<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">
<chapter>
<header>
<copyright>
<year>2000</year><year>2015</year>
<holder>Ericsson AB. All Rights Reserved.</holder>
</copyright>
<legalnotice>
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.
</legalnotice>
<title>NIFs</title>
<prepared></prepared>
<docno></docno>
<date></date>
<rev></rev>
<file>nif.xml</file>
</header>
<p>This section outlines an example of how to solve the example
problem in <seealso marker="example">Problem Example</seealso>
by using Native Implemented Functions (NIFs).</p>
<p>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 <c>foo</c> and <c>bar</c> in the example, that
do some relatively short calculations without side effects and
return the result.</p>
<p>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.</p>
<p>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.</p>
<section>
<title>Erlang Program</title>
<p>Even if all functions of a module are NIFs, an Erlang
module is still needed for two reasons:</p>
<list type="bulleted">
<item>The NIF library must be explicitly loaded by
Erlang code in the same module.</item>
<item>All NIFs of a module must have an Erlang implementation
as well.</item>
</list>
<p>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.</p>
<p>NIF libraries are loaded by calling <c>erlang:load_nif/2</c>,
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:</p>
<codeinclude file="complex6.erl" tag="" type="none"></codeinclude>
<p>Here, the directive <c>on_load</c> is used to get function
<c>init</c> to be automatically called when the module is
loaded. If <c>init</c> returns anything other than <c>ok</c>,
such when the loading of the NIF library fails in this example,
the module is unloaded and calls to functions within it,
fail.</p>
<p>Loading the NIF library overrides the stub implementations
and cause calls to <c>foo</c> and <c>bar</c> to be dispatched to
the NIF implementations instead.</p>
</section>
<section>
<title>NIF Library Code</title>
<p>The NIFs of the module are compiled and linked into a
shared library. Each NIF is implemented as a normal C function. The macro
<c>ERL_NIF_INIT</c> together with an array of structures defines the names,
arity, and function pointers of all the NIFs in the module. The header
file <c>erl_nif.h</c> must be included. As the library is a shared
module, not a program, no main function is to be present.</p>
<p>The function arguments passed to a NIF appears in an array <c>argv</c>,
with <c>argc</c> as the length of the array, and thus the arity of the
function. The Nth argument of the function can be accessed as
<c>argv[N-1]</c>. 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:</p>
<codeinclude file="complex6_nif.c" tag="" type="none"></codeinclude>
<p>Here,<c>ERL_NIF_INIT</c> has the following arguments:</p>
<list type="bulleted">
<item><p>The first argument must be the name of the
Erlang module as a C-identifier. It will be stringified by the
macro.</p>
</item>
<item>The second argument is the array of <c>ErlNifFunc</c>
structures containing name, arity, and function pointer of
each NIF.</item>
<item>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 <c>NULL</c>.</item>
</list>
<p>Function arguments and return values are represented as values
of type <c>ERL_NIF_TERM</c>. Here, functions like <c>enif_get_int</c>
and <c>enif_make_int</c> are used to convert between Erlang term
and C-type.
If the function argument <c>argv[0]</c> is not an integer,
<c>enif_get_int</c> returns false, in which case it returns
by throwing a <c>badarg</c>-exception with <c>enif_make_badarg</c>.</p>
</section>
<section>
<title>Running the Example</title>
<p><em>Step 1.</em> Compile the C code:</p>
<pre>
unix> <input>gcc -o complex6_nif.so -fpic -shared complex.c complex6_nif.c</input>
windows> <input>cl -LD -MD -Fe complex6_nif.dll complex.c complex6_nif.c</input></pre>
<p><em>Step 2:</em> Start Erlang and compile the Erlang code:</p>
<pre>
> <input>erl</input>
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> <input>c(complex6).</input>
{ok,complex6}</pre>
<p><em>Step 3:</em> Run the example:</p>
<pre>
3> <input>complex6:foo(3).</input>
4
4> <input>complex6:bar(5).</input>
10
5> <input>complex6:foo("not an integer").</input>
** exception error: bad argument
in function complex6:foo/1
called as comlpex6:foo("not an integer")
</pre>
</section>
</chapter>