ok
*DBG* code_lock consume cast {button,3} in state locked
7> code_lock:button(4).
*DBG* code_lock receive cast {button,4} in state locked
ok
Unlock
*DBG* code_lock consume cast {button,4} in state locked
*DBG* code_lock receive state_timeout lock in state open
Lock
*DBG* code_lock consume state_timeout lock in state open
8> sys:statistics(code_lock, get).
{ok,[{start_time,{{2017,4,21},{16,8,7}}},
{current_time,{{2017,4,21},{16,9,42}}},
{reductions,2973},
{messages_in,5},
{messages_out,0}]}
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 325
9.6 sys and proc_lib
9> sys:statistics(code_lock, false).
ok
10> sys:trace(code_lock, false).
ok
11> sys:get_status(code_lock).
{status,<0.63.0>,
{module,gen_statem},
[[{'$initial_call',{code_lock,init,1}},
{'$ancestors',[<0.61.0>]}],
running,<0.61.0>,[],
[{header,"Status for state machine code_lock"},
{data,[{"Status",running},
{"Parent",<0.61.0>},
{"Logged Events",[]},
{"Postponed",[]}]},
{data,[{"State",
{locked,#{code => [1,2,3,4],remaining => [1,2,3,4]}}}]}]]}
9.6.2 Special Processes
This section describes how to write a process that complies to the OTP design principles, without using a standard
behaviour. Such a process is to:
•
•
•
Be started in a way that makes the process fit into a supervision tree
Support the sys debug facilities
Take care of system messages.
System messages are messages with a special meaning, used in the supervision tree. Typical system messages are
requests for trace output, and requests to suspend or resume process execution (used during release handling). Processes
implemented using standard behaviours automatically understand these messages.
Example
The simple server from Overview, implemented using sys and proc_lib so it fits into a supervision tree:
-module(ch4).
-export([start_link/0]).
-export([alloc/0, free/1]).
-export([init/1]).
-export([system_continue/3, system_terminate/4,
write_debug/3,
system_get_state/1, system_replace_state/2]).
start_link() ->
proc_lib:start_link(ch4, init, [self()]).
alloc() ->
ch4 ! {self(), alloc},
receive
{ch4, Res} ->
Res
end.
free(Ch) ->
ch4 ! {free, Ch},
ok.
init(Parent) ->
register(ch4, self()),
Chs = channels(),
326 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.6 sys and proc_lib
Deb = sys:debug_options([]),
proc_lib:init_ack(Parent, {ok, self()}),
loop(Chs, Parent, Deb).
loop(Chs, Parent, Deb) ->
receive
{From, alloc} ->
Deb2 = sys:handle_debug(Deb, fun ch4:write_debug/3,
ch4, {in, alloc, From}),
{Ch, Chs2} = alloc(Chs),
From ! {ch4, Ch},
Deb3 = sys:handle_debug(Deb2, fun ch4:write_debug/3,
ch4, {out, {ch4, Ch}, From}),
loop(Chs2, Parent, Deb3);
{free, Ch} ->
Deb2 = sys:handle_debug(Deb, fun ch4:write_debug/3,
ch4, {in, {free, Ch}}),
Chs2 = free(Ch, Chs),
loop(Chs2, Parent, Deb2);
{system, From, Request} ->
sys:handle_system_msg(Request, From, Parent,
ch4, Deb, Chs)
end.
system_continue(Parent, Deb, Chs) ->
loop(Chs, Parent, Deb).
system_terminate(Reason, _Parent, _Deb, _Chs) ->
exit(Reason).
system_get_state(Chs) ->
{ok, Chs}.
system_replace_state(StateFun, Chs) ->
NChs = StateFun(Chs),
{ok, NChs, NChs}.
write_debug(Dev, Event, Name) ->
io:format(Dev, "~p event = ~p~n", [Name, Event]).
Example on how the simple debugging functions in the sys module can also be used for ch4:
% erl
Erlang (BEAM) emulator version 5.2.3.6 [hipe] [threads:0]
Eshell V5.2.3.6 (abort with ^G)
1> ch4:start_link().
{ok,<0.30.0>}
2> sys:statistics(ch4, true).
ok
3> sys:trace(ch4, true).
ok
4> ch4:alloc().
ch4 event = {in,alloc,<0.25.0>}
ch4 event = {out,{ch4,ch1},<0.25.0>}
ch1
5> ch4:free(ch1).
ch4 event = {in,{free,ch1}}
ok
6> sys:statistics(ch4, get).
{ok,[{start_time,{{2003,6,13},{9,47,5}}},
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 327
9.6 sys and proc_lib
{current_time,{{2003,6,13},{9,47,56}}},
{reductions,109},
{messages_in,2},
{messages_out,1}]}
7> sys:statistics(ch4, false).
ok
8> sys:trace(ch4, false).
ok
9> sys:get_status(ch4).
{status,<0.30.0>,
{module,ch4},
[[{'$ancestors',[<0.25.0>]},{'$initial_call',{ch4,init,[<0.25.0>]}}],
running,<0.25.0>,[],
[ch1,ch2,ch3]]}
Starting the Process
A function in the proc_lib module is to be used to start the process. Several functions are available, for example,
spawn_link/3,4 for asynchronous start and start_link/3,4,5 for synchronous start.
A process started using one of these functions stores information (for example, about the ancestors and initial call)
that is needed for a process in a supervision tree.
If the process terminates with another reason than normal or shutdown, a crash report is generated. For more
information about the crash report, see the SASL User's Guide.
In the example, synchronous start is used. The process starts by calling ch4:start_link():
start_link() ->
proc_lib:start_link(ch4, init, [self()]).
ch4:start_link calls the function proc_lib:start_link. This function takes a module name, a function
name, and an argument list as arguments, spawns, and links to a new process. The new process starts by executing
the given function, here ch4:init(Pid), where Pid is the pid (self()) of the first process, which is the parent
process.
All initialization, including name registration, is done in init. The new process must also acknowledge that it has
been started to the parent:
init(Parent) ->
...
proc_lib:init_ack(Parent, {ok, self()}),
loop(...).
proc_lib:start_link is synchronous and does not return until proc_lib:init_ack has been called.
Debugging
To support the debug facilites in sys, a debug structure is needed. The Deb term is initialized using
sys:debug_options/1:
init(Parent) ->
...
Deb = sys:debug_options([]),
...
328 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.6 sys and proc_lib
loop(Chs, Parent, Deb).
sys:debug_options/1 takes a list of options as argument. Here the list is empty, which means no debugging is
enabled initially. For information about the possible options, see the sys(3) manual page in STDLIB.
Then, for each system event to be logged or traced, the following function is to be called.
sys:handle_debug(Deb, Func, Info, Event) => Deb1
Here:
•
•
•
•
Deb is the debug structure.
Func is a fun specifying a (user-defined) function used to format trace output. For each system event, the
format function is called as Func(Dev, Event, Info), where:
• Dev is the I/O device to which the output is to be printed. See the io(3) manual page in STDLIB.
• Event and Info are passed as is from handle_debug.
Info is used to pass more information to Func. It can be any term and is passed as is.
Event is the system event. It is up to the user to define what a system event is and how it is to be represented.
Typically at least incoming and outgoing messages are considered system events and represented by the tuples
{in,Msg[,From]} and {out,Msg,To}, respectively.
handle_debug returns an updated debug structure Deb1.
In the example, handle_debug is called for each incoming and outgoing message. The format function Func is
the function ch4:write_debug/3, which prints the message using io:format/3.
loop(Chs, Parent, Deb) ->
receive
{From, alloc} ->
Deb2 = sys:handle_debug(Deb, fun ch4:write_debug/3,
ch4, {in, alloc, From}),
{Ch, Chs2} = alloc(Chs),
From ! {ch4, Ch},
Deb3 = sys:handle_debug(Deb2, fun ch4:write_debug/3,
ch4, {out, {ch4, Ch}, From}),
loop(Chs2, Parent, Deb3);
{free, Ch} ->
Deb2 = sys:handle_debug(Deb, fun ch4:write_debug/3,
ch4, {in, {free, Ch}}),
Chs2 = free(Ch, Chs),
loop(Chs2, Parent, Deb2);
...
end.
write_debug(Dev, Event, Name) ->
io:format(Dev, "~p event = ~p~n", [Name, Event]).
Handling System Messages
System messages are received as:
{system, From, Request}
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 329
9.6 sys and proc_lib
The content and meaning of these messages do not need to be interpreted by the process. Instead the following function
is to be called:
sys:handle_system_msg(Request, From, Parent, Module, Deb, State)
This function does not return. It handles the system message and then either calls the following if process execution
is to continue:
Module:system_continue(Parent, Deb, State)
Or calls the following if the process is to terminate:
Module:system_terminate(Reason, Parent, Deb, State)
A process in a supervision tree is expected to terminate with the same reason as its parent.
•
•
•
•
•
Request and From are to be passed as is from the system message to the call to handle_system_msg.
Parent is the pid of the parent.
Module is the name of the module.
Deb is the debug structure.
State is a term describing the internal state and is passed to system_continue/system_terminate/
system_get_state/system_replace_state.
If the process is to return its state, handle_system_msg calls:
Module:system_get_state(State)
If the process is to replace its state using the fun StateFun, handle_system_msg calls:
Module:system_replace_state(StateFun, State)
In the example:
loop(Chs, Parent, Deb) ->
receive
...
{system, From, Request} ->
sys:handle_system_msg(Request, From, Parent,
ch4, Deb, Chs)
end.
system_continue(Parent, Deb, Chs) ->
loop(Chs, Parent, Deb).
system_terminate(Reason, Parent, Deb, Chs) ->
exit(Reason).
330 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.6 sys and proc_lib
system_get_state(Chs) ->
{ok, Chs, Chs}.
system_replace_state(StateFun, Chs) ->
NChs = StateFun(Chs),
{ok, NChs, NChs}.
If the special process is set to trap exits and if the parent process terminates, the expected behavior is to terminate
with the same reason:
init(...) ->
...,
process_flag(trap_exit, true),
...,
loop(...).
loop(...) ->
receive
...
{'EXIT', Parent, Reason} ->
..maybe some cleaning up here..
exit(Reason);
...
end.
9.6.3 User-Defined Behaviours
To implement a user-defined behaviour, write code similar to code for a special process, but call functions in a callback
module for handling specific tasks.
If the compiler is to warn for missing callback functions, as it does for the OTP behaviours, add -callback attributes
in the behaviour module to describe the expected callbacks:
-callback Name1(Arg1_1, Arg1_2, ..., Arg1_N1) -> Res1.
-callback Name2(Arg2_1, Arg2_2, ..., Arg2_N2) -> Res2.
...
-callback NameM(ArgM_1, ArgM_2, ..., ArgM_NM) -> ResM.
NameX are the names of the expected callbacks. ArgX_Y and ResX are types as they are described in Types and
Function Specifications. The whole syntax of the -spec attribute is supported by the -callback attribute.
Callback functions that are optional for the user of the behaviour to implement are specified by use of the optional_callbacks attribute:
-optional_callbacks([OptName1/OptArity1, ..., OptNameK/OptArityK]).
where each OptName/OptArity specifies the name and arity of a callback function. Note that the optional_callbacks attribute is to be used together with the -callback attribute; it cannot be combined with
the behaviour_info() function described below.
Tools
that
need
to
know
about
optional
callback
functions
can
call
Behaviour:behaviour_info(optional_callbacks) to get a list of all optional callback functions.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 331
9.6 sys and proc_lib
Note:
We recommend using the -callback attribute rather than the behaviour_info() function. The reason is
that the extra type information can be used by tools to produce documentation or find discrepancies.
As an alternative to the -callback and -optional_callbacks attributes you may directly implement and
export behaviour_info():
behaviour_info(callbacks) ->
[{Name1, Arity1},...,{NameN, ArityN}].
where each {Name, Arity} specifies the name and arity of a callback function. This function is otherwise
automatically generated by the compiler using the -callback attributes.
When the compiler encounters the module attribute -behaviour(Behaviour). in a module Mod, it calls
Behaviour:behaviour_info(callbacks) and compares the result with the set of functions actually
exported from Mod, and issues a warning if any callback function is missing.
Example:
%% User-defined behaviour module
-module(simple_server).
-export([start_link/2, init/3, ...]).
-callback
-callback
-callback
-callback
init(State :: term()) -> 'ok'.
handle_req(Req :: term(), State :: term()) -> {'ok', Reply :: term()}.
terminate() -> 'ok'.
format_state(State :: term()) -> term().
-optional_callbacks([format_state/1]).
%% Alternatively you may define:
%%
%% -export([behaviour_info/1]).
%% behaviour_info(callbacks) ->
%%
[{init,1},
%%
{handle_req,2},
%%
{terminate,0}].
start_link(Name, Module) ->
proc_lib:start_link(?MODULE, init, [self(), Name, Module]).
init(Parent, Name, Module) ->
register(Name, self()),
...,
Dbg = sys:debug_options([]),
proc_lib:init_ack(Parent, {ok, self()}),
loop(Parent, Module, Deb, ...).
...
In a callback module:
-module(db).
-behaviour(simple_server).
332 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.7 Applications
-export([init/1, handle_req/2, terminate/0]).
...
The contracts specified with -callback attributes in behaviour modules can be further refined by adding -spec
attributes in callback modules. This can be useful as -callback contracts are usually generic. The same callback
module with contracts for the callbacks:
-module(db).
-behaviour(simple_server).
-export([init/1, handle_req/2, terminate/0]).
-record(state, {field1 :: [atom()], field2 :: integer()}).
-type state()
:: #state{}.
-type request() :: {'store', term(), term()};
{'lookup', term()}.
...
-spec handle_req(request(), state()) -> {'ok', term()}.
...
Each -spec contract is to be a subtype of the respective -callback contract.
9.7 Applications
This section is to be read with the app(4) and application(3) manual pages in Kernel.
9.7.1 Application Concept
When you have written code implementing some specific functionality you might want to make the code into an
application, that is, a component that can be started and stopped as a unit, and which can also be reused in other
systems.
To do this, create an application callback module, and describe how the application is to be started and stopped.
Then, an application specification is needed, which is put in an application resource file. Among other things, this
file specifies which modules the application consists of and the name of the callback module.
If you use systools, the Erlang/OTP tools for packaging code (see Releases), the code for each application is placed
in a separate directory following a pre-defined directory structure.
9.7.2 Application Callback Module
How to start and stop the code for the application, that is, the supervision tree, is described by two callback functions:
start(StartType, StartArgs) -> {ok, Pid} | {ok, Pid, State}
stop(State)
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 333
9.7 Applications
•
•
•
•
start is called when starting the application and is to create the supervision tree by starting the top supervisor.
It is expected to return the pid of the top supervisor and an optional term, State, which defaults to []. This
term is passed as is to stop.
StartType is usually the atom normal. It has other values only in the case of a takeover or failover, see
Distributed Applications.
StartArgs is defined by the key mod in the application resource file.
stop/1 is called after the application has been stopped and is to do any necessary cleaning up. The actual
stopping of the application, that is, the shutdown of the supervision tree, is handled automatically as described
in Starting and Stopping Applications.
Example of an application callback module for packaging the supervision tree from Supervisor Behaviour:
-module(ch_app).
-behaviour(application).
-export([start/2, stop/1]).
start(_Type, _Args) ->
ch_sup:start_link().
stop(_State) ->
ok.
A library application that cannot be started or stopped, does not need any application callback module.
9.7.3 Application Resource File
To define an application, an application specification is created, which is put in an application resource file, or in
short an .app file:
{application, Application, [Opt1,...,OptN]}.
•
•
Application, an atom, is the name of the application. The file must be named Application.app.
Each Opt is a tuple {Key,Value}, which define a certain property of the application. All keys are optional.
Default values are used for any omitted keys.
The contents of a minimal .app file for a library application libapp looks as follows:
{application, libapp, []}.
The contents of a minimal .app file ch_app.app for a supervision tree application like ch_app looks as follows:
{application, ch_app,
[{mod, {ch_app,[]}}]}.
The key mod defines the callback module and start argument of the application, in this case ch_app and [],
respectively. This means that the following is called when the application is to be started:
334 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.7 Applications
ch_app:start(normal, [])
The following is called when the application is stopped.
ch_app:stop([])
When using systools, the Erlang/OTP tools for packaging code (see Section Releases), the keys description,
vsn, modules, registered, and applications are also to be specified:
{application, ch_app,
[{description, "Channel allocator"},
{vsn, "1"},
{modules, [ch_app, ch_sup, ch3]},
{registered, [ch3]},
{applications, [kernel, stdlib, sasl]},
{mod, {ch_app,[]}}
]}.
•
•
•
•
•
description - A short description, a string. Defaults to "".
vsn - Version number, a string. Defaults to "".
modules - All modules introduced by this application. systools uses this list when generating boot scripts
and tar files. A module must be defined in only one application. Defaults to [].
registered - All names of registered processes in the application. systools uses this list to detect name
clashes between applications. Defaults to [].
applications - All applications that must be started before this application is started. systools uses this
list to generate correct boot scripts. Defaults to []. Notice that all applications have dependencies to at least
Kernel and STDLIB.
Note:
For details about the syntax and contents of the application resource file, see the app manual page in Kernel.
9.7.4 Directory Structure
When packaging code using systools, the code for each application is placed in a separate directory, lib/
Application-Vsn, where Vsn is the version number.
This can be useful to know, even if systools is not used, since Erlang/OTP is packaged according to the OTP
principles and thus comes with a specific directory structure. The code server (see the code(3) manual page in
Kernel) automatically uses code from the directory with the highest version number, if more than one version of an
application is present.
Directory Structure guidelines for a Development Environment
Any directory structure for development will suffice as long as the released directory structure adhere to the description
below, but it is encouraged that the same directory structure also be used in a development environment. The version
number should be omitted from the application directory name since this is an artifact of the release step.
Some sub-directories are required. Some sub-directories are optional, meaning that it should only be used if the
application itself requires it. Finally, some sub-directories are recommended, meaning it is encouraged that it is used
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 335
9.7 Applications
and used as described here. For example, both documentation and tests are encouraged to exist in an application for
it to be deemed a proper OTP application.
─ ${application}
├── doc
│ ├── internal
│ ├── examples
│ └── src
├── include
├── priv
├── src
│ └── ${application}.app.src
└── test
•
•
•
•
•
•
•
•
src - Required. Contains the Erlang source code, the source of the .app file and internal include files used by
the application itself. Additional sub-directories within src can be used as namespaces to organize source files.
These directories should never be deeper than one level.
priv - Optional. Used for application specific files.
include - Optional. Used for public include files that must be reachable from other applications.
doc - Recommended. Any source documentation should be placed in sub-directories here.
doc/internal - Recommended. Any documentation that describes implementation details about this
application, not intended for publication, should be placed here.
doc/examples - Recommended. Source code for examples on how to use this application should be placed
here. It is encouraged that examples are sourced to the public documentation from this directory.
doc/src - Recommended. All source files for documentation, such as Markdown, AsciiDoc or XML-files,
should be placed here.
test - Recommended. All files regarding tests, such as test suites and test specifications, should be placed
here.
Other directories in the development environment may be needed. If source code from languages other than Erlang is
used, for instance C-code for NIFs, that code should be placed in a separate directory. By convention it is recommended
to prefix such directories with the language name, for example c_src for C, java_src for Java or go_src for
Go. Directories with _src suffix indicates that it is a part of the application and the compilation step. The final build
artifacts should target the priv/lib or priv/bin directories.
The priv directory holds assets that the application needs during runtime. Executables should reside in priv/bin
and dynamically-linked libraries should reside in priv/lib. Other assets are free to reside within the priv directory
but it is recommended it does so in a structured manner.
Source files from other languages that generate Erlang code, such as ASN.1 or Mibs, should be placed in directories,
at the top level or in src, with the same name as the source language, for example asn1 and mibs. Build artifacts
should be placed in their respective language directory, such as src for Erlang code or java_src for Java code.
The .app file for release may reside in the ebin-directory in a development environment but it is encouraged that
this is an artifact of the build step. By convention a .app.src file is used, which resides in the src directory. This
file is nearly identical as the .app file but certain fields may be replaced during the build step, such as the application
version.
Directory names should not be capitalized.
It is encouraged to omit empty directories.
Directory Structure for a Released System
A released application must follow a certain structure.
336 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.7 Applications
─ ${application}-${version}
├── bin
├── doc
│ ├── html
│ ├── man[1-9]
│ ├── pdf
│ ├── internal
│ └── examples
├── ebin
│ └── ${application}.app
├── include
├── priv
│ ├── lib
│ └── bin
└── src
•
•
•
•
•
•
•
•
•
•
•
•
•
src - Optional. Contains the Erlang source code and internal include files used by the application itself. This
directory is no longer required in a released application.
ebin - Required. Contains the Erlang object code, the beam files. The .app file must also be placed here.
priv - Optional. Used for application specific files. code:priv_dir/1 is to be used to access this
directory.
priv/lib - Recommended. Any shared-object files that are used by the application, such as NIFs or linkedin-drivers, should be placed here.
priv/bin - Recommended. Any executable that is used by the application, such as port-programs, should be
placed here.
include - Optional. Used for public include files that must be reachable from other applications.
bin - Optional. Any executable that is a product of the application, such as escripts or shell-scripts, should be
placed here.
doc - Optional. Any released documentation should be placed in sub-directories here.
doc/man1 - Recommended. Man pages for Application executables.
doc/man3 - Recommended. Man pages for module APIs.
doc/man6 - Recommended. Man pages for Application overview.
doc/html - Optional. HTML pages for the entire Application.
doc/pdf - Optional. PDF documentation for the entire Application.
The src directory could be useful to release for debugging purposes but is not required. The include directory
should only be released if the applications has public include files.
The only documentation that is recommended to be released in this way are the man pages. HTML and PDF will
normally be distributed in some other manner.
It is encouraged to omit empty directories.
9.7.5 Application Controller
When an Erlang runtime system is started, a number of processes are started as part of the Kernel application. One of
these processes is the application controller process, registered as application_controller.
All operations on applications are coordinated by the application controller. It is interacted through the functions in
the module application, see the application(3) manual page in Kernel. In particular, applications can be
loaded, unloaded, started, and stopped.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 337
9.7 Applications
9.7.6 Loading and Unloading Applications
Before an application can be started, it must be loaded. The application controller reads and stores the information
from the .app file:
1> application:load(ch_app).
ok
2> application:loaded_applications().
[{kernel,"ERTS CXC 138 10","2.8.1.3"},
{stdlib,"ERTS CXC 138 10","1.11.4.3"},
{ch_app,"Channel allocator","1"}]
An application that has been stopped, or has never been started, can be unloaded. The information about the application
is erased from the internal database of the application controller.
3> application:unload(ch_app).
ok
4> application:loaded_applications().
[{kernel,"ERTS CXC 138 10","2.8.1.3"},
{stdlib,"ERTS CXC 138 10","1.11.4.3"}]
Note:
Loading/unloading an application does not load/unload the code used by the application. Code loading is done
the usual way.
9.7.7 Starting and Stopping Applications
An application is started by calling:
5> application:start(ch_app).
ok
6> application:which_applications().
[{kernel,"ERTS CXC 138 10","2.8.1.3"},
{stdlib,"ERTS CXC 138 10","1.11.4.3"},
{ch_app,"Channel allocator","1"}]
If the application is not already loaded, the application controller first loads it using application:load/1. It
checks the value of the applications key, to ensure that all applications that are to be started before this application
are running.
The application controller then creates an application master for the application. The application master is the group
leader of all the processes in the application. The application master starts the application by calling the application
callback function start/2 in the module, and with the start argument, defined by the mod key in the .app file.
An application is stopped, but not unloaded, by calling:
7> application:stop(ch_app).
ok
338 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.7 Applications
The application master stops the application by telling the top supervisor to shut down. The top supervisor tells all its
child processes to shut down, and so on; the entire tree is terminated in reversed start order. The application master
then calls the application callback function stop/1 in the module defined by the mod key.
9.7.8 Configuring an Application
An application can be configured using configuration parameters. These are a list of {Par,Val} tuples specified
by a key env in the .app file:
{application, ch_app,
[{description, "Channel allocator"},
{vsn, "1"},
{modules, [ch_app, ch_sup, ch3]},
{registered, [ch3]},
{applications, [kernel, stdlib, sasl]},
{mod, {ch_app,[]}},
{env, [{file, "/usr/local/log"}]}
]}.
Par is to be an atom. Val is any term. The application can retrieve the value of a configuration parameter by calling
application:get_env(App, Par) or a number of similar functions, see the application(3) manual
page in Kernel.
Example:
% erl
Erlang (BEAM) emulator version 5.2.3.6 [hipe] [threads:0]
Eshell V5.2.3.6 (abort with ^G)
1> application:start(ch_app).
ok
2> application:get_env(ch_app, file).
{ok,"/usr/local/log"}
The values in the .app file can be overridden by values in a system configuration file. This is a file that contains
configuration parameters for relevant applications:
[{Application1, [{Par11,Val11},...]},
...,
{ApplicationN, [{ParN1,ValN1},...]}].
The system configuration is to be called Name.config and Erlang is to be started with the command-line argument
-config Name. For details, see the config(4) manual page in Kernel.
Example:
A file test.config is created with the following contents:
[{ch_app, [{file, "testlog"}]}].
The value of file overrides the value of file as defined in the .app file:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 339
9.7 Applications
% erl -config test
Erlang (BEAM) emulator version 5.2.3.6 [hipe] [threads:0]
Eshell V5.2.3.6 (abort with ^G)
1> application:start(ch_app).
ok
2> application:get_env(ch_app, file).
{ok,"testlog"}
If release handling is used, exactly one system configuration file is to be used and that file is to be called sys.config.
The values in the .app file and the values in a system configuration file can be overridden directly from the command
line:
% erl -ApplName Par1 Val1 ... ParN ValN
Example:
% erl -ch_app file '"testlog"'
Erlang (BEAM) emulator version 5.2.3.6 [hipe] [threads:0]
Eshell V5.2.3.6 (abort with ^G)
1> application:start(ch_app).
ok
2> application:get_env(ch_app, file).
{ok,"testlog"}
9.7.9 Application Start Types
A start type is defined when starting the application:
application:start(Application, Type)
application:start(Application) is the same as calling application:start(Application,
temporary). The type can also be permanent or transient:
•
•
•
If a permanent application terminates, all other applications and the runtime system are also terminated.
If a transient application terminates with reason normal, this is reported but no other applications are
terminated. If a transient application terminates abnormally, that is with any other reason than normal, all
other applications and the runtime system are also terminated.
If a temporary application terminates, this is reported but no other applications are terminated.
An application can always be stopped explicitly by calling application:stop/1. Regardless of the mode, no
other applications are affected.
The transient mode is of little practical use, since when a supervision tree terminates, the reason is set to shutdown,
not normal.
340 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.8 Included Applications
9.8 Included Applications
9.8.1 Introduction
An application can include other applications. An included application has its own application directory and .app
file, but it is started as part of the supervisor tree of another application.
An application can only be included by one other application.
An included application can include other applications.
An application that is not included by any other application is called a primary application.
Figure 8.1: Primary Application and Included Applications
The application controller automatically loads any included applications when loading a primary application, but does
not start them. Instead, the top supervisor of the included application must be started by a supervisor in the including
application.
This means that when running, an included application is in fact part of the primary application, and a process in an
included application considers itself belonging to the primary application.
9.8.2 Specifying Included Applications
Which applications to include is defined by the included_applications key in the .app file:
{application, prim_app,
[{description, "Tree application"},
{vsn, "1"},
{modules, [prim_app_cb, prim_app_sup, prim_app_server]},
{registered, [prim_app_server]},
{included_applications, [incl_app]},
{applications, [kernel, stdlib, sasl]},
{mod, {prim_app_cb,[]}},
{env, [{file, "/usr/local/log"}]}
]}.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 341
9.8 Included Applications
9.8.3 Synchronizing Processes during Startup
The supervisor tree of an included application is started as part of the supervisor tree of the including application. If
there is a need for synchronization between processes in the including and included applications, this can be achieved
by using start phases.
Start phases are defined by the start_phases key in the .app file as a list of tuples {Phase,PhaseArgs},
where Phase is an atom and PhaseArgs is a term.
The value of the mod key of the including application must be set to {application_starter,
[Module,StartArgs]}, where Module as usual is the application callback module. StartArgs is a term
provided as argument to the callback function Module:start/2:
{application, prim_app,
[{description, "Tree application"},
{vsn, "1"},
{modules, [prim_app_cb, prim_app_sup, prim_app_server]},
{registered, [prim_app_server]},
{included_applications, [incl_app]},
{start_phases, [{init,[]}, {go,[]}]},
{applications, [kernel, stdlib, sasl]},
{mod, {application_starter,[prim_app_cb,[]]}},
{env, [{file, "/usr/local/log"}]}
]}.
{application, incl_app,
[{description, "Included application"},
{vsn, "1"},
{modules, [incl_app_cb, incl_app_sup, incl_app_server]},
{registered, []},
{start_phases, [{go,[]}]},
{applications, [kernel, stdlib, sasl]},
{mod, {incl_app_cb,[]}}
]}.
When starting a primary application with included applications, the primary application is started the normal way,
that is:
•
•
The application controller creates an application master for the application
The application master calls Module:start(normal, StartArgs) to start the top supervisor.
Then, for the primary application and each included application in top-down, left-to-right order, the application master
calls Module:start_phase(Phase, Type, PhaseArgs) for each phase defined for the primary application,
in that order. If a phase is not defined for an included application, the function is not called for this phase and
application.
The following requirements apply to the .app file for an included application:
•
•
•
The {mod, {Module,StartArgs}} option must be included. This option is used to find the callback
module Module of the application. StartArgs is ignored, as Module:start/2 is called only for the
primary application.
If the included application itself contains included applications, instead the {mod,
{application_starter, [Module,StartArgs]}} option must be included.
The {start_phases, [{Phase,PhaseArgs}]} option must be included, and the set of specified
phases must be a subset of the set of phases specified for the primary application.
When starting prim_app as defined above, the application controller calls the following callback functions before
application:start(prim_app) returns a value:
342 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.9 Distributed Applications
application:start(prim_app)
=> prim_app_cb:start(normal, [])
=> prim_app_cb:start_phase(init, normal, [])
=> prim_app_cb:start_phase(go, normal, [])
=> incl_app_cb:start_phase(go, normal, [])
ok
9.9 Distributed Applications
9.9.1 Introduction
In a distributed system with several Erlang nodes, it can be necessary to control applications in a distributed manner.
If the node, where a certain application is running, goes down, the application is to be restarted at another node.
Such an application is called a distributed application. Notice that it is the control of the application that is distributed.
All applications can be distributed in the sense that they, for example, use services on other nodes.
Since a distributed application can move between nodes, some addressing mechanism is required to ensure that it can
be addressed by other applications, regardless on which node it currently executes. This issue is not addressed here,
but the global or pg2 modules in Kernel can be used for this purpose.
9.9.2 Specifying Distributed Applications
Distributed applications are controlled by both the application controller and a distributed application controller
process, dist_ac. Both these processes are part of the Kernel application. Distributed applications are thus specified
by configuring the Kernel application, using the following configuration parameter (see also kernel(6)):
distributed = [{Application, [Timeout,] NodeDesc}]
•
•
•
Specifies where the application Application = atom() can execute.
>NodeDesc = [Node | {Node,...,Node}] is a list of node names in priority order. The order
between nodes in a tuple is undefined.
Timeout = integer() specifies how many milliseconds to wait before restarting the application at
another node. It defaults to 0.
For distribution of application control to work properly, the nodes where a distributed application can run must contact
each other and negotiate where to start the application. This is done using the following configuration parameters in
Kernel:
•
•
•
sync_nodes_mandatory = [Node] - Specifies which other nodes must be started (within the time-out
specified by sync_nodes_timeout).
sync_nodes_optional = [Node] - Specifies which other nodes can be started (within the time-out
specified by sync_nodes_timeout).
sync_nodes_timeout = integer() | infinity - Specifies how many milliseconds to wait for the
other nodes to start.
When started, the node waits for all nodes specified by sync_nodes_mandatory and sync_nodes_optional
to come up. When all nodes are up, or when all mandatory nodes are up and the time specified by
sync_nodes_timeout has elapsed, all applications start. If not all mandatory nodes are up, the node terminates.
Example:
An application myapp is to run at the node cp1@cave. If this node goes down, myapp is to be restarted at cp2@cave
or cp3@cave. A system configuration file cp1.config for cp1@cave can look as follows:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 343
9.9 Distributed Applications
[{kernel,
[{distributed, [{myapp, 5000, [cp1@cave, {cp2@cave, cp3@cave}]}]},
{sync_nodes_mandatory, [cp2@cave, cp3@cave]},
{sync_nodes_timeout, 5000}
]
}
].
The system configuration files for cp2@cave and cp3@cave are identical, except for the list of mandatory nodes,
which is to be [cp1@cave, cp3@cave] for cp2@cave and [cp1@cave, cp2@cave] for cp3@cave.
Note:
All involved nodes must have the same value for distributed and sync_nodes_timeout. Otherwise the
system behaviour is undefined.
9.9.3 Starting and Stopping Distributed Applications
When all involved (mandatory) nodes have been started, the distributed application can be started by calling
application:start(Application) at all of these nodes.
A boot script (see Releases) can be used that automatically starts the application.
The application is started at the first operational node that is listed in the list of nodes in the distributed
configuration parameter. The application is started as usual. That is, an application master is created and calls the
application callback function:
Module:start(normal, StartArgs)
Example:
Continuing the example from the previous section, the three nodes are started, specifying the system configuration file:
> erl -sname cp1 -config cp1
> erl -sname cp2 -config cp2
> erl -sname cp3 -config cp3
When all nodes are operational, myapp can be started. This is achieved by calling application:start(myapp)
at all three nodes. It is then started at cp1, as shown in the following figure:
Figure 9.1: Application myapp - Situation 1
Similarly, the application must be stopped by calling application:stop(Application) at all involved nodes.
344 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.9 Distributed Applications
9.9.4 Failover
If the node where the application is running goes down, the application is restarted (after the specified time-out) at
the first operational node that is listed in the list of nodes in the distributed configuration parameter. This is
called a failover.
The application is started the normal way at the new node, that is, by the application master calling:
Module:start(normal, StartArgs)
An exception is if the application has the start_phases key defined (see Included Applications). The application
is then instead started by calling:
Module:start({failover, Node}, StartArgs)
Here Node is the terminated node.
Example:
If cp1 goes down, the system checks which one of the other nodes, cp2 or cp3, has the least number of running
applications, but waits for 5 seconds for cp1 to restart. If cp1 does not restart and cp2 runs fewer applications than
cp3, myapp is restarted on cp2.
Figure 9.2: Application myapp - Situation 2
Suppose now that cp2 goes also down and does not restart within 5 seconds. myapp is now restarted on cp3.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 345
9.9 Distributed Applications
Figure 9.3: Application myapp - Situation 3
9.9.5 Takeover
If a node is started, which has higher priority according to distributed than the node where a distributed
application is running, the application is restarted at the new node and stopped at the old node. This is called a takeover.
The application is started by the application master calling:
Module:start({takeover, Node}, StartArgs)
Here Node is the old node.
Example:
If myapp is running at cp3, and if cp2 now restarts, it does not restart myapp, as the order between the cp2 and
cp3 nodes is undefined.
Figure 9.4: Application myapp - Situation 4
346 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.10 Releases
However, if cp1 also restarts, the function application:takeover/2 moves myapp to cp1, as cp1 has
a higher priority than cp3 for this application. In this case, Module:start({takeover, cp3@cave},
StartArgs) is executed at cp1 to start the application.
Figure 9.5: Application myapp - Situation 5
9.10 Releases
This section is to be read with the rel(4), systools(3), and script(4) manual pages in SASL.
9.10.1 Release Concept
When you have written one or more applications, you might want to create a complete system with these applications
and a subset of the Erlang/OTP applications. This is called a release.
To do this, create a release resource file that defines which applications are included in the release.
The release resource file is used to generate boot scripts and release packages. A system that is transferred to and
installed at another site is called a target system. How to use a release package to create a target system is described
in System Principles.
9.10.2 Release Resource File
To define a release, create a release resource file, or in short a .rel file. In the file, specify the name and version of
the release, which ERTS version it is based on, and which applications it consists of:
{release, {Name,Vsn}, {erts, EVsn},
[{Application1, AppVsn1},
...
{ApplicationN, AppVsnN}]}.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 347
9.10 Releases
Name, Vsn, EVsn, and AppVsn are strings.
The file must be named Rel.rel, where Rel is a unique name.
Each Application (atom) and AppVsn is the name and version of an application included in the release. The
minimal release based on Erlang/OTP consists of the Kernel and STDLIB applications, so these applications must
be included in the list.
If the release is to be upgraded, it must also include the SASL application.
Example: A release of ch_app from Applications has the following .app file:
{application, ch_app,
[{description, "Channel allocator"},
{vsn, "1"},
{modules, [ch_app, ch_sup, ch3]},
{registered, [ch3]},
{applications, [kernel, stdlib, sasl]},
{mod, {ch_app,[]}}
]}.
The .rel file must also contain kernel, stdlib, and sasl, as these applications are required by ch_app. The
file is called ch_rel-1.rel:
{release,
{"ch_rel", "A"},
{erts, "5.3"},
[{kernel, "2.9"},
{stdlib, "1.12"},
{sasl, "1.10"},
{ch_app, "1"}]
}.
9.10.3 Generating Boot Scripts
systools in the SASL application includes tools to build and check releases. The functions read the rel and .app
files and performs syntax and dependency checks. The systools:make_script/1,2 function is used to generate
a boot script (see System Principles):
1> systools:make_script("ch_rel-1", [local]).
ok
This creates a boot script, both the readable version, ch_rel-1.script, and the binary version,
ch_rel-1.boot, used by the runtime system.
•
•
"ch_rel-1" is the name of the .rel file, minus the extension.
local is an option that means that the directories where the applications are found are used in the boot script,
instead of $ROOT/lib ($ROOT is the root directory of the installed release).
This is a useful way to test a generated boot script locally.
When starting Erlang/OTP using the boot script, all applications from the .rel file are automatically loaded and
started:
348 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.10 Releases
% erl -boot ch_rel-1
Erlang (BEAM) emulator version 5.3
Eshell V5.3 (abort with ^G)
1>
=PROGRESS REPORT==== 13-Jun-2003::12:01:15 ===
supervisor: {local,sasl_safe_sup}
started: [{pid,<0.33.0>},
{name,alarm_handler},
{mfa,{alarm_handler,start_link,[]}},
{restart_type,permanent},
{shutdown,2000},
{child_type,worker}]
...
=PROGRESS REPORT==== 13-Jun-2003::12:01:15 ===
application: sasl
started_at: nonode@nohost
...
=PROGRESS REPORT==== 13-Jun-2003::12:01:15 ===
application: ch_app
started_at: nonode@nohost
9.10.4 Creating a Release Package
The systools:make_tar/1,2 function takes a .rel file as input and creates a zipped tar file with the code for
the specified applications, a release package:
1> systools:make_script("ch_rel-1").
ok
2> systools:make_tar("ch_rel-1").
ok
The release package by default contains:
•
•
•
•
The .app files
The .rel file
The object code for all applications, structured according to the application directory structure
The binary boot script renamed to start.boot
% tar tf ch_rel-1.tar
lib/kernel-2.9/ebin/kernel.app
lib/kernel-2.9/ebin/application.beam
...
lib/stdlib-1.12/ebin/stdlib.app
lib/stdlib-1.12/ebin/beam_lib.beam
...
lib/sasl-1.10/ebin/sasl.app
lib/sasl-1.10/ebin/sasl.beam
...
lib/ch_app-1/ebin/ch_app.app
lib/ch_app-1/ebin/ch_app.beam
lib/ch_app-1/ebin/ch_sup.beam
lib/ch_app-1/ebin/ch3.beam
releases/A/start.boot
releases/A/ch_rel-1.rel
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 349
9.10 Releases
releases/ch_rel-1.rel
A new boot script was generated, without the local option set, before the release package was made. In the release
package, all application directories are placed under lib. You do not know where the release package will be installed,
so no hard-coded absolute paths are allowed.
The release resource file mysystem.rel is duplicated in the tar file. Originally, this file was only stored in the
releases directory to make it possible for the release_handler to extract this file separately. After unpacking
the tar file, release_handler would automatically copy the file to releases/FIRST. However, sometimes the
tar file is unpacked without involving the release_handler (for example, when unpacking the first target system)
and the file is therefore now instead duplicated in the tar file so no manual copying is necessary.
If a relup file and/or a system configuration file called sys.config is found, these files are also included in the
release package. See Release Handling.
Options can be set to make the release package include source code and the ERTS binary as well.
For information on how to install the first target system, using a release package, see System Principles. For information
on how to install a new release package in an existing system, see Release Handling.
9.10.5 Directory Structure
The directory structure for the code installed by the release handler from a release package is as follows:
$ROOT/lib/App1-AVsn1/ebin
/priv
/App2-AVsn2/ebin
/priv
...
/AppN-AVsnN/ebin
/priv
/erts-EVsn/bin
/releases/Vsn
/bin
•
•
•
•
lib - Application directories
erts-EVsn/bin - Erlang runtime system executables
releases/Vsn - .rel file and boot script start.boot; if present in the release package, relup and/or
sys.config
bin - Top-level Erlang runtime system executables
Applications are not required to be located under directory $ROOT/lib. Several installation directories, which contain
different parts of a system, can thus exist. For example, the previous example can be extended as follows:
$SECOND_ROOT/.../SApp1-SAVsn1/ebin
/priv
/SApp2-SAVsn2/ebin
/priv
...
/SAppN-SAVsnN/ebin
/priv
$THIRD_ROOT/TApp1-TAVsn1/ebin
/priv
/TApp2-TAVsn2/ebin
/priv
350 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.10 Releases
...
/TAppN-TAVsnN/ebin
/priv
$SECOND_ROOT and $THIRD_ROOT
systools:make_script/2 function.
are
introduced
as
variables
in
the
call
to
the
Disk-Less and/or Read-Only Clients
If a complete system consists of disk-less and/or read-only client nodes, a clients directory is to be added to the
$ROOT directory. A read-only node is a node with a read-only file system.
The clients directory is to have one subdirectory per supported client node. The name of each client directory
is to be the name of the corresponding client node. As a minimum, each client directory is to contain the bin and
releases subdirectories. These directories are used to store information about installed releases and to appoint the
current release to the client. The $ROOT directory thus contains the following:
$ROOT/...
/clients/ClientName1/bin
/releases/Vsn
/ClientName2/bin
/releases/Vsn
...
/ClientNameN/bin
/releases/Vsn
This structure is to be used if all clients are running the same type of Erlang machine. If there are clients running
different types of Erlang machines, or on different operating systems, the clients directory can be divided into one
subdirectory per type of Erlang machine. Alternatively, one $ROOT can be set up per type of machine. For each type,
some of the directories specified for the $ROOT directory are to be included:
$ROOT/...
/clients/Type1/lib
/erts-EVsn
/bin
/ClientName1/bin
/releases/Vsn
/ClientName2/bin
/releases/Vsn
...
/ClientNameN/bin
/releases/Vsn
...
/TypeN/lib
/erts-EVsn
/bin
...
With this structure, the root directory for clients of Type1 is $ROOT/clients/Type1.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 351
9.11 Release Handling
9.11 Release Handling
9.11.1 Release Handling Principles
An important feature of the Erlang programming language is the ability to change module code in runtime, code
replacement, as described in the Erlang Reference Manual.
Based on this feature, the OTP application SASL provides a framework for upgrading and downgrading between
different versions of an entire release in runtime. This is called release handling.
The framework consists of:
•
•
Offline support - systools for generating scripts and building release packages
Online support - release_handler for unpacking and installing release packages
The minimal system based on Erlang/OTP, enabling release handling, thus consists of the Kernel, STDLIB, and SASL
applications.
Release Handling Workflow
Step 1) A release is created as described in Releases.
Step 2) The release is transferred to and installed at target environment. For information of how to install the first
target system, see System Principles.
Step 3) Modifications, for example, error corrections, are made to the code in the development environment.
Step 4) At some point, it is time to make a new version of release. The relevant .app files are updated and a new
.rel file is written.
Step 5) For each modified application, an application upgrade file, .appup, is created. In this file, it is described
how to upgrade and/or downgrade between the old and new version of the application.
Step 6) Based on the .appup files, a release upgrade file called relup, is created. This file describes how to upgrade
and/or downgrade between the old and new version of the entire release.
Step 7) A new release package is made and transferred to the target system.
Step 8) The new release package is unpacked using the release handler.
Step 9) The new version of the release is installed, also using the release handler. This is done by evaluating the
instructions in relup. Modules can be added, deleted, or reloaded, applications can be started, stopped, or restarted,
and so on. In some cases, it is even necessary to restart the entire emulator.
•
•
If the installation fails, the system can be rebooted. The old release version is then automatically used.
If the installation succeeds, the new version is made the default version, which is to now be used if there is a
system reboot.
Release Handling Aspects
Appup Cookbook, contains examples of .appup files for typical cases of upgrades/downgrades that are normally easy
to handle in runtime. However, many aspects can make release handling complicated, for example:
•
Complicated or circular dependencies can make it difficult or even impossible to decide in which order things
must be done without risking runtime errors during an upgrade or downgrade. Dependencies can be:
•
•
•
Between nodes
Between processes
Between modules
352 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.11 Release Handling
•
During release handling, non-affected processes continue normal execution. This can lead to time-outs or other
problems. For example, new processes created in the time window between suspending processes using a certain
module, and loading a new version of this module, can execute old code.
It is thus recommended that code is changed in as small steps as possible, and always kept backwards compatible.
9.11.2 Requirements
For release handling to work properly, the runtime system must have knowledge about which release it is running.
It must also be able to change (in runtime) which boot script and system configuration file to use if the system is
rebooted, for example, by heart after a failure. Thus, Erlang must be started as an embedded system; for information
on how to do this, see Embedded System.
For system reboots to work properly, it is also required that the system is started with heartbeat monitoring, see the
erl(1) manual page in ERTS and the heart(3) manual page in Kernel
Other requirements:
•
The boot script included in a release package must be generated from the same .rel file as the release package
itself.
•
Information about applications is fetched from the script when an upgrade or downgrade is performed.
The system must be configured using only one system configuration file, called sys.config.
•
If found, this file is automatically included when a release package is created.
All versions of a release, except the first one, must contain a relup file.
If found, this file is automatically included when a release package is created.
9.11.3 Distributed Systems
If the system consists of several Erlang nodes, each node can use its own version of the release. The release handler
is a locally registered process and must be called at each node where an upgrade or downgrade is required. A release
handling instruction, sync_nodes, can be used to synchronize the release handler processes at a number of nodes,
see the appup(4) manual page in SASL.
9.11.4 Release Handling Instructions
OTP supports a set of release handling instructions that are used when creating .appup files. The release handler
understands a subset of these, the low-level instructions. To make it easier for the user, there are also a number of
high-level instructions, which are translated to low-level instructions by systools:make_relup.
Some of the most frequently used instructions are described in this section. The complete list of instructions is included
in the appup(4) manual page in SASL.
First, some definitions:
•
•
Residence module - The module where a process has its tail-recursive loop function(s). If these functions are
implemented in several modules, all those modules are residence modules for the process.
Functional module - A module that is not a residence module for any process.
For a process implemented using an OTP behaviour, the behaviour module is the residence module for that process.
The callback module is a functional module.
load_module
If a simple extension has been made to a functional module, it is sufficient to load the new version of the module into
the system, and remove the old version. This is called simple code replacement and for this the following instruction
is used:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 353
9.11 Release Handling
{load_module, Module}
update
If a more complex change has been made, for example, a change to the format of the internal state of a gen_server,
simple code replacement is not sufficient. Instead, it is necessary to:
•
•
•
•
Suspend the processes using the module (to avoid that they try to handle any requests before the code
replacement is completed).
Ask them to transform the internal state format and switch to the new version of the module.
Remove the old version.
Resume the processes.
This is called synchronized code replacement and for this the following instructions are used:
{update, Module, {advanced, Extra}}
{update, Module, supervisor}
update with argument {advanced,Extra} is used when changing the internal state of a behaviour as described
above. It causes behaviour processes to call the callback function code_change, passing the term Extra and some
other information as arguments. See the manual pages for the respective behaviours and Appup Cookbook.
update with argument supervisor is used when changing the start specification of a supervisor. See Appup
Cookbook.
When a module is to be updated, the release handler finds which processes that are using the module by traversing the
supervision tree of each running application and checking all the child specifications:
{Id, StartFunc, Restart, Shutdown, Type, Modules}
A process uses a module if the name is listed in Modules in the child specification for the process.
If Modules=dynamic, which is the case for event managers, the event manager process informs the release handler
about the list of currently installed event handlers (gen_event), and it is checked if the module name is in this list
instead.
The release handler suspends, asks for code change, and resumes processes by calling the functions
sys:suspend/1,2, sys:change_code/4,5, and sys:resume/1,2, respectively.
add_module and delete_module
If a new module is introduced, the following instruction is used:
{add_module, Module}
The instruction loads the module and is necessary when running Erlang in embedded mode. It is not strictly required
when running Erlang in interactive (default) mode, since the code server then automatically searches for and loads
unloaded modules.
The opposite of add_module is delete_module, which unloads a module:
354 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.11 Release Handling
{delete_module, Module}
Any process, in any application, with Module as residence module, is killed when the instruction is evaluated. The
user must therefore ensure that all such processes are terminated before deleting the module, to avoid a situation with
failing supervisor restarts.
Application Instructions
The following is the instruction for adding an application:
{add_application, Application}
Adding an application means that the modules defined by the modules key in the .app file are loaded using a
number of add_module instructions, and then the application is started.
The following is the instruction for removing an application:
{remove_application, Application}
Removing an application means that the application is stopped, the modules are unloaded using a number of
delete_module instructions, and then the application specification is unloaded from the application controller.
The following is the instruction for restarting an application:
{restart_application, Application}
Restarting an application means that the application is stopped and then started again similar to using the instructions
remove_application and add_application in sequence.
apply (Low-Level)
To call an arbitrary function from the release handler, the following instruction is used:
{apply, {M, F, A}}
The release handler evalutes apply(M, F, A).
restart_new_emulator (Low-Level)
This instruction is used when changing to a new emulator version, or when any of the core applications Kernel,
STDLIB, or SASL is upgraded. If a system reboot is needed for another reason, the restart_emulator instruction
is to be used instead.
This instruction requires that the system is started with heartbeat monitoring, see the erl(1) manual page in ERTS
and the heart(3) manual page in Kernel.
The restart_new_emulator instruction must always be the first instruction in a relup. If the relup is generated
by systools:make_relup/3,4, this is automatically ensured.
When the release handler encounters the instruction, it first generates a temporary boot file, which starts the new
versions of the emulator and the core applications, and the old version of all other applications. Then it shuts down the
current emulator by calling init:reboot(), see the init(3) manual page in Kernel. All processes are terminated
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 355
9.11 Release Handling
gracefully and the system is rebooted by the heart program, using the temporary boot file. After the reboot, the rest
of the relup instructions are executed. This is done as a part of the temporary boot script.
Warning:
This mechanism causes the new versions of the emulator and core applications to run with the old version of
other applications during startup. Thus, take extra care to avoid incompatibility. Incompatible changes in the core
applications can in some situations be necessary. If possible, such changes are preceded by deprecation over two
major releases before the actual change. To ensure the application is not crashed by an incompatible change,
always remove any call to deprecated functions as soon as possible.
An info report is written when the upgrade is completed. To programmatically find out if the upgrade is complete,
call release_handler:which_releases(current) and check if it returns the expected (that is, the new)
release.
The new release version must be made permanent when the new emulator is operational. Otherwise, the old version
will be used if there is a new system reboot.
On UNIX, the release handler tells the heart program which command to use to reboot the system. The environment
variable HEART_COMMAND, normally used by the heart program, is ignored in this case. The command instead
defaults to $ROOT/bin/start. Another command can be set by using the SASL configuration parameter
start_prg, see the sasl(6) manual page.
restart_emulator (Low-Level)
This instruction is not related to upgrades of ERTS or any of the core applications. It can be used by any application
to force a restart of the emulator after all upgrade instructions are executed.
A relup script can only have one restart_emulator instruction and it must always be placed at the end. If the
relup is generated by systools:make_relup/3,4, this is automatically ensured.
When the release handler encounters the instruction, it shuts down the emulator by calling init:reboot(), see the
init(3) manual page in Kernel. All processes are terminated gracefully and the system can then be rebooted by the
heart program using the new release version. No more upgrade instruction is executed after the restart.
9.11.5 Application Upgrade File
To define how to upgrade/downgrade between the current version and previous versions of an application, an
application upgrade file, or in short an .appup file is created. The file is to be called Application.appup,
where Application is the application name:
{Vsn,
[{UpFromVsn1,
...,
{UpFromVsnK,
[{DownToVsn1,
...,
{DownToVsnK,
•
•
•
•
InstructionsU1},
InstructionsUK}],
InstructionsD1},
InstructionsDK}]}.
Vsn, a string, is the current version of the application, as defined in the .app file.
Each UpFromVsn is a previous version of the application to upgrade from.
Each DownToVsn is a previous version of the application to downgrade to.
Each Instructions is a list of release handling instructions.
356 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.11 Release Handling
For information about the syntax and contents of the .appup file, see the appup(4) manual page in SASL.
Appup Cookbook includes examples of .appup files for typical upgrade/downgrade cases.
Example: Consider the release ch_rel-1 from Releases. Assume you want to add a function available/0 to
server ch3, which returns the number of available channels (when trying out the example, change in a copy of the
original directory, so that the first versions are still available):
-module(ch3).
-behaviour(gen_server).
-export([start_link/0]).
-export([alloc/0, free/1]).
-export([available/0]).
-export([init/1, handle_call/3, handle_cast/2]).
start_link() ->
gen_server:start_link({local, ch3}, ch3, [], []).
alloc() ->
gen_server:call(ch3, alloc).
free(Ch) ->
gen_server:cast(ch3, {free, Ch}).
available() ->
gen_server:call(ch3, available).
init(_Args) ->
{ok, channels()}.
handle_call(alloc, _From, Chs) ->
{Ch, Chs2} = alloc(Chs),
{reply, Ch, Chs2};
handle_call(available, _From, Chs) ->
N = available(Chs),
{reply, N, Chs}.
handle_cast({free, Ch}, Chs) ->
Chs2 = free(Ch, Chs),
{noreply, Chs2}.
A new version of the ch_app.app file must now be created, where the version is updated:
{application, ch_app,
[{description, "Channel allocator"},
{vsn, "2"},
{modules, [ch_app, ch_sup, ch3]},
{registered, [ch3]},
{applications, [kernel, stdlib, sasl]},
{mod, {ch_app,[]}}
]}.
To upgrade ch_app from "1" to "2" (and to downgrade from "2" to "1"), you only need to load the new (old)
version of the ch3 callback module. Create the application upgrade file ch_app.appup in the ebin directory:
{"2",
[{"1", [{load_module, ch3}]}],
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 357
9.11 Release Handling
[{"1", [{load_module, ch3}]}]
}.
9.11.6 Release Upgrade File
To define how to upgrade/downgrade between the new version and previous versions of a release, a release upgrade
file, or in short relup file, is to be created.
This file does not need to be created manually, it can be generated by systools:make_relup/3,4. The relevant
versions of the .rel file, .app files, and .appup files are used as input. It is deducted which applications are to
be added and deleted, and which applications that must be upgraded and/or downgraded. The instructions for this are
fetched from the .appup files and transformed into a single list of low-level instructions in the right order.
If the relup file is relatively simple, it can be created manually. It it only to contain low-level instructions.
For details about the syntax and contents of the release upgrade file, see the relup(4) manual page in SASL.
Example, continued from the previous section: You have a new version "2" of ch_app and an .appup file. A
new version of the .rel file is also needed. This time the file is called ch_rel-2.rel and the release version
string is changed from "A" to "B":
{release,
{"ch_rel", "B"},
{erts, "5.3"},
[{kernel, "2.9"},
{stdlib, "1.12"},
{sasl, "1.10"},
{ch_app, "2"}]
}.
Now the relup file can be generated:
1> systools:make_relup("ch_rel-2", ["ch_rel-1"], ["ch_rel-1"]).
ok
This generates a relup file with instructions for how to upgrade from version "A" ("ch_rel-1") to version
"B" ("ch_rel-2") and how to downgrade from version "B" to version "A".
Both the old and new versions of the .app and .rel files must be in the code path, as well as the .appup and (new)
.beam files. The code path can be extended by using the option path:
1> systools:make_relup("ch_rel-2", ["ch_rel-1"], ["ch_rel-1"],
[{path,["../ch_rel-1",
"../ch_rel-1/lib/ch_app-1/ebin"]}]).
ok
9.11.7 Installing a Release
When you have made a new version of a release, a release package can be created with this new version and transferred
to the target environment.
To install the new version of the release in runtime, the release handler is used. This is a process belonging to the
SASL application, which handles unpacking, installation, and removal of release packages. It is communicated through
the release_handler module. For details, see the release_handler(3) manual page in SASL.
358 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.11 Release Handling
Assuming there is an operational target system with installation root directory $ROOT, the release package with the
new version of the release is to be copied to $ROOT/releases.
First, unpack the release package. The files are then extracted from the package:
release_handler:unpack_release(ReleaseName) => {ok, Vsn}
•
•
ReleaseName is the name of the release package except the .tar.gz extension.
Vsn is the version of the unpacked release, as defined in its .rel file.
A directory $ROOT/lib/releases/Vsn is created, where the .rel file, the boot script start.boot, the
system configuration file sys.config, and relup are placed. For applications with new version numbers, the
application directories are placed under $ROOT/lib. Unchanged applications are not affected.
An unpacked release can be installed. The release handler then evaluates the instructions in relup, step by step:
release_handler:install_release(Vsn) => {ok, FromVsn, []}
If an error occurs during the installation, the system is rebooted using the old version of the release. If installation
succeeds, the system is afterwards using the new version of the release, but if anything happens and the system is
rebooted, it starts using the previous version again.
To be made the default version, the newly installed release must be made permanent, which means the previous
version becomes old:
release_handler:make_permanent(Vsn) => ok
The system keeps information about which versions are old and permanent in the files $ROOT/releases/
RELEASES and $ROOT/releases/start_erl.data.
To downgrade from Vsn to FromVsn, install_release must be called again:
release_handler:install_release(FromVsn) => {ok, Vsn, []}
An installed, but not permanent, release can be removed. Information about the release is then deleted from $ROOT/
releases/RELEASES and the release-specific code, that is, the new application directories and the $ROOT/
releases/Vsn directory, are removed.
release_handler:remove_release(Vsn) => ok
Example (continued from the previous sections)
Step 1) Create a target system as described in System Principles of the first version "A" of ch_rel from Releases.
This time sys.config must be included in the release package. If no configuration is needed, the file is to contain
the empty list:
[].
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 359
9.11 Release Handling
Step 2) Start the system as a simple target system. In reality, it is to be started as an embedded system. However, using
erl with the correct boot script and config file is enough for illustration purposes:
% cd $ROOT
% bin/erl -boot $ROOT/releases/A/start -config $ROOT/releases/A/sys
...
$ROOT is the installation directory of the target system.
Step 3) In another Erlang shell, generate start scripts and create a release package for the new version "B". Remember
to include (a possible updated) sys.config and the relup file, see Release Upgrade File.
1> systools:make_script("ch_rel-2").
ok
2> systools:make_tar("ch_rel-2").
ok
The new release package now also contains version "2" of ch_app and the relup file:
% tar tf ch_rel-2.tar
lib/kernel-2.9/ebin/kernel.app
lib/kernel-2.9/ebin/application.beam
...
lib/stdlib-1.12/ebin/stdlib.app
lib/stdlib-1.12/ebin/beam_lib.beam
...
lib/sasl-1.10/ebin/sasl.app
lib/sasl-1.10/ebin/sasl.beam
...
lib/ch_app-2/ebin/ch_app.app
lib/ch_app-2/ebin/ch_app.beam
lib/ch_app-2/ebin/ch_sup.beam
lib/ch_app-2/ebin/ch3.beam
releases/B/start.boot
releases/B/relup
releases/B/sys.config
releases/B/ch_rel-2.rel
releases/ch_rel-2.rel
Step 4) Copy the release package ch_rel-2.tar.gz to the $ROOT/releases directory.
Step 5) In the running target system, unpack the release package:
1> release_handler:unpack_release("ch_rel-2").
{ok,"B"}
The new application version ch_app-2 is installed under $ROOT/lib next to ch_app-1. The kernel, stdlib,
and sasl directories are not affected, as they have not changed.
Under $ROOT/releases, a new directory B is created, containing ch_rel-2.rel, start.boot,
sys.config, and relup.
Step 6) Check if the function ch3:available/0 is available:
360 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.11 Release Handling
2> ch3:available().
** exception error: undefined function ch3:available/0
Step 7) Install the new release. The instructions in $ROOT/releases/B/relup are executed one by one, resulting
in the new version of ch3 being loaded. The function ch3:available/0 is now available:
3> release_handler:install_release("B").
{ok,"A",[]}
4> ch3:available().
3
5> code:which(ch3).
".../lib/ch_app-2/ebin/ch3.beam"
6> code:which(ch_sup).
".../lib/ch_app-1/ebin/ch_sup.beam"
Processes in ch_app for which code have not been updated, for example, the supervisor, are still evaluating code
from ch_app-1.
Step 8) If the target system is now rebooted, it uses version "A" again. The "B" version must be made permanent, to
be used when the system is rebooted.
7> release_handler:make_permanent("B").
ok
9.11.8 Updating Application Specifications
When a new version of a release is installed, the application specifications are automatically updated for all loaded
applications.
Note:
The information about the new application specifications is fetched from the boot script included in the release
package. Thus, it is important that the boot script is generated from the same .rel file as is used to build the
release package itself.
Specifically, the application configuration parameters are automatically updated according to (in increasing priority
order):
•
•
•
The data in the boot script, fetched from the new application resource file App.app
The new sys.config
Command-line arguments -App Par Val
This means that parameter values set in the other system configuration files and values set using
application:set_env/3 are disregarded.
When an installed release is made permanent, the system process init is set to point out the new sys.config.
After the installation, the application controller compares the old and new configuration parameters for all running
applications and call the callback function:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 361
9.12 Appup Cookbook
Module:config_change(Changed, New, Removed)
•
•
•
Module is the application callback module as defined by the mod key in the .app file.
Changed and New are lists of {Par,Val} for all changed and added configuration parameters, respectively.
Removed is a list of all parameters Par that have been removed.
The function is optional and can be omitted when implementing an application callback module.
9.12 Appup Cookbook
This section includes examples of .appup files for typical cases of upgrades/downgrades done in runtime.
9.12.1 Changing a Functional Module
When a functional module has been changed, for example, if a new function has been added or a bug has been corrected,
simple code replacement is sufficient, for example:
{"2",
[{"1", [{load_module, m}]}],
[{"1", [{load_module, m}]}]
}.
9.12.2 Changing a Residence Module
In a system implemented according to the OTP design principles, all processes, except system processes and special
processes, reside in one of the behaviours supervisor, gen_server, gen_fsm, gen_statem or gen_event.
These belong to the STDLIB application and upgrading/downgrading normally requires an emulator restart.
OTP thus provides no support for changing residence modules except in the case of special processes.
9.12.3 Changing a Callback Module
A callback module is a functional module, and for code extensions simple code replacement is sufficient.
Example: When adding a function to ch3, as described in the example in Release Handling, ch_app.appup looks
as follows:
{"2",
[{"1", [{load_module, ch3}]}],
[{"1", [{load_module, ch3}]}]
}.
OTP also supports changing the internal state of behaviour processes, see Changing Internal State.
9.12.4 Changing Internal State
In this case, simple code replacement is not sufficient. The process must explicitly transform its state using the
callback function code_change before switching to the new version of the callback module. Thus, synchronized
code replacement is used.
Example: Consider gen_server ch3 from gen_server Behaviour. The internal state is a term Chs representing
the available channels. Assume you want to add a counter N, which keeps track of the number of alloc requests so
far. This means that the format must be changed to {Chs,N}.
362 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.12 Appup Cookbook
The .appup file can look as follows:
{"2",
[{"1", [{update, ch3, {advanced, []}}]}],
[{"1", [{update, ch3, {advanced, []}}]}]
}.
The third element of the update instruction is a tuple {advanced,Extra}, which says that the affected processes
are to do a state transformation before loading the new version of the module. This is done by the processes calling the
callback function code_change (see the gen_server(3) manual page in STDLIB). The term Extra, in this
case [], is passed as is to the function:
-module(ch3).
...
-export([code_change/3]).
...
code_change({down, _Vsn}, {Chs, N}, _Extra) ->
{ok, Chs};
code_change(_Vsn, Chs, _Extra) ->
{ok, {Chs, 0}}.
The first argument is {down,Vsn} if there is a downgrade, or Vsn if there is a upgrade. The term Vsn is fetched
from the 'original' version of the module, that is, the version you are upgrading from, or downgrading to.
The version is defined by the module attribute vsn, if any. There is no such attribute in ch3, so in this case the version
is the checksum (a huge integer) of the beam file, an uninteresting value, which is ignored.
The other callback functions of ch3 must also be modified and perhaps a new interface function must be added, but
this is not shown here.
9.12.5 Module Dependencies
Assume that a module is extended by adding an interface function, as in the example in Release Handling, where a
function available/0 is added to ch3.
If a call is added to this function, say in module m1, a runtime error could can occur during release upgrade if the new
version of m1 is loaded first and calls ch3:available/0 before the new version of ch3 is loaded.
Thus, ch3 must be loaded before m1, in the upgrade case, and conversely in the downgrade case. m1 is said to be
dependent on ch3. In a release handling instruction, this is expressed by the DepMods element:
{load_module, Module, DepMods}
{update, Module, {advanced, Extra}, DepMods}
DepMods is a list of modules, on which Module is dependent.
Example: The module m1 in application myapp is dependent on ch3 when upgrading from "1" to "2", or downgrading
from "2" to "1":
myapp.appup:
{"2",
[{"1", [{load_module, m1, [ch3]}]}],
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 363
9.12 Appup Cookbook
[{"1", [{load_module, m1, [ch3]}]}]
}.
ch_app.appup:
{"2",
[{"1", [{load_module, ch3}]}],
[{"1", [{load_module, ch3}]}]
}.
If instead m1 and ch3 belong to the same application, the .appup file can look as follows:
{"2",
[{"1",
[{load_module,
{load_module,
[{"1",
[{load_module,
{load_module,
}.
ch3},
m1, [ch3]}]}],
ch3},
m1, [ch3]}]}]
m1 is dependent on ch3 also when downgrading. systools knows the difference between up- and downgrading
and generates a correct relup, where ch3 is loaded before m1 when upgrading, but m1 is loaded before ch3 when
downgrading.
9.12.6 Changing Code for a Special Process
In this case, simple code replacement is not sufficient. When a new version of a residence module for a special process
is loaded, the process must make a fully qualified call to its loop function to switch to the new code. Thus, synchronized
code replacement must be used.
Note:
The name(s) of the user-defined residence module(s) must be listed in the Modules part of the child specification
for the special process. Otherwise the release handler cannot find the process.
Example: Consider the example ch4 in sys and proc_lib. When started by a supervisor, the child specification can
look as follows:
{ch4, {ch4, start_link, []},
permanent, brutal_kill, worker, [ch4]}
If ch4 is part of the application sp_app and a new version of the module is to be loaded when upgrading from version
"1" to "2" of this application, sp_app.appup can look as follows:
{"2",
[{"1", [{update, ch4, {advanced, []}}]}],
[{"1", [{update, ch4, {advanced, []}}]}]
}.
364 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.12 Appup Cookbook
The update instruction must contain the tuple {advanced,Extra}. The instruction makes the special process
call the callback function system_code_change/4, a function the user must implement. The term Extra, in this
case [], is passed as is to system_code_change/4:
-module(ch4).
...
-export([system_code_change/4]).
...
system_code_change(Chs, _Module, _OldVsn, _Extra) ->
{ok, Chs}.
•
•
•
The first argument is the internal state State, passed from function
sys:handle_system_msg(Request, From, Parent, Module, Deb, State), and called by
the special process when a system message is received. In ch4, the internal state is the set of available channels
Chs.
The second argument is the name of the module (ch4).
The third argument is Vsn or {down,Vsn}, as described for gen_server:code_change/3 in Changing
Internal State.
In this case, all arguments but the first are ignored and the function simply returns the internal state again. This is
enough if the code only has been extended. If instead the internal state is changed (similar to the example in Changing
Internal State), this is done in this function and {ok,Chs2} returned.
9.12.7 Changing a Supervisor
The supervisor behaviour supports changing the internal state, that is, changing the restart strategy and maximum
restart frequency properties, as well as changing the existing child specifications.
Child processes can be added or deleted, but this is not handled automatically. Instructions must be given by in the
.appup file.
Changing Properties
Since the supervisor is to change its internal state, synchronized code replacement is required. However, a special
update instruction must be used.
First, the new version of the callback module must be loaded, both in the case of upgrade and downgrade. Then the
new return value of init/1 can be checked and the internal state be changed accordingly.
The following upgrade instruction is used for supervisors:
{update, Module, supervisor}
Example: To change the restart strategy of ch_sup (from Supervisor Behaviour) from one_for_one to
one_for_all, change the callback function init/1 in ch_sup.erl:
-module(ch_sup).
...
init(_Args) ->
{ok, {#{strategy => one_for_all, ...}, ...}}.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 365
9.12 Appup Cookbook
The file ch_app.appup:
{"2",
[{"1", [{update, ch_sup, supervisor}]}],
[{"1", [{update, ch_sup, supervisor}]}]
}.
Changing Child Specifications
The instruction, and thus the .appup file, when changing an existing child specification, is the same as when changing
properties as described earlier:
{"2",
[{"1", [{update, ch_sup, supervisor}]}],
[{"1", [{update, ch_sup, supervisor}]}]
}.
The changes do not affect existing child processes. For example, changing the start function only specifies how the
child process is to be restarted, if needed later on.
The id of the child specification cannot be changed.
Changing the Modules field of the child specification can affect the release handling process itself, as this field is
used to identify which processes are affected when doing a synchronized code replacement.
Adding and Deleting Child Processes
As stated earlier, changing child specifications does not affect existing child processes. New child specifications are
automatically added, but not deleted. Child processes are not automatically started or terminated, this must be done
using apply instructions.
Example: Assume a new child process m1 is to be added to ch_sup when upgrading ch_app from "1" to "2". This
means m1 is to be deleted when downgrading from "2" to "1":
{"2",
[{"1",
[{update, ch_sup, supervisor},
{apply, {supervisor, restart_child, [ch_sup, m1]}}
]}],
[{"1",
[{apply, {supervisor, terminate_child, [ch_sup, m1]}},
{apply, {supervisor, delete_child, [ch_sup, m1]}},
{update, ch_sup, supervisor}
]}]
}.
The order of the instructions is important.
The supervisor must be registered as ch_sup for the script to work. If the supervisor is not registered, it
cannot be accessed directly from the script. Instead a help function that finds the pid of the supervisor and calls
supervisor:restart_child, and so on, must be written. This function is then to be called from the script using
the apply instruction.
If the module m1 is introduced in version "2" of ch_app, it must also be loaded when upgrading and deleted when
downgrading:
366 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.12 Appup Cookbook
{"2",
[{"1",
[{add_module, m1},
{update, ch_sup, supervisor},
{apply, {supervisor, restart_child, [ch_sup, m1]}}
]}],
[{"1",
[{apply, {supervisor, terminate_child, [ch_sup, m1]}},
{apply, {supervisor, delete_child, [ch_sup, m1]}},
{update, ch_sup, supervisor},
{delete_module, m1}
]}]
}.
As stated earlier, the order of the instructions is important. When upgrading, m1 must be loaded, and the supervisor
child specification changed, before the new child process can be started. When downgrading, the child process must
be terminated before the child specification is changed and the module is deleted.
9.12.8 Adding or Deleting a Module
Example: A new functional module m is added to ch_app:
{"2",
[{"1", [{add_module, m}]}],
[{"1", [{delete_module, m}]}]
9.12.9 Starting or Terminating a Process
In a system structured according to the OTP design principles, any process would be a child process belonging to a
supervisor, see Adding and Deleting Child Processes in Changing a Supervisor.
9.12.10 Adding or Removing an Application
When adding or removing an application, no .appup file is needed. When generating relup, the .rel files are
compared and the add_application and remove_application instructions are added automatically.
9.12.11 Restarting an Application
Restarting an application is useful when a change is too complicated to be made without restarting the processes, for
example, if the supervisor hierarchy has been restructured.
Example: When adding a child m1 to ch_sup, as in Adding and Deleting Child Processes in Changing a Supervisor,
an alternative to updating the supervisor is to restart the entire application:
{"2",
[{"1", [{restart_application, ch_app}]}],
[{"1", [{restart_application, ch_app}]}]
}.
9.12.12 Changing an Application Specification
When installing a release, the application specifications are automatically updated before evaluating the relup script.
Thus, no instructions are needed in the .appup file:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 367
9.12 Appup Cookbook
{"2",
[{"1", []}],
[{"1", []}]
}.
9.12.13 Changing Application Configuration
Changing an application configuration by updating the env key in the .app file is an instance of changing an
application specification, see the previous section.
Alternatively, application configuration parameters can be added or updated in sys.config.
9.12.14 Changing Included Applications
The release handling instructions for adding, removing, and restarting applications apply to primary applications only.
There are no corresponding instructions for included applications. However, since an included application is really a
supervision tree with a topmost supervisor, started as a child process to a supervisor in the including application, a
relup file can be manually created.
Example: Assume there is a release containing an application prim_app, which have a supervisor prim_sup in
its supervision tree.
In a new version of the release, the application ch_app is to be included in prim_app. That is, its topmost supervisor
ch_sup is to be started as a child process to prim_sup.
The workflow is as follows:
Step 1) Edit the code for prim_sup:
init(...) ->
{ok, {...supervisor flags...,
[...,
{ch_sup, {ch_sup,start_link,[]},
permanent,infinity,supervisor,[ch_sup]},
...]}}.
Step 2) Edit the .app file for prim_app:
{application, prim_app,
[...,
{vsn, "2"},
...,
{included_applications, [ch_app]},
...
]}.
Step 3) Create a new .rel file, including ch_app:
{release,
...,
[...,
{prim_app, "2"},
{ch_app, "1"}]}.
368 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.12 Appup Cookbook
The included application can be started in two ways. This is described in the next two sections.
Application Restart
Step 4a) One way to start the included application is to restart the entire prim_app application. Normally, the
restart_application instruction in the .appup file for prim_app would be used.
However, if this is done and a relup file is generated, not only would it contain instructions for restarting (that is,
removing and adding) prim_app, it would also contain instructions for starting ch_app (and stopping it, in the case
of downgrade). This is because ch_app is included in the new .rel file, but not in the old one.
Instead, a correct relup file can be created manually, either from scratch or by editing the generated version. The
instructions for starting/stopping ch_app are replaced by instructions for loading/unloading the application:
{"B",
[{"A",
[],
[{load_object_code,{ch_app,"1",[ch_sup,ch3]}},
{load_object_code,{prim_app,"2",[prim_app,prim_sup]}},
point_of_no_return,
{apply,{application,stop,[prim_app]}},
{remove,{prim_app,brutal_purge,brutal_purge}},
{remove,{prim_sup,brutal_purge,brutal_purge}},
{purge,[prim_app,prim_sup]},
{load,{prim_app,brutal_purge,brutal_purge}},
{load,{prim_sup,brutal_purge,brutal_purge}},
{load,{ch_sup,brutal_purge,brutal_purge}},
{load,{ch3,brutal_purge,brutal_purge}},
{apply,{application,load,[ch_app]}},
{apply,{application,start,[prim_app,permanent]}}]}],
[{"A",
[],
[{load_object_code,{prim_app,"1",[prim_app,prim_sup]}},
point_of_no_return,
{apply,{application,stop,[prim_app]}},
{apply,{application,unload,[ch_app]}},
{remove,{ch_sup,brutal_purge,brutal_purge}},
{remove,{ch3,brutal_purge,brutal_purge}},
{purge,[ch_sup,ch3]},
{remove,{prim_app,brutal_purge,brutal_purge}},
{remove,{prim_sup,brutal_purge,brutal_purge}},
{purge,[prim_app,prim_sup]},
{load,{prim_app,brutal_purge,brutal_purge}},
{load,{prim_sup,brutal_purge,brutal_purge}},
{apply,{application,start,[prim_app,permanent]}}]}]
}.
Supervisor Change
Step 4b) Another way to start the included application (or stop it in the case of downgrade) is by combining instructions
for adding and removing child processes to/from prim_sup with instructions for loading/unloading all ch_app
code and its application specification.
Again, the relup file is created manually. Either from scratch or by editing a generated version. Load all code
for ch_app first, and also load the application specification, before prim_sup is updated. When downgrading,
prim_sup is to updated first, before the code for ch_app and its application specification are unloaded.
{"B",
[{"A",
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 369
9.12 Appup Cookbook
[],
[{load_object_code,{ch_app,"1",[ch_sup,ch3]}},
{load_object_code,{prim_app,"2",[prim_sup]}},
point_of_no_return,
{load,{ch_sup,brutal_purge,brutal_purge}},
{load,{ch3,brutal_purge,brutal_purge}},
{apply,{application,load,[ch_app]}},
{suspend,[prim_sup]},
{load,{prim_sup,brutal_purge,brutal_purge}},
{code_change,up,[{prim_sup,[]}]},
{resume,[prim_sup]},
{apply,{supervisor,restart_child,[prim_sup,ch_sup]}}]}],
[{"A",
[],
[{load_object_code,{prim_app,"1",[prim_sup]}},
point_of_no_return,
{apply,{supervisor,terminate_child,[prim_sup,ch_sup]}},
{apply,{supervisor,delete_child,[prim_sup,ch_sup]}},
{suspend,[prim_sup]},
{load,{prim_sup,brutal_purge,brutal_purge}},
{code_change,down,[{prim_sup,[]}]},
{resume,[prim_sup]},
{remove,{ch_sup,brutal_purge,brutal_purge}},
{remove,{ch3,brutal_purge,brutal_purge}},
{purge,[ch_sup,ch3]},
{apply,{application,unload,[ch_app]}}]}]
}.
9.12.15 Changing Non-Erlang Code
Changing code for a program written in another programming language than Erlang, for example, a port program, is
application-dependent and OTP provides no special support for it.
Example: When changing code for a port program, assume that the Erlang process controlling the port is a
gen_server portc and that the port is opened in the callback function init/1:
init(...) ->
...,
PortPrg = filename:join(code:priv_dir(App), "portc"),
Port = open_port({spawn,PortPrg}, [...]),
...,
{ok, #state{port=Port, ...}}.
If the port program is to be updated, the code for the gen_server can be extended with a code_change function,
which closes the old port and opens a new port. (If necessary, the gen_server can first request data that must be
saved from the port program and pass this data to the new port):
code_change(_OldVsn, State, port) ->
State#state.port ! close,
receive
{Port,close} ->
true
end,
PortPrg = filename:join(code:priv_dir(App), "portc"),
Port = open_port({spawn,PortPrg}, [...]),
{ok, #state{port=Port, ...}}.
Update the application version number in the .app file and write an .appup file:
370 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
9.12 Appup Cookbook
["2",
[{"1", [{update, portc, {advanced,port}}]}],
[{"1", [{update, portc, {advanced,port}}]}]
].
Ensure that the priv directory, where the C program is located, is included in the new release package:
1> systools:make_tar("my_release", [{dirs,[priv]}]).
...
9.12.16 Emulator Restart and Upgrade
Two upgrade instructions restart the emulator:
•
restart_new_emulator
•
Intended when ERTS, Kernel, STDLIB, or SASL is upgraded. It is automatically added when the relup file
is generated by systools:make_relup/3,4. It is executed before all other upgrade instructions. For more
information about this instruction, see restart_new_emulator (Low-Level) in Release Handling Instructions.
restart_emulator
Used when a restart of the emulator is required after all other upgrade instructions are executed. For more
information about this instruction, see restart_emulator (Low-Level) in Release Handling Instructions.
If an emulator restart is necessary and no upgrade instructions are needed, that is, if the restart itself is enough for the
upgraded applications to start running the new versions, a simple relup file can be created manually:
{"B",
[{"A",
[],
[restart_emulator]}],
[{"A",
[],
[restart_emulator]}]
}.
In this case, the release handler framework with automatic packing and unpacking of release packages, automatic path
updates, and so on, can be used without having to specify .appup files.
9.12.17 Emulator Upgrade From Pre OTP R15
From OTP R15, an emulator upgrade is performed by restarting the emulator with new versions of the core applications
(Kernel, STDLIB, and SASL) before loading code and running upgrade instruction for other applications. For this to
work, the release to upgrade from must include OTP R15 or later.
For the case where the release to upgrade from includes an earlier emulator version, systools:make_relup
creates a backwards compatible relup file. This means that all upgrade instructions are executed before the emulator
is restarted. The new application code is therefore loaded into the old emulator. If the new code is compiled with the
new emulator, there can be cases where the beam format has changed and beam files cannot be loaded. To overcome
this problem, compile the new code with the old emulator.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 371
10.1 Introduction
10 OAM Principles
10.1 Introduction
The Operation and Maintenance (OAM) support in OTP consists of a generic model for management subsystems in
OTP, and some components to be used in these subsystems. This section describes the model.
The main idea in the model is that it is not tied to any specific management protocol. An Application Programming
Interface (API) is defined, which can be used to write adaptations for specific management protocols.
Each OAM component in OTP is implemented as one sub-application, which can be included in a management
application for the system. Notice that such a complete management application is not in the scope of this generic
functionality. However, this section includes examples illustrating how such an application can be built.
10.1.1 Terminology
The protocol-independent architectural model on the network level is the well-known client-server model for
management operations. This model is based on the client-server principle, where the manager (client) sends a request
from a manager to an agent (server) when it accesses management information. The agent sends a reply back to the
manager. There are two main differences to the normal client-server model:
•
•
Usually a few managers communicate with many agents.
The agent can spontaneously send a notification, for example, an alarm, to the manager.
The following picture illustrates the idea:
Figure 1.1: Terminology
The manager is often referred to as the Network Management System (NMS), to emphasize that it usually is realized
as a program that presents data to an operator.
372 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
10.1 Introduction
The agent is an entity that executes within a Network Element (NE). In OTP, the NE can be a distributed system,
meaning that the distributed system is managed as one entity. Of course, the agent can be configured to be able to run
on one of several nodes, making it a distributed OTP application.
The management information is defined in a Management Information Base (MIB). It is a formal definition of
which information the agent makes available to the manager. The manager accesses the MIB through a management
protocol, such as SNMP, CMIP, HTTP, or CORBA. Each protocol has its own MIB definition language. In SNMP, it
is a subset of ASN.1, in CMIP it is GDMO, in HTTP it is implicit, and using CORBA, it is IDL.
Usually, the entities defined in the MIB are called Managed Objects (MOs), although they do not have to be objects
in the object-oriented way. For example, a simple scalar variable defined in a MIB is called an MO. The MOs are
logical objects, not necessarily with a one-to-one mapping to the resources.
10.1.2 Model
This section presents the generic protocol-independent model for use within an OTP-based NE. This model is used
by all OAM components and can be used by the applications. The advantage of the model is that it clearly separates
the resources from the management protocol. The resources do not need to be aware of which management protocol
is used to manage the system. The same resources can therefore be managed with different protocols.
The entities involved in this model are the agent, which terminates the management protocol, and the resources,
which is to be managed, that is, the actual application entities. The resources should in general have no knowledge
of the management protocol used, and the agent should have no knowledge of the managed resources. This implies
that a translation mechanism is needed, to translate the management operations to operations on the resources. This
translation mechanism is usually called instrumentation and the function that implements it is called instrumentation
function. The instrumentation functions are written for each combination of management protocol and resource to be
managed. For example, if an application is to be managed by SNMP and HTTP, two sets of instrumentation functions
are defined; one that maps SNMP requests to the resources, and one that, for example, generates an HTML page for
some resources.
When a manager makes a request to the agent, the following illustrates the situation:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 373
10.1 Introduction
Figure 1.2: Request to An Agent by a Manager
The mapping between an instrumentation function and a resource is not necessarily 1-1. It is also possible to write one
instrumentation function for each resource, and use that function from different protocols.
The agent receives a request and maps it to calls to one or more instrumentation functions. These functions perform
operations on the resources to implement the semantics associated with the MO.
For example, a system that is managed with SNMP and HTTP can be structured as follows:
374 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
10.1 Introduction
Figure 1.3: Structure of a System Managed with SNMP and HTTP
The resources can send notifications to the manager as well. Examples of notifications are events and alarms. The
resource needs to generate protocol-independent notifications. The following picture illustrates how this is achieved:
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 375
10.1 Introduction
Figure 1.4: Notification Handling
The main idea is that the resource sends the notifications as Erlang terms to a dedicated gen_event process. Into
this process, handlers for the different management protocols are installed. When an event is received by this process,
it is forwarded to each installed handler. The handlers are responsible for translating the event into a notification to be
sent over the management protocol. For example, a handler for SNMP translates each event into an SNMP trap.
10.1.3 SNMP-Based OAM
For all OAM components, SNMP adaptations are provided. Other adaptations might be defined in the future.
The OAM components, and some other OTP applications, define SNMP MIBs. These MIBs are written in SNMPv2
SMI syntax, as defined in RFC 1902. For convenience we also deliver the SNMPv1 SMI equivalent. All MIBs are
designed to be v1/v2 compatible, that is, the v2 MIBs do not use any construct not available in v1.
MIB Structure
The top-level OTP MIB is called OTP-REG and it is included in the SASL application. All other OTP MIBs import
some objects from this MIB.
376 | Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation
10.1 Introduction
Each MIB is contained in one application. The MIB text files are stored under mibs/<MIB>.mib in the application
directory. The generated .hrl files with constant declarations are stored under include/<MIB>.hrl, and the
compiled MIBs are stored under priv/mibs/<MIB>.bin. For example, the OTP-MIB is included in the SASL
application:
sasl-1.3/mibs/OTP-MIB.mib
include/OTP-MIB.hrl
priv/mibs/OTP-MIB.bin
An application that needs to import this MIB into another MIB is to use the il option to the SNMP MIB compiler:
snmp:c("MY-MIB", [{il, ["sasl/priv/mibs"]}]).
If the application needs to include the generated .hrl file, it is to use the -include_lib directive to the Erlang
compiler:
-module(my_mib).
-include_lib("sasl/include/OTP-MIB.hrl").
The following MIBs are defined in the OTP system:
•
•
•
•
•
•
•
•
•
OTP-REG) (in SASL) contains the top-level OTP registration objects, used by all other MIBs.
OTP-TC (in SASL) contains the general Textual Conventions, which can be used by any other MIB.
OTP-MIB (in SASL) contains objects for instrumentation of the Erlang nodes, the Erlang machines, and the
applications in the system.
OTP-OS-MON-MIB (in oc_mon) contains objects for instrumentation of disk, memory, and CPU use of the
nodes in the system.
OTP-SNMPEA-MIB (in snmp) contains objects for instrumentation and control of the extensible SNMP agent
itself. The agent also implements the standard SNMPv2-MIB (or v1 part of MIB-II, if SNMPv1 is used).
OTP-EVA-MIB (in eva) contains objects for instrumentation and control of the events and alarms in the system.
OTP-LOG-MIB (in eva) contains objects for instrumentation and control of the logs and FTP transfer of logs.
OTP-EVA-LOG-MIB (in eva) contains objects for instrumentation and control of the events and alarm logs in
the system.
OTP-SNMPEA-LOG-MIB (in eva) contains objects for instrumentation and control of the SNMP audit trail log
in the system.
The different applications use different strategies for loading the MIBs into the agent. Some MIB implementations are
code-only, while others need a server. One way, used by the code-only MIB implementations, is for the user to call
a function such as otp_mib:init(Agent) to load the MIB, and otp_mib:stop(Agent) to unload the MIB.
See the manual page for each application for a description of how to load each MIB.
Ericsson AB. All Rights Reserved.: Erlang/OTP System Documentation | 377