The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

NAME

Apache::Handlers

SYNOPSIS

In code:

  use Apache::Handlers qw(CLEANUP);

  our $global;
  my $session : PerlCleanupHandler;

  CLEANUP {
    our $global = undef;
  };

In httpd.conf:

  PerlModule Apache::Handlers
  PerlChildInitHandler Apache::Handlers
  PerlPostReadRequestHandler Apache::Handlers
  <Perl>
    Apache::Handlers -> reset;
  </Perl>

DESCRIPTION

Apache::Handlers provides two different methods of declaring when code snippets should be run during the Apache request phase.

The code defined with the constructs provided by this module do not directly affect the success or failure of the request. Thus, this module does not provide a replacement for content, access, or other handlers.

The code is executed in the order it is encountered except for CHILDEXIT, CLEANUP, PerlChildExitHandler, and PerlCleanupHandler code. These are executed in the reverse order, similar to the pairing of BEGIN and END blocks.

The block construct or attribute must be run before the phase it refers to. Otherwise, it won't be run in that phase. The phases are run in the following order:

CHILDINIT TRANS HEADERPARSER ACCESS AUTHEN AUTHZ TYPE FIXUP CONTENT LOG CLEANUP CHILDEXIT

The RESTART phase is not an actual Apache request phase and has no effect after the server has started. It is used to define code that should run during the server startup phase when Apache reads the server configuration the second time or is gracefully (or not so gracefully) restarted. It should be used to clean up so the second configuration process won't duplicate information or cause errors.

If this module is called during the ChildInit phase, then it will only call that code associated with CHILDINIT blocks. Otherwise, the CHILDINIT code will be run at the first opportunity (basically, the first request made of the child process). Thus the two Perl*Handler configuration directives in the Synopsis.

Running without mod_perl

When developing outside mod_perl, all code associated with CHILDINIT, TRANS, HEADERPARSER, ACCESS, AUTHEN, AUTHZ, TYPE, FIXUP, and CONTENT is run in an INIT block. All code associated with LOG, CLEANUP, and CHILDEXIT is run in an END block.

Block Constructs

The following allow for blocks of code to be run at the specified phase. Note that these are subroutines taking a single code reference argument and thus require a terminating semi-colon (;). They are named to be like the BEGIN, END, etc., constructs in Perl, though they are not quite at the same level in the language.

If the code is seen and handled before Apache has handled a request, it will be run for each request. Otherwise, it is pushed on the handler stack, run, and then removed at the end of the request.

These are named the same as the Apache/mod_perl configuration directives except the Perl and Handler strings have been removed and the remainder has been capitalized.

ACCESS
AUTHEN
AUTHZ
CHILDEXIT
CHILDINIT
CLEANUP
CONTENT
FIXUP
HEADERPARSER
LOG
POSTREADREQUEST
RESTART
TRANS
TYPE

Attributes

If Attribute::Handlers is available, then the following attributes are available (N.B.: Attribute::Handlers requires Perl 5.6.0). These are named the same as the Apache/mod_perl configuration directives.

If the attribute argument is a constant value (non-CODE reference), then the variable is assigned that value. Otherwise, it is assigned the value that the CODE reference returns.

If the attribute is being applied to a subroutine, then that subroutine is called during that phase. For example, the following two snippets result in the same code being run at the same time.

 my $something  = sub : PerlChildExitHandler {
   print "We did it!\n";
 };

 sub something : PerlChildExitHandler {
   print "We did it!\n";
 };

When an attribute is applied to a subroutine, the argument is ignored.

When the attribute argument is itself a CODE reference, the referent (the variable the attribute applies to) is passed as a reference:

 my $global : PerlChildInitHandler(sub { print "global: $$_[0]\n" });

This will print the value of $global and set it equal to 1 (or the value of the print statement).

PerlAccessHandler
PerlAuthenHandler
PerlAuthzHandler
PerlChildInitHandler
PerlChildExitHandler
PerlCleanupHandler
PerlFixupHandler
PerlHandler
PerlHeaderParserHandler
PerlLogHandler
PerlPostReadRequestHandler
PerlRestartHandler
PerlTransHandler
PerlTypeHandler

Other Methods

dump

This will dump the current set of code references and return the string. This uses Data::Dumper.

reset

This will clear out all previously set code. This should only be used in the startup.pl or equivalent so that code doesn't get run twice during a request (when it should only be run once). This will also run any RESET blocks that have been defined.

run_phase

Given a list of phases (using the names for the block constructs above), this will run through the code for that phase, dieing (outside mod_perl) or logging (if in mod_perl) if there is an error. For example,

  run_phase( qw: CONTENT LOG CLEANUP : );

will run any code associated with the CONTENT, LOG, and CLEANUP phases.

CAVEATS

Caveats are things that at first glance might be bugs, but end up potentially useful. So I am going to make this section into a kind of cookbook for non-obvious uses for these potential bugs.

Authentication and Authorization

Be aware that these two phases only run if Apache has reason to believe they are needed. This can be a bit handy since the following snippet should tell you if the authentication phase was run. Of course, if an authentication handler runs before this and returns OK, then this may not run.

  my $authentication_ran : PerlTransHandler(0) PerlAuthenHandler(1);

  LOG {
    if($authentication_ran) {
      # log something special
    }
  };

Errors

If code causes an error (such that an eval would set $@), then the request will throw a SERVER_ERROR and write $@ to either STDERR (if not in mod_perl and there is no die handler, such as the Error module) or to the Apache error log with a log level of debug.

Useing modules

Any of the block constructs or attributes provided by this module that are used in the body of a module that is brought in via the use keyword will be considered to take place before the child is spawned. This means that any code designated to run during a particular phase will be run at the appropriate time as if the module had been loaded during the server startup.

Modules can now rest assured that using a CLEANUP block in their file will mean that code is run at the end of every request, even if the module was loaded in the child process and not during server startup.

This is done by looking for code run during the BEGIN phase.

BUGS

Unlike caveats, bugs are features that are undesirable and/or get in the way of doing something useful. I'm sure there are some. Please let me know when you find them.

Security

There is no way (currently) to limit registration of code for later processing during a particular phase. Ideas are welcome for how this should be designed.

SEE ALSO

Apache, Attribute::Handlers, Data::Dumper.

AUTHOR

James G. Smith <jsmith@cpan.org>

COPYRIGHT

Copyright (C) 2002 Texas A&M University. All Rights Reserved.

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

3 POD Errors

The following errors were encountered while parsing the POD:

Around line 362:

=back doesn't take any parameters, but you said =back 4

Around line 429:

=back doesn't take any parameters, but you said =back 4

Around line 459:

=back doesn't take any parameters, but you said =back 4