|
NAMEFFI::Platypus::Bundle - Bundle foreign code with your Perl moduleVERSIONversion 1.56SYNOPSIS"ffi/foo.c":#include <ffi_platypus_bundle.h> #include <string.h> typedef struct { char *name; int value; } foo_t; foo_t* foo__new(const char *class_name, const char *name, int value) { (void)class_name; foo_t *self = malloc( sizeof( foo_t ) ); self->name = strdup(name); self->value = value; return self; } const char * foo__name(foo_t *self) { return self->name; } int foo__value(foo_t *self) { return self->value; } void foo__DESTROY(foo_t *self) { free(self->name); free(self); } "lib/Foo.pm": package Foo; use strict; use warnings; use FFI::Platypus 1.00; { my $ffi = FFI::Platypus->new( api => 1 ); $ffi->type('object(Foo)' => 'foo_t'); $ffi->mangler(sub { my $name = shift; $name =~ s/^/foo__/; $name; }); $ffi->bundle; $ffi->attach( new => [ 'string', 'string', 'int' ] => 'foo_t' ); $ffi->attach( name => [ 'foo_t' ] => 'string' ); $ffi->attach( value => [ 'foo_t' ] => 'int' ); $ffi->attach( DESTROY => [ 'foo_t' ] => 'void' ); } 1; "t/foo.t" use Test2::V0; use Foo; my $foo = Foo->new("platypus", 10); isa_ok $foo, 'Foo'; is $foo->name, "platypus"; is $foo->value, 10; done_testing; "Makefile.PL": use ExtUtils::MakeMaker; use FFI::Build::MM; my $fbmm = FFI::Build::MM->new; WriteMakefile( $fbmm->mm_args( NAME => 'Foo', DISTNAME => 'Foo', VERSION => '1.00', # ... ) ); sub MY::postamble { $fbmm->mm_postamble; } or "dist.ini": name = Foo version = 0.01 ... [FFI::Build] version = 1.04 DESCRIPTIONThis document serves as a tutorial for using the new bundling interface provided by FFI::Platypus as of api version 1. It requires FFI::Platypus of at least 1.00.Sometimes when writing FFI bindings you need to include a little C code (or your favorite compiled language) to finish things off. Alternatively, you might just want to write some C code (or your favorite compiled language) to include with your Perl module to make a tight loop faster. The bundling interface has you covered. Basic exampleTo illustrate we will go through the files in the synopsis and explain how and why they work. To start with we have some C code which emulates object oriented code using "foo__" as a prefix. We use a C struct that we call "foo_t" to store our object data. On the C level the struct acts as a class, when combined with its functions that act as methods. The constructor just allocates the memory it needs for the "foo_t" instance, fills in the appropriate fields and returns the pointer:foo_t* foo__new(const char *class_name, const char *name, int value) { (void) class_name; foo_t *self = malloc( sizeof( foo_t ) ); self->name = strdup(name); self->value = value; return self; } We include a class name as the first argument, because Perl will include that when calling the constructor, but we do not use it here. An exercise for the reader would be to add hierarchical inheritance. There are also some methods which return member values. This class has only read only members, but you could have read/write or other methods depending on your needs. const char * foo__name(foo_t *self) { return self->name; } We also include a destructor so that the memory owned by the object can be freed when it is no longer needed. void foo__DESTROY(foo_t *self) { free(self->name); free(self); } This might start to look a little like a Perl module, and when we look at the Perl code that binds to this code, you will see why. First lets prepare the FFI::Platypus instance and specify the correct api version: my $ffi = FFI::Platypus->new( api => 1 ); The bundle interface is only supported with api version 1, so if you try to use version 0 it will not work. Next we define an object type for "foo_t" which will associate it with the Perl class "Foo". $ffi->type('object(Foo)' => 'foo_t'); As object type is a blessed reference to an opaque (default) or integer type which can be used as a Perl object. Platypus does the translating of Perl object to and from the foo_t pointers that the C code understands. For more details on Platypus types see FFI::Platypus::Type. Next we set the mangler on the Platypus instance so that we can refer to function names without the "foo__" prefix. You could just not use the prefix in your C code and skip this step, or you could refer to the function names in their full in your Perl code, however, this saves extra typing and allows you to bundle more than one class with your Perl code without having to worry about name conflicts. $ffi->mangler(sub { my $name = shift; $name =~ s/^/foo__/; $name; }); Finally we let Platypus know that we will be bundling code. $ffi->bundle; By default, this searches for the appropriate place for your dynamic libraries using the current package. In some cases you may need to override this, for example if your dist is named "Foo-Bar" but your specific class is named "Foo::Bar::Baz", you'd want something like this: package Foo::Bar::Baz; use FFI::Platypus 1.00; my $ffi = FFI::Platypus->new( api => 1 ); $ffi->bundle('Foo::Bar'); ... Now, finally we can attach the methods for our class: $ffi->attach( new => [ 'string', 'int' ] => 'foo_t' ); $ffi->attach( name => [ 'foo_t' ] => 'string' ); $ffi->attach( value => [ 'foo_t' ] => 'int' ); $ffi->attach( DESTROY => [ 'foo_t' ] => 'void' ); Note that we do not have to include the "foo__" prefix because of the way we set up the mangler. If we hadn't done that then we could instead attach with the full names: $ffi->attach( [ 'foo__new' => 'new' ] => [ 'string', 'int' ] => 'foo_t' ); $ffi->attach( [ 'foo__name' => 'name' ] => [ 'foo_t' ] => 'string' ); ... You're done! You can now use this class. Lets write a test to make sure it works, use Test2::V0; use Foo; my $foo = Foo->new("platypus", 10); isa_ok $foo, 'Foo'; is $foo->name, "platypus"; is $foo->value, 10; done_testing; and use "prove" to check that it works: % prove -lvm t/foo.t .. ok 1 - An object of class 'Foo' isa 'Foo' ok 2 ok 3 1..3 ok All tests successful. Files=1, Tests=3, 0 wallclock secs ( 0.02 usr 0.00 sys + 0.14 cusr 0.03 csys = 0.19 CPU) Result: PASS Platypus automatically compiles and links the dynamic library for you: % ls ffi/_build foo.c.o libFoo.so The C code will be rebuilt next time if the source code is newer than the object or dynamic libraries files. If the source files are not changed, then it won't be rebuilt to save time. If you are using the code without MakeMaker, or another build system you are responsible for cleaning up these files. This is intended as a convenience to allow you to test your code without having to invoke MakeMaker, or "dzil" or whatever build system you are using. When you distribute your module though, you will want the dynamic library built just once at build-time and installed correctly so that it can be found at run-time. You don't need to make any changes to your C or Perl code, but you do need to tell MakeMaker to build and install the appropriate files using FFI::Build::MM: use ExtUtils::MakeMaker; use FFI::Build::MM; my $fbmm = FFI::Build::MM->new; WriteMakefile( $fbmm->mm_args( NAME => 'Foo', DISTNAME => 'Foo', VERSION => '1.00', # ... ) ); sub MY::postamble { $fbmm->mm_postamble; } And we can invoke all the normal MakeMaker style stuff and our C code will be compiled, linked and installed at the appropriate steps. % perl Makefile.PL Generating a Unix-style Makefile Writing Makefile for Foo Writing MYMETA.yml and MYMETA.json % make cp lib/Foo.pm blib/lib/Foo.pm "/Users/ollisg/perl5/perlbrew/perls/perl-5.30.0/bin/perl" -MFFI::Build::MM=cmd -e fbx_build CC ffi/foo.c LD blib/lib/auto/share/dist/Foo/lib/libFoo.dylib % make test "/Users/ollisg/perl5/perlbrew/perls/perl-5.30.0/bin/perl" -MFFI::Build::MM=cmd -e fbx_build "/Users/ollisg/perl5/perlbrew/perls/perl-5.30.0/bin/perl" -MFFI::Build::MM=cmd -e fbx_test PERL_DL_NONLAZY=1 "/Users/ollisg/perl5/perlbrew/perls/perl-5.30.0/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/foo.t .. ok All tests successful. Files=1, Tests=3, 0 wallclock secs ( 0.01 usr 0.00 sys + 0.06 cusr 0.01 csys = 0.08 CPU) Result: PASS If the "Makefile.PL" file above looks overly complicated, you can use the Dist::Zilla::Plugin::FFI::Build plugin to simplify your life if you are using Dist::Zilla: [FFI::Build] version = 1.04 Specifying version 1.04 will ensure that any ".o" or ".so" files are pruned from your build tree and not distributed by mistake. Initialization exampleThe bundle interface also gives you entry points which will be called automatically when your code is loaded and unloaded if they are found.
Here is an example that passes the version and a callback back into Perl space that emulates the Perl 5.10 "say" feature. "ffi/init.c": #include <ffi_platypus_bundle.h> char buffer[512]; const char *version; void (*say)(const char *); void ffi_pl_bundle_init(const char *package, int argc, void *argv[]) { version = argv[0]; say = argv[1]; say("in init!"); snprintf(buffer, 512, "package = %s, version = %s", package, version); say(buffer); snprintf(buffer, 512, "args = %d", argc); say(buffer); } void ffi_pl_bundle_fini(const char *package) { say("in fini!"); } "lib/Init.pm": package Init; use strict; use warnings; use FFI::Platypus 1.00; our $VERSION = '1.00'; { my $ffi = FFI::Platypus->new( api => 1 ); my $say = $ffi->closure(sub { my $string = shift; print "$string\n"; }); $ffi->bundle([ $ffi->cast( 'string' => 'opaque', $VERSION ), $ffi->cast( '(string)->void' => 'opaque', $say ), ]); undef $ffi; undef $say; } 1; The deinitialization order for the $say callback and the $ffi instance is essential here, so we do it manually with "undef": undef $ffi; undef $say; First we deallocate $ffi which calls "ffi_pl_bundle_fini", which calls $say, so we want to make sure the latter is still allocated. Once "ffi_pl_bundle_fini" is done, we can safely deallocate $say. If "ffi_pl_bundle_fini" didn't call back into Perl space like this then we don't have to be as careful about deallocating things in Perl space. Compiler or linker flags exampleThere are times when you will want to specify your own compiler and linker flags for the C code that you are bundling. The "TL;DR" is that you can put a ".fbx" file in your "ffi" directory. This is a Perl script that returns a hash reference that is passed into the FFI::Build constructor. This allows you to set a number of options, including compiler and linker flags. A more detailed example follows:You may want or need to set compiler and linker flags for your bundled C code. For example, say we have a header file, but instead of putting it in the "ffi" directory we want to put it in a separate directory called "include". "include/answer.h": #ifndef ANSWER_H #define ANSWER_H int answer(void); #endif "ffi/answer.c": #include <answer.h> int answer(void) { /* the answer to life the universe and everything */ return 42; } "lib/Answer.pm": package Answer; use strict; use warnings; use FFI::Platypus 1.00; use Exporter qw( import ); our @EXPORT = qw( answer ); my $ffi = FFI::Platypus->new( api => 1 ); $ffi->bundle; $ffi->attach( answer => [] => 'int' ); 1; If you try to use this module just as-is you will get an error, about not being able to find the header file. Probably something like this: ffi/answer.c:1:10: fatal error: 'answer.h' file not found So we put a "answer.fbx" file in the "ffi" directory. (In case you are wondering FBX stands for "Ffi Build and file eXtensions should whenever possible be three characters long"). The name of the file can be anything so long as it ends in ".fbx", we just choose "answer" here because that is the name of the project. "ffi/answer.fbx": our $DIR; return { cflags => "-I/include", source => "$DIR/*.c", } The $DIR variable is provided by the builder code. It is the root of the distribution, and is helpful if you need a fully qualified path. In this case you could have also used "ffi/*.c". The script returns a hash reference which is passed into the FFI::Build constructor, so you can use any of the options supported by that class. Now we should be able to use our bundled module: % perl -Ilib -MAnswer=answer -E 'say answer' 42 Using bundled code with Alien.A useful technique is to use Platypus with Alien technology. The Alien namespace is reserved for providing external non-Perl dependencies for CPAN modules. The nominal Alien module when installed looks for the library locally, and if it can't be found it fetches it from the internet, builds it, and installs it in a private directory so that it can be used by other CPAN modules. For Aliens that provide shared libraries, and that have simple interfaces that do not require additional C code you can easily just pass the shared libraries to Platypus directly. For modules that require some bundled C code and an Alien you have to link the Alien library with your bundled code. If the Alien uses the Alien::Base interface then all you have to do is give the name of the Alien to FFI::Build.For example, the "bzip2" library provides an interface that requires the caller to allocate a C "struct" and then pass it to its various functions. The "struct" is actually pretty simple and you could use FFI::C or FFI::Platypus::Record, but here is an example of how you would connect bundled C code with an Alien. "ffi/compress.c": #include <bzlib.h> #include <stdlib.h> int bzip2__new(bz_stream **stream, int blockSize100k, int verbosity, int workFactor ) { *stream = malloc(sizeof(bz_stream)); (*stream)->bzalloc = NULL; (*stream)->bzfree = NULL; (*stream)->opaque = NULL; return BZ2_bzCompressInit(*stream, blockSize100k, verbosity, workFactor ); } "lib/Bzip2.pm": package Bzip2; use strict; use warnings; use FFI::Platypus 1.00; use FFI::Platypus::Memory qw( free ); my $ffi = FFI::Platypus->new( api => 1 ); $ffi->bundle; $ffi->mangler(sub { my $name = shift; $name =~ s/^/bzip2__/ unless $name =~ /^BZ2_/; $name; }); =head2 new my $bzip2 = Bzip2->new($block_size_100k, $verbosity, $work_flow); =cut $ffi->attach( new => ['opaque*', 'int', 'int', 'int'] => 'int' => sub { my $xsub = shift; my $class = shift; my $ptr; my $ret = $xsub->(\$ptr, @_); return bless \$ptr, $class; }); $ffi->attach( [ BZ2_bzCompressEnd => 'DESTROY' ] => ['opaque'] => 'int' => sub { my $xsub = shift; my $self = shift; my $ret = $xsub->($$self); free $$self; }); 1; The ".fbx" file that goes with this to make it work with Alien::Libbz2 is now pretty trivial: "ffi/bz2.fbx": { alien => ['Alien::Libbz2'], source => ['ffi/*.c'], }; AUTHORAuthor: Graham Ollis <plicease@cpan.org>Contributors: Bakkiaraj Murugesan (bakkiaraj) Dylan Cali (calid) pipcet Zaki Mughal (zmughal) Fitz Elliott (felliott) Vickenty Fesunov (vyf) Gregor Herrmann (gregoa) Shlomi Fish (shlomif) Damyan Ivanov Ilya Pavlov (Ilya33) Petr Písař (ppisar) Mohammad S Anwar (MANWAR) Håkon Hægland (hakonhagland, HAKONH) Meredith (merrilymeredith, MHOWARD) Diab Jerius (DJERIUS) Eric Brine (IKEGAMI) szTheory José Joaquín Atria (JJATRIA) Pete Houston (openstrike, HOUSTON) COPYRIGHT AND LICENSEThis software is copyright (c) 2015,2016,2017,2018,2019,2020 by Graham Ollis.This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.
Visit the GSP FreeBSD Man Page Interface. |