API CookBook

This is a collection of recipes for mixing C and Perl objects, methods, and classes. The reader should be familiar with the XS language and with the process of creating new extensions. The XS language is described in perlxs, and perlxstut is a tutorial describing how to create new extensions. The reader should also be familiar with Perl's internal data types and API macros and functions (i.e. SV, HV, AV, SvIV, newRV) as described in perlguts and perlcall.

Linking a C struct to a blessed or unblessed Perl object.

Ex1 demonstrates how to write a Perl constructor in C. The constructor allocates a C structure and creates an opaque Perl object to hold an address to the structure. Methods are implemented in C and Perl sends messages to the opaque object. A destructor for the opaque C struct is implemented in C.

Creating and destroying objects.

Ex1 and Ex6 demonstrate how to create and destroy object types that Perl doesn't know about, such as C structures. Ex3 demonstrates the same with object types that Perl knows about, such as HV or AV types. A destructor is needed only when Perl does not know about the object's type.

Creating a Perl subclass of a C superclass.

Ex1, Ex3, Ex4, and Ex6 implement Perl classes in C. The test Ex1/t/test1.t demonstrates a Perl-based subclass of a C-based superclass and Ex3/t/test1.t and Ex4/t/test1.t demonstrate a C-based subclass of a C-based superclass.

Creating a C subclass of a Perl superclass.

This is similar to the C-based subclass of a C-based superclass example found in Ex4.

Sharing a variable between Perl and C.

Ex2 demonstrates how to let Perl and C share variables. A variable can live in Perl or it can live in C. A Perl tie() is demonstrated.

Using a Perl hash object in C.

Ex3 demonstrates how C can use a Perl hash object. The hash can be a blessed or unblessed object and be created by Perl or by C. Store and fetch operations are demonstrated. The example shows how to control an object's refcount in a C-based constructor.

Using a Perl array object in C.

Ex4 demonstrates how C can use a Perl array object. The array can be a blessed or unblessed object and be created by Perl or by C. Push and pop operations are demonstrated. The example shows how to control an object's refcount in a C-based constructor.

Using a C array in Perl.

This is best done by having a C function copy the C array into a Perl array or by supplying C functions that Perl can call to request each element of the array. Consult the examples in Ex4 for using Perl array objects in C.

Illuminating an opaque C object for Perl.

Ex1 demonstrates a C struct with accessor functions for its fields. This allows Perl to know the names of the structure fields and to access those fields.

Illuminating an opaque Perl object for C.

The Perl object will usually be an array object or a hash object. Ex3 and Ex4 demonstrate how to fiddle these objects in C.

Sending messages to Perl objects from C.

This topic is addressed in detail in perlcall but a twist is covered here. The perlcall manpage assumes it has a reference to a blessed Perl scalar object. Ex5 uses a Perl array object, and it's already been dereferenced by the typemap so the XSUB can fiddle the array. To use the array object in a method call a reference must be created and the array object's reference count must be watched so it doesn't leak.

Storing a Perl object inside a C object.

Ex6 shows how to store a reference to a Perl object inside a C object. The example discusses the reference counting issues for the Perl object; when to increment, when to decrement, and why.

Creating double-typed Perl variables.

Ex_SDV demonstrates a way to have a Perl scalar variable be context-sensitive. In string context the variable will print a string and in numeric context the variable will print a number. This is similar to the behavior of the $] variable in Perl4.

Sharing a char** type between C and Perl

Ex8 shows two ways to share a char** type between C and Perl. The first approach uses an opaque object and the second uses full illumination via XS_pack_* and XS_unpack_* functions.

Linking Perl and C++ classes and objects.

Ex7 demonstrates how to create a Perl interface to a C++ class. The example shows how to create a Perl object which is linked to a C++ dynamic object. Constructors, destructors, and methods are demonstrated.

The CalcCC example demonstrates how to create a Perl interface to a C++ class which has superclasses. In an inheritance hierarchy the subclass is said to possess the features of its superclasses. All methods in the C++ class's hierarchy are listed in the XS code. These methods are treated as if they are part of the subclass (they are, in an OO sense). This example also demonstrates how to add a Perl-based method to a C++ class.

Set the $CC variable in the Makefile.PL to your C++ compiler and then run the Makefile.PL. Then type make perl to get a new perl binary in the build directory. The test suite will use this new perl binary.

You can build dynamic-loading extensions if you have a Perl which has perlmain.c compiled with a C++ compiler.

This requires 5.002.

Knowing your typemap.

It's a good idea to read the INPUT/OUTPUT maps for your type before using them. An INPUT map which uses SvRV() is dereferencing to get to the object. Some things, like perl_call_method(), as used in Ex5, don't want the object to be dereferenced and so you'll have to rebuild the reference with newRV().

An OUTPUT map which uses newRV() will cause the object to have its reference count incremented. Sometimes, as shown in the dropit() XSUB of Ex6, or the new() XSUB of Ex3, this effect of newRV() will have to be taken into account. OUTPUT maps will almost always return the object wrapped in newRV().

cpanm CookBookA::Ex6

perl -MCPAN -e shell
install CookBookA::Ex6