Ken Williams
and 2 contributors

NAME

Apache::SSI - Implement Server Side Includes in Perl

SYNOPSIS

In httpd.conf:

    <Files *.phtml>  # or whatever
    SetHandler perl-script
    PerlHandler Apache::SSI
    </Files>

You may wish to subclass Apache::SSI for your own extensions. If so, compile mod_perl with PERL_METHOD_HANDLERS=1 (so you can use object-oriented inheritance), and create a module like this:

    package MySSI;
    use Apache::SSI ();
    @ISA = qw(Apache::SSI);

    #embedded syntax:
    #<!--#something param=value -->
    sub ssi_something {
       my($self, $attr) = @_;
       my $cmd = $attr->{param};
       ...
       return $a_string;   
    }
 
 Then in httpd.conf:
 
    <Files *.phtml>
     SetHandler perl-script
     PerlHandler MySSI
    </Files>

DESCRIPTION

Apache::SSI implements the functionality of mod_include for handling server-parsed html documents. It runs under Apache's mod_perl.

In my mind, there are two main reasons you might want to use this module: you can sub-class it to implement your own custom SSI directives, and/or you can parse the output of other mod_perl handlers, or send the SSI output through another handler (use Apache::Filter to do this).

Each SSI directive is handled by an Apache::SSI method with the prefix "ssi_". For example, <!--#printenv--> is handled by the ssi_printenv method. attribute=value pairs inside the SSI tags are parsed and passed to the method in a hash reference.

'Echo' directives are handled by the ssi_echo method, which delegates lookup to methods with the prefix "echo_". For instance, <!--#echo var=DOCUMENT_NAME--> is handled by the echo_DOCUMENT_NAME method.

You can customize behavior by inheriting from Apache::SSI and overriding 'ssi_*' and 'echo_*' methods, or writing new ones.

SSI Directives

This module supports the same directives as mod_include. At least, that's the goal. =) For methods listed below but not documented, please see mod_include's online documentation at http://www.apache.org/ .

  • config

  • echo

  • exec

  • fsize

  • flastmod

  • include

  • printenv

  • set

  • perl

    There are two ways to call a Perl function, and two ways to supply it with arguments. The function can be specified either as an anonymous subroutine reference, or as the name of a function defined elsewhere:

     <!--#perl sub="sub { localtime() }"-->
     <!--#perl sub="time::now"-->

    If the 'sub' argument matches the regular expression /^\s*sub[^\w:]/, it is assumed to be a subroutine reference. Otherwise it's assumed to be the name of a function. In the latter case, the string "main::" will be prepended to the function name if the name doesn't contain "::" (this forces the function to be in the main package, or a package you specify). Note that it's a pretty bad idea to put your code in the main package, so I only halfheartedly endorse this feature.

    In general, it will be slower to use anonymous subroutines, because each one has to be eval()'ed and there is no caching. For best results, pre-load any code you need in the parent process, then call it by name.

    If you're calling a subroutine like "&Package::SubPack::handler", you can omit the "handler" portion, making your directive like this:

     <!--#perl sub="Package::Subpack"-->

    If you want to supply a list of arguments to the function, you use either the "arg" or the "args" parameter:

     <!--#perl sub="sub {$_[0] * 7}" arg=7-->
     <!--#perl sub=holy::matrimony arg=Hi arg=Lois-->
     <!--#perl sub=holy::matrimony args=Hi,Lois-->

    The "args" parameter will simply split on commas, meaning that currently there's no way to embed a comma in arguments passed via the "args" parameter. Use the "arg" parameter for this.

    If you give a key-value pair and the key is not 'sub', 'arg', 'args', or 'pass_request' (see below), then your routine will be passed both the key and the value. This lets you pass a hash of key-value pairs to your function:

     <!--#perl sub=holy::matrimony groom=Hi bride=Lois-->
     Will call &holy::matrimony('groom', 'Hi', 'bride', 'Lois');

    As of version 1.95, we pass the current Apache request object ($r) as the first argument to the function. To turn off this behavior, give the key-value pair 'pass_request=no', or put 'PerlSetVar SSIPerlPass_Request no' in your server's config file.

    See http://perl.apache.org/src/mod_perl.html for more information on Perl SSI calls.

  • if

  • elif

  • else

  • endif

    These four directives can be used just like in mod_include, with one important difference: the boolean expression is evaluated using Perl's eval(). This means you use == or eq instead of = to test equality. It also means you can use pre-loaded Perl subroutines in the conditions:

     <!--#if expr="&Movies::is_by_Coen_Brothers($MOVIE)"-->
      This movie is by the Coen Brothers.
     <!--#else-->
      This movie is not by the Coen Brothers.
     <!--#endif-->

    It can't handle very sophistocated Perl though, because it manually looks for variables (of the form $var or ${var}, just like mod_include), and will get tripped up on expressions like $object->method or $hash{'key'}. I'll welcome any suggestions for how to allow arbitrary Perl expressions while still filling in Apache variables.

CHAINING HANDLERS

There are two fairly simple ways for this module to exist in a stacked handler chain. The first uses Apache::Filter, and your httpd.conf would look something like this:

 PerlModule Apache::Filter
 PerlModule Apache::SSI
 PerlModule My::BeforeSSI
 PerlModule My::AfterSSI
 <Files ~ "\.ssi$">
  SetHandler perl-script
  PerlSetVar Filter On
  PerlHandler My::BeforeSSI Apache::SSI My::AfterSSI
 </Files>

The "PerlSetVar Filter On" directive tells the three stacked handlers that they should use their filtering mode. It's mandatory.

The second uses Apache::OutputChain, and your httpd.conf would look something like this:

 PerlModule Apache::OutputChain
 PerlModule Apache::SSIChain
 PerlModule My::BeforeSSI
 PerlModule My::AfterSSI
 <Files ~ "\.ssi$">
  SetHandler perl-script
  PerlHandler Apache::OutputChain My::AfterSSI Apache::SSIChain My::BeforeSSI
 </Files>

Note that the order of handlers is reversed in the two different methods. One reason I wrote Apache::Filter is to get the order to be more intuitive. Another reason is that Apache::SSI itself can be used in a handler stack using Apache::Filter, whereas it needs to be wrapped in Apache::SSIChain to be used with Apache::OutputChain.

Please see the documentation for Apache::OutputChain and Apache::Filter for more specific information. And look at the note in CAVEATS too.

CAVEATS

* When chaining handlers via Apache::Filter, if you use <!--#include ...--> or <!--#exec cgi=...-->, then Apache::SSI must be the last filter in the chain. This is because Apache::SSI uses $r->lookup_uri(...)->run to include the files, and this sends the output through C's stdout rather than Perl's STDOUT. Thus Apache::Filter can't catch it and filter it.

If Apache::SSI is the last filter in the chain, or if you stick to simpler SSI directives like <!--#fsize-->, <!--#flastmod-->, etc. you'll be fine.

* Currently, the way <!--#echo var=whatever--> looks for variables is to first try $r->subprocess_env, then try %ENV, then the five extra environment variables mod_include supplies. Is this the correct order?

TO DO

Revisit http://www.apache.org/docs/mod/mod_include.html and see what else there I can implement.

It would be nice to have a "PerlSetVar ASSI_Subrequests 0|1" option that would let you choose between executing a full-blown subrequest when including a file, or just opening it and printing it.

I'd like to know how to use Apache::test for the real.t test.

SEE ALSO

mod_include, mod_perl(3), Apache(3), HTML::Embperl(3), Apache::ePerl(3), Apache::OutputChain(3)

AUTHOR

Ken Williams ken@mathforum.org

Concept based on original version by Doug MacEachern dougm@osf.org . Implementation different.

COPYRIGHT

Copyright 1998 Swarthmore College. All rights reserved.

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