Ed J


Inline::Struct -- Manipulate C structures directly from Perl.


  use Inline C => 'DATA', structs => ['Foo'];

  my $obj = Inline::Struct::Foo->new;

  print myfunc($obj), "\n";

  struct Foo {
    int num;
    char *str;
  typedef struct Foo Foo;

  SV *myfunc(Foo *f) {
    return newSVpvf("myfunc: num=%i, str='%s'", f->num, f->str);

This complete program prints:

  myfunc: num=10, str='Hello'


Inline::Struct is not a new language. It's a language extension designed to be used by Inline::C. It parses struct definitions and creates typemaps and XS code which bind each struct into a Perl class. This code is passed to Inline::C, which compiles it in the normal way.

NOTE: Inline::Struct parses only C-style structs. It doesn't know about any C++ extensions to structs like scopes, constructors or methods. If you want such functionality you should use Inline::CPP to parse your structs.

Using Inline::Struct

Inline::Struct has a Parse::RecDescent grammar to parse C structs. If a struct is recognized, it can be bound to Perl. If the struct's definition is not recognized (usually because it has a member with no typemap), it will not be bound to Perl, but will be available from other functions in C or C++.

The following example shows how a simple struct might look to a Perl programmer.

Example 1:

  use Inline C => <<'END', enable => 'structs';
  struct Fraction {
    long numer;
    long denom;

  my $o = Inline::Struct::Fraction->new(4, 3);
  print $o->numer, $o->denom, "\n";

After the code above has been compiled, Perl's namespace looks a lot like the following:

  package Inline::Struct::Fraction;
  sub new { }
  sub DESTROY { }
  sub _KEYS { }
  sub _VALUES { }
  sub _HASH { }
  sub numer { }
  sub denom { }

Note that these are actually XS subs written in C, not Perl subs. But that's what it looks like.

The Struct Interface

The following sections define the interface of each subroutine. Note: this interface is likely to change in future versions of Inline::Struct. Please don't rely on Inline::Struct in production code quite yet.

When a struct is bound by Inline::Struct, a new namespace is created underneath Inline::Struct. So if you have a struct named 'Foo', the package of the Perl class will be 'Inline::Struct::Foo'.


If no arguments are provided, all fields are zeroed out. If you provide values, they should be appropriate for the field type, and in the same order as they are defined in the struct.


The destructor. Should never be called by the programmer -- this is called automatically when the Perl variable holding the struct is destroyed. Frees the memory associated with the struct. If the struct holds pointers to malloc'd memory, they will not be freed. If you run into such a situation, consider using C++ and Inline::CPP instead.


A read-only method, this returns a reference to an array containing the names of the fields in the struct. The fields are in the order they appear in the C source code.


A read-only method, this returns a reference to an array containing the values of the fields in the struct. The values are returned in the same order as the fields.


A read-only method, this returns a reference to a hash, mapping field names to field values.


For each field in the struct, an accessor method will be created which lets you get or set the value in the struct. If no arguments are provided, the method returns the value of that field. If any arguments are provided, the field is set to the first argument, and the modified structure is returned. This makes setting multiple fields easy:


C and C++ Configuration Options

Inline::Struct has no configuration options of its own, but it does provide a new configuration option for C or C++.


Specifies that structs are to be bound to Perl. There are several meanings to this option, so I'll explain with an example:

   use Inline C => config => structs => 'Foo';

Adds 'Foo' to the list of structs to bind to Perl.

   use Inline C => config => structs => ['Foo', 'Bar'];

Adds 'Foo' and 'Bar' to the list of structs to bind to Perl.

   use Inline C => config => structs => undef;

Clears the list of structs to bind to Perl.

   use Inline C => config => enable => 'structs';
   use Inline C => config => structs => 1;

Enable binding structs to Perl, without specifying any structs to search for. As shown, this would bind all structs to Perl.

   use Inline C => config => disable => 'structs';


   use Inline C => config => structs => 0;

Disable binding structs to Perl.



A script, benchmark, that benchmarks a simple C struct against a pure-Perl data structure, is supplied. It should be run a couple of times to get everything cached. A typical results run is as follows:

  Faster type          % faster
  ISF dnum read        24%
  PP dnum write        247%
  ISF inum read        39%
  PP inum write        231%
  ISF str read         18%
  PP str write         264%

This shows that reading the struct is faster than a simple object implemented as a hash-ref, while writing to a struct in the current implementation is several times slower. If the Perl object is instead implemented as an array-ref, in the class PP::Foo::Array, the numbers do not change significantly.


The same script also compares memory usage. A typical results run:

  Memory usage
  10000 x bless [ 7, "string" ], "main": 34592
  10000 x Inline::Struct::Foo->new: 45648
  100000 x bless [ 7, "string" ], "main": 139344
  100000 x Inline::Struct::Foo->new: 248080
  1000000 x bless [ 7, "string" ], "main": 1187024
  1000000 x Inline::Struct::Foo->new: 2257968

The memory usage of the struct is around twice as large.


For more information about using C from Perl, see Inline::C. For more information about using C++ from Perl, see Inline::CPP.


Neil Watkiss (NEILW@cpan.org)


Copyright (C) 2001, Neil Watkiss.

This module is free software. It may be used, redistributed and/or modified under the same terms as Perl itself.

See http://dev.perl.org/licenses/