Torsten Förtsch > Class-Member-1.6 > Class::Member



Annotate this POD

View/Report Bugs
Module Version: 1.6   Source  


Class::Member - A set of modules to make the module developement easier


 package MyModule;
 use Class::Member::HASH qw/member_A member_B -CLASS_MEMBERS
                            -NEW=new -INIT=init/;
 package MyModule;
 use Class::Member::GLOB qw/member_A member_B -CLASS_MEMBERS
                            -NEW=new -INIT=init/;
 package MyModule;
 use Class::Member qw/member_A member_B -CLASS_MEMBERS/;
 package MyModule;
 use Class::Member::Dynamic qw/member_A member_B -CLASS_MEMBERS/;


Perl class instances are mostly blessed HASHes or GLOBs and store member variables either as $self->{membername} or ${*$self}{membername} respectively.

This is very error prone when you start to develope derived classes based on such modules. The developer of the derived class must watch the member variables of the base class to avoid name conflicts.

To avoid that Class::Member::XXX stores member variables in its own namespace prepending the package name to the variable name, e.g.

 package My::New::Module;

 use Class::Member::HASH qw/member_A memberB/;

will store member_A as $self->{'My::New::Module::member_A'}.

To make access to these members easier it exports access functions into the callers namespace. To access member_A you simply call.

 $self->member_A;               # read access
 $self->member_A($new_value);   # write access
 $self->member_A=$new_value;    # write access (used as lvalue)

Class::Member::HASH and Class::Member::GLOB are used if your objects are HASH or GLOB references. But sometimes you do not know whether your instances are GLOBs or HASHes (Consider developement of derived classes where the base class is likely to be changed.). In this case use Class::Member and the methods are defined at compile time to handle each type of objects, GLOBs and HASHes. But the first access to a method redefines it according to the actual object type. Thus, the first access will last slightly longer but all subsequent calls are executed at the same speed as Class::Member::GLOB or Class::Member::HASH.

Class::Member::Dynamic is used if your objects can be GLOBs and HASHes at the same time. The actual type is determined at each access and the appropriate action is taken.

In addition to member names there are a few options that can be given: -CLASS_MEMBERS. It lets the import() function create an array named @CLASS_MEMBERS in the caller's namespace that contains the names of all methods it defines. Thus, you can create a contructor that expects named parameters where each name corresponds to a class member:

 use Class::Member qw/member_A member_B -CLASS_MEMBERS/;
 sub new {
   my $parent=shift;
   my $class=ref($parent) || $parent;
   my $I=bless {}=>$class;
   my %o=@_;
   if( ref($parent) ) {         # inherit first
     foreach my $m (@CLASS_MEMBERS) {
   # then override with named parameters
   foreach my $m (@CLASS_MEMBERS) {
     $I->$m=$o{$m} if( exists $o{$m} );
   return $I;

Further, if you use one of Class::Member::HASH or Class::Member::GLOB a constructor method can be created automatically. Just add -NEW or -NEW=name to the use() call. The first form creates a new() method that is implemented as shown except of the $I->init call. The 2nd form can be used if your constructor must not be named new.

What happens if one Class::Member based class inherits the constructor from another Class::Member based class? In this case the inherited contructor works for the @CLASS_MEMBERS of the base class as well as the derived class. For example:

 package Base;
 use Class::Member::HASH qw/-NEW -CLASS_MEMBERS el1 el2/;

 package Inherited;
 use Class::Member::HASH qw/-CLASS_MEMBERS el3/;
 use base qw/Base/;

Now Inherited-new> calls the constructor of the base class but one can pass el1, el2 as well as el3 parameters.

The $I->init call is added by specifying the -INIT or -INIT=name option. If given a new function &{I N I T} is created in the caller's namespace to hold the name of the init() method. Yes, the symbol name does contain spaces to make it harder to change by chance. You don't normally have to care about it. Again, the -INIT=name form is used if your init() method is not named init.

The init() method itself is provided by you.

More detailed here is how the initializer is called:

 my $init=$self->can('I N I T');
 if( $init ) {

That means the constructor looks if the class itself or one of the base classes provides a I N I T method (the name includes spaces between each pair of characters). If so it calls that method to fetch the initializer name. The last step calls the initializer itself.


Torsten Foertsch <>


Class::Member::HASH, Class::Member::GLOB, Class::Member::Dynamic


Copyright 2003-2008 Torsten Foertsch.

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

syntax highlighting: