The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.
 / 
Affix-0.10_03-TRIAL
River stage zero No dependents
2 ++
/ Affix

NAME

Affix - A Foreign Function Interface eXtension

SYNOPSIS

    use Affix;

    affix( 'libfoo', 'bar', [Str, Float] => Double );
    print bar( 'Baz', 3.14 );

    # or

    my $bar = wrap( 'libfoo', 'bar', [Str, Float] => Double );
    print $bar->( 'Baz', 3.14 );

    # or

    sub bar : Native('libfoo') : Signature([Str, Float] => Double);
    print bar( 'Baz', 10.9 );

    # bind to exported values

DESCRIPTION

Affix is a wrapper around dyncall. If you're looking to design your own low level FFI, see Dyn.pm.

But if you're just looking for a fast FFI system, keep reading.

Note: This is experimental software and is subject to change as long as this disclaimer is here.

Basic Usage

The basic API here is rather simple but not lacking in power.

affix( ... )

    affix( 'C:\Windows\System32\user32.dll', 'pow', [Double, Double] => Double );
    warn pow( 3, 5 );

Attaches a given symbol in a named perl sub.

Parameters include:

$lib

path of the library as a string or pointer returned by dlLoadLibrary( ... )

$symbol_name

the name of the symbol to call

$parameters

signature defining argument types in an array

$return

return type

$name

optional name of affixed sub; $symbol_name by default

Returns a code reference on success.

wrap( ... )

Creates a wrapper around a given symbol in a given library.

    my $pow = wrap( 'C:\Windows\System32\user32.dll', 'pow', [Double, Double] => Double );
    warn $pow->(5, 10); # 5**10

Parameters include:

$lib

pointer returned by dlLoadLibrary( ... ) or the path of the library as a string

$symbol_name

the name of the symbol to call

$parameters

signature defining argument types in an array

$return

return type

wrap( ... ) behaves exactly like affix( ... ) but returns an anonymous subroutine.

:Native CODE attribute

All the sugar is right here in the :Native code attribute. This API is inspired by Raku's native trait.

A simple example would look like this:

    use Affix;
    sub some_argless_function :Native('something');
    some_argless_function();

The first line imports various code attributes and types. The next line looks like a relatively ordinary Perl sub declaration--with a twist. We use the :Native attribute in order to specify that the sub is actually defined in a native library. The platform-specific extension (e.g., .so or .dll), as well as any customary prefixes (e.g., lib) will be added for you.

The first time you call "some_argless_function", the "libsomething" will be loaded and the "some_argless_function" will be located in it. A call will then be made. Subsequent calls will be faster, since the symbol handle is retained.

Of course, most functions take arguments or return values--but everything else that you can do is just adding to this simple pattern of declaring a Perl sub, naming it after the symbol you want to call and marking it with the :Native-related attributes.

Except in the case you are using your own compiled libraries, or any other kind of bundled library, shared libraries are versioned, i.e., they will be in a file libfoo.so.x.y.z, and this shared library will be symlinked to libfoo.so.x. By default, Affix will pick up that file if it's the only existing one. This is why it's safer, and advisable, to always include a version, this way:

    sub some_argless_function :Native('foo', v1.2.3)

Please check the section on the ABI/API version for more information.

Changing names

Sometimes you want the name of your Perl subroutine to be different from the name used in the library you're loading. Maybe the name is long or has different casing or is otherwise cumbersome within the context of the module you are trying to create.

Affix provides the :Symbol attribute for you to specify the name of the native routine in your library that may be different from your Perl subroutine name.

    package Foo;
    use Affix;
    sub init :Native('foo') :Symbol('FOO_INIT');

Inside of libfoo there is a routine called FOO_INIT but, since we're creating a module called Foo and we'd rather call the routine as Foo::init (instead of Foo::FOO_INIT), we use the symbol trait to specify the name of the symbol in libfoo and call the subroutine whatever we want (init in this case).

Passing and returning values

Normal Perl signatures do not convey the type of arguments a native function expects and what it returns so you must define them with our final attribute: :Signature

    use Affix;
    sub add :Native("calculator") :Signature([Int, Int] => Int);

Here, we have declared that the function takes two 32-bit integers and returns a 32-bit integer. You can find the other types that you may pass further down this page.

Signatures

Affix's advisory signatures are required to give us a little hint about what we should expect.

    [ Int, ArrayRef[ Int, 100 ], Str ] => Int

Arguments are defined in a list: [ Int, ArrayRef[ Char, 5 ], Str ]

The return value comes next: Int

To call the function with such a signature, your Perl would look like this:

    mh $int = func( 500, [ 'a', 'b', 'x', '4', 'H' ], 'Test');

See the aptly named sections entitled Types for more on the possible types and "Calling Conventions" in Calling Conventions for flags that may also be defined as part of your signature.

Library Paths and Names

The :Native attribute, affix( ... ), and wrap( ... ) all accept the library name, the full path, or a subroutine returning either of the two. When using the library name, the name is assumed to be prepended with lib and appended with .so (or just appended with .dll on Windows), and will be searched for in the paths in the LD_LIBRARY_PATH (PATH on Windows) environment variable.

You can also put an incomplete path like './foo' and Affix will automatically put the right extension according to the platform specification. If you wish to suppress this expansion, simply pass the string as the body of a block.

    sub bar :Native({ './lib/Non Standard Naming Scheme' });

BE CAREFUL: the :Native attribute and constant might be evaluated at compile time.

ABI/API version

If you write :Native('foo'), Affix will search libfoo.so under Unix like system (libfoo.dynlib on macOS, foo.dll on Windows). In most modern system it will require you or the user of your module to install the development package because it's recommended to always provide an API/ABI version to a shared library, so libfoo.so ends often being a symbolic link provided only by a development package.

To avoid that, the :Native attribute allows you to specify the API/ABI version. It can be a full version or just a part of it. (Try to stick to Major version, some BSD code does not care for Minor.)

    use Affix;
    sub foo1 :Native('foo', v1); # Will try to load libfoo.so.1
    sub foo2 :Native('foo', v1.2.3); # Will try to load libfoo.so.1.2.3

    sub pow : Native('m', v6) : Signature([Double, Double] => Double);

Calling into the standard library

If you want to call a function that's already loaded, either from the standard library or from your own program, you can omit the library value or pass and explicit undef.

For example on a UNIX-like operating system, you could use the following code to print the home directory of the current user:

    use Affix;
    use Data::Dumper;
    typedef PwStruct => Struct [
        name  => Str,     # username
        pass  => Str,     # hashed pass if shadow db isn't in use
        uuid  => UInt,    # user
        guid  => UInt,    # group
        gecos => Str,     # real name
        dir   => Str,     # ~/
        shell => Str      # bash, etc.
    ];
    sub getuid : Native : Signature([]=>Int);
    sub getpwuid : Native : Signature([Int]=>Pointer[PwStruct]);
    my $data = main::getpwuid( getuid() );
    print Dumper( ptr2sv( $data, Pointer [ PwStruct() ] ) );

Exported Variables

Variables exported by a library - also names "global" or "extern" variables - can be accessed using pin( ... ).

pin( ... )

    pin( $errno, 'libc', 'errno', Int );
    print $errno;
    $errno = 0;

This code applies magic to $error that binds it to the integer variable named "errno" as exported by the libc library.

Expected parameters include:

$var

Perl scalar that will be bound to the exported variable.

$lib

pointer returned by dlLoadLibrary( ... ) or the path of the library as a string

$symbol_name

the name of the exported variable

$type

type that data will be coerced in or out of as required

This is likely broken on BSD. Patches welcome.

Memory Functions

To help toss raw data around, some standard memory related functions are exposed here. You may import them by name or with the :memory or :all tags.

malloc( ... )

    my $ptr = malloc( $size );

Allocates $size bytes of uninitialized storage.

calloc( ... )

    my $ptr = calloc( $num, $size );

Allocates memory for an array of $num objects of $size and initializes all bytes in the allocated storage to zero.

realloc( ... )

    $ptr = realloc( $ptr, $new_size );

Reallocates the given area of memory. It must be previously allocated by malloc( ... ), calloc( ... ), or realloc( ... ) and not yet freed with a call to free( ... ) or realloc( ... ). Otherwise, the results are undefined.

free( ... )

    free( $ptr );

Deallocates the space previously allocated by malloc( ... ), calloc( ... ), or realloc( ... ).

memchr( ... )

    memchr( $ptr, $ch, $count );

Finds the first occurrence of $ch in the initial $count bytes (each interpreted as unsigned char) of the object pointed to by $ptr.

memcmp( ... )

    my $cmp = memcmp( $lhs, $rhs, $count );

Compares the first $count bytes of the objects pointed to by $lhs and $rhs. The comparison is done lexicographically.

memset( ... )

    memset( $dest, $ch, $count );

Copies the value $ch into each of the first $count characters of the object pointed to by $dest.

memcpy( ... )

    memcpy( $dest, $src, $count );

Copies $count characters from the object pointed to by $src to the object pointed to by $dest.

memmove( ... )

    memmove( $dest, $src, $count );

Copies $count characters from the object pointed to by $src to the object pointed to by $dest.

sizeof( ... )

    my $size = sizeof( Int );
    my $size1 = sizeof( Struct[ name => Str, age => Int ] );

Returns the size, in bytes, of the type passed to it.

offsetof( ... )

    my $struct = Struct[ name => Str, age => Int ];
    my $offset = offsetof( $struct, 'age' );

Returns the offset, in bytes, from the beginning of a structure including padding, if any.

Utility Functions

Here's some thin cushions for the rougher edges of wrapping libraries.

They may be imported by name for now but might be renamed, removed, or changed in the future.

cast( ... )

    my $hash = cast( $ptr, Struct[i => Int, ... ] );

This function will parse a pointer into a given target type.

The source pointer would have normally been obtained from a call to a native subroutine that returned a pointer, a lvalue pointer to a native subroutine, or, as part of a Struct[ ... ].

DumpHex( ... )

    DumpHex( $ptr, $length );

Dumps $length bytes of raw data from a given point in memory.

This is a debugging function that probably shouldn't find its way into your code and might not be public in the future.

Types

While Raku offers a set of native types with a fixed, and known, representation in memory but this is Perl so we need to do the work ourselves and design and build a pseudo-type system. Affix supports the fundamental types (void, int, etc.) and aggregates (struct, array, union).

Fundamental Types with Native Representation

    Affix       C99/C++     Rust    C#          pack()  Raku
    -----------------------------------------------------------------------
    Void        void/NULL   ->()    void/NULL   -
    Bool        _Bool       bool    bool        -       bool
    Char        int8_t      i8      sbyte       c       int8
    UChar       uint8_t     u8      byte        C       byte, uint8
    Short       int16_t     i16     short       s       int16
    UShort      uint16_t    u16     ushort      S       uint16
    Int         int32_t     i32     int         i       int32
    UInt        uint32_t    u32     uint        I       uint32
    Long        int64_t     i64     long        l       int64, long
    ULong       uint64_t    u64     ulong       L       uint64, ulong
    LongLong    -           i128                q       longlong
    ULongLong   -           u128                Q       ulonglong
    Float       float       f32                 f       num32
    Double      double      f64                 d       num64
    SSize_t     SSize_t                                 SSize_t
    Size_t      size_t                                  size_t
    Str         char *

Given sizes are minimums measured in bits

Void

The Void type corresponds to the C void type. It is generally found in typed pointers representing the equivalent to the void * pointer in C.

    sub malloc :Native :Signature([Size_t] => Pointer[Void]);
    my $data = malloc( 32 );

As the example shows, it's represented by a parameterized Pointer[ ... ] type, using as parameter whatever the original pointer is pointing to (in this case, void). This role represents native pointers, and can be used wherever they need to be represented in a Perl script.

In addition, you may place a Void in your signature to skip a passed argument.

Bool

Boolean type may only have room for one of two values: true or false.

Char

Signed character. It's guaranteed to have a width of at least 8 bits.

Pointers (Pointer[Char]) might be better expressed with a Str.

UChar

Unsigned character. It's guaranteed to have a width of at least 8 bits.

Short

Signed short integer. It's guaranteed to have a width of at least 16 bits.

UShort

Unsigned short integer. It's guaranteed to have a width of at least 16 bits.

Int

Basic signed integer type.

It's guaranteed to have a width of at least 16 bits. However, on 32/64 bit systems it is almost exclusively guaranteed to have width of at least 32 bits.

UInt

Basic unsigned integer type.

It's guaranteed to have a width of at least 16 bits. However, on 32/64 bit systems it is almost exclusively guaranteed to have width of at least 32 bits.

Long

Signed long integer type. It's guaranteed to have a width of at least 32 bits.

ULong

Unsigned long integer type. It's guaranteed to have a width of at least 32 bits.

LongLong

Signed long long integer type. It's guaranteed to have a width of at least 64 bits.

ULongLong

Unsigned long long integer type. It's guaranteed to have a width of at least 64 bits.

Float

Single precision floating-point type.

Double

Double precision floating-point type.

SSize_t

Signed integer type.

Size_t

Unsigned integer type often expected as the result of sizeof or offsetof but can be found elsewhere.

Str

Automatically handle null terminated character pointers with this rather than trying using Pointer[Char] and doing it yourself.

You'll learn a bit more about Pointer[...] and other parameterized types in the next section.

Parameterized Types

Some types must be provided with more context data.

Pointer[ ... ]

    Pointer[Int]  ~~ int *
    Pointer[Void] ~~ void *

Create pointers to (almost) all other defined types including Struct and Void.

To handle a pointer to an object, see InstanceOf.

Void pointers (Pointer[Void]) might be created with malloc and other memory related functions.

Struct[ ... ]

    Struct[                    struct {
        dob => Struct[              struct {
            year  => Int,               int year;
            month => Int,   ~~          int month;
            day   => Int                int day;
        ],                          } dob;
        name => Str,                char *name;
        wId  => Long                long wId;
    ];                          };

A struct consists of a sequence of members with storage allocated in an ordered sequence (as opposed to Union, which is a type consisting of a sequence of members where storage overlaps).

A C struct that looks like this:

    struct {
        char *make;
        char *model;
        int   year;
    };

...would be defined this way:

    Struct[
        make  => Str,
        model => Str,
        year  => Int
    ];

All fundamental and aggregate types may be found inside of a Struct.

ArrayRef[ ... ]

The elements of the array must pass the additional size constraint.

An array length must be given:

    ArrayRef[Int, 5];   # int arr[5]
    ArrayRef[Any, 20];  # SV * arr[20]
    ArrayRef[Char, 5];  # char arr[5]
    ArrayRef[Str, 10];  # char *arr[10]

Union[ ... ]

A union is a type consisting of a sequence of members with overlapping storage (as opposed to Struct, which is a type consisting of a sequence of members whose storage is allocated in an ordered sequence).

The value of at most one of the members can be stored in a union at any one time and the union is only as big as necessary to hold its largest member (additional unnamed trailing padding may also be added). The other members are allocated in the same bytes as part of that largest member.

A C union that looks like this:

    union {
        char  c[5];
        float f;
    };

...would be defined this way:

    Union[
        c => ArrayRef[Char, 5],
        f => Float
    ];

CodeRef[ ... ]

A value where ref($value) equals CODE. This would be how callbacks are defined.

The argument list and return value must be defined. For example, CodeRef[[Int, Int]=Int]> ~~ typedef int (*fuc)(int a, int b);; that is to say our function accepts two integers and returns an integer.

    CodeRef[[] => Void];                # typedef void (*function)();
    CodeRef[[Pointer[Int]] => Int];     # typedef Int (*function)(int * a);
    CodeRef[[Str, Int] => Struct[...]]; # typedef struct Person (*function)(chat * name, int age);

InstanceOf[ ... ]

    InstanceOf['Some::Class']

A blessed object of a certain type. When used as an lvalue, the result is properly blessed. As an rvalue, the reference is checked to be a subclass of the given package.

Any

Anything you dump here will be passed along unmodified. We hand off a pointer to the SV* perl gives us without copying it.

Enum[ ... ]

The value of an Enum is defined by its underlying type which includes Int, Char, etc.

This type is declared with an list of strings.

    Enum[ 'ALPHA', 'BETA' ];
    # ALPHA = 0
    # BETA  = 1

Unless an enumeration constant is defined in an array reference, its value is the value one greater than the value of the previous enumerator in the same enumeration. The value of the first enumerator (if it is not defined) is zero.

    Enum[ 'A', 'B', [C => 10], 'D', [E => 1], 'F', [G => 'F + C'] ];
    # A = 0
    # B = 1
    # C = 10
    # D = 11
    # E = 1
    # F = 2
    # G = 12

    Enum[ [ one => 'a' ], 'two', [ 'three' => 'one' ] ]
    # one   = a
    # two   = b
    # three = a

As you can see, enum values may allude to earlier defined values and even basic arithmetic is supported.

Additionally, if you typedef the enum into a given namespace, you may refer to elements by name. They are defined as dualvars so that works:

    typedef color => Enum[ 'RED', 'GREEN', 'BLUE' ];
    print color::RED();     # RED
    print int color::RED(); # 0

IntEnum[ ... ]

Same as Enum.

UIntEnum[ ... ]

Enum but with unsigned integers.

CharEnum[ ... ]

Enum but with signed chars.

Calling Conventions

Handle with care! Using these without understanding them can break your code!

Refer to the dyncall manual, http://www.angelcode.com/dev/callconv/callconv.html, https://en.wikipedia.org/wiki/Calling_convention, and your local university's Comp Sci department for a deeper explanation.

Anyway, here are the current options:

CC_DEFAULT
CC_THISCALL
CC_ELLIPSIS
CC_ELLIPSIS_VARARGS
CC_CDECL
CC_STDCALL
CC_FASTCALL_MS
CC_FASTCALL_GNU
CC_THISCALL_MS
CC_THISCALL_GNU
CC_ARM_ARM
CC_ARM_THUMB
CC_SYSCALL

When used in "Signatures" in signatures, most of these cause the internal argument stack to be reset. The exception is CC_ELLIPSIS_VARARGS which is used prior to binding varargs of variadic functions.

Examples

The best example of use might be LibUI. Brief examples will be found in eg/. Very short examples might find their way here.

Microsoft Windows

Here is an example of a Windows API call:

    use Affix;
    sub MessageBoxA :Native('user32') :Signature([Int, Str, Str, Int] => Int);
    MessageBoxA(0, "We have NativeCall", "ohai", 64);

Short tutorial on calling a C function

This is an example for calling a standard function and using the returned information.

getaddrinfo is a POSIX standard function for obtaining network information about a network node, e.g., google.com. It is an interesting function to look at because it illustrates a number of the elements of Affix.

The Linux manual provides the following information about the C callable function:

    int getaddrinfo(const char *node, const char *service,
       const struct addrinfo *hints,
       struct addrinfo **res);

The function returns a response code 0 for success and 1 for error. The data are extracted from a linked list of addrinfo elements, with the first element pointed to by res.

From the table of Affix types we know that an int is Int. We also know that a char * is best expressed with Str. But addrinfo is a structure, which means we will need to write our own type class. However, the function declaration is straightforward:

    TODO

Features

Not all features of dyncall are supported on all platforms, for those, the underlying library defines macros you can use to detect support. These values are exposed under the Affix::Feature package:

Affix::Feature::Syscall()

If true, your platform supports a syscall calling conventions.

Affix::Feature::AggrByVal()

If true, your platform supports passing around aggregates (struct, union) by value.

See Also

Check out FFI::Platypus for a more robust and mature FFI.

Examples found in eg/.

LibUI for a larger demo project based on Affix

Types::Standard for the inspiration of the advisory types system.

LICENSE

Copyright (C) Sanko Robinson.

This library is free software; you can redistribute it and/or modify it under the terms found in the Artistic License 2. Other copyrights, terms, and conditions may apply to data transmitted through this module.

AUTHOR

Sanko Robinson <sanko@cpan.org>