The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


Template::Extract - Use TT2 syntax to extract data from documents


This document describes version 0.40 of Template::Extract, released September 17, 2005.


    use Template::Extract;
    use Data::Dumper;

    my $obj = Template::Extract->new;
    my $template = << '.';
    <ul>[% FOREACH record %]
    <li><A HREF="[% url %]">[% title %]</A>: [% rate %] - [% comment %].
    [% ... %]
    [% END %]</ul>

    my $document = << '.';
    <html><head><title>Great links</title></head><body>
    <ul><li><A HREF="">News for nerds.</A>: A+ - nice.
    this text is ignored.</li>
    <li><A HREF="">Where do you want...</A>: Z! - yeah.
    this text is ignored, too.</li></ul>

    print Data::Dumper::Dumper(
        $obj->extract($template, $document)


This module adds template extraction functionality to the Template toolkit. It can take a rendered document and its template together, and get the original data structure back, effectively reversing the Template::process function.



Constructor. Currently all options are passed into the underlying Template::Parser object. The same set of options are also passed to classes responsible to compile and run the extraction process, but they are currently ignored.

extract($template, $document, \%values)

This method takes three arguments: the template string, or a reference to it; a document string to match against; and an optional hash reference to supply initial values, as well as storing the extracted values into.

The return value is \%values upon success, and undef on failure. If \%values is omitted from the argument list, a new hash reference will be constructed and returned.

Extraction is done by transforming the result from Template::Parser to a highly esoteric regular expression, which utilizes the (?{...}) construct to insert matched parameters into the hash reference.

The special [% ... %] directive is taken as the /.*?/s regex, i.e. ignore everything (as short as possible) between this identifier and the next one. For backward compatibility, [% _ %] and [% __ %] are also accepted.

The special [% // %] directive is taken as a non-capturing regex, embedded inside /(?:)/s; for example, [% /\d*/ %] matches any number of digits. Capturing parentheses may not be used with this directive, but you can use the [% var =~ // %] directive to capture the match into var.

You may set $Template::Extract::DEBUG to a true value to display generated regular expressions.

The extraction process defaults to succeed even with a partial match. To match the entire document only, set $Template::Extract::EXACT to a true value.


Use Template::Extract::Compile to perform the first phase of extract, by returning the regular expression compiled from $template.

run($regex, $document, \%values)

Use Template::Extract::Run to perform the second phase of extract, by applying the regular expression on $document and returning the resulting \%values.


If you would like to use different modules to parse, compile and run the extraction process, simply subclass Template::Extract and override the COMPILE_CLASS, PARSER_CLASS and RUN_CLASS methods to return alternate class names.


Currently, the extract method only supports [% GET %], [% SET %] and [% FOREACH %] directives, because [% WHILE %], [% CALL %] and [% SWITCH %] blocks are next to impossible to extract correctly.

[% SET key = "value" %] only works for simple scalar values.

Outermost [% FOREACH %] blocks must match at least once in the document, but inner ones may occur zero times. This is to prevent the regex optimizer from failing prematurely.

There is no support for different PRE_CHOMP and POST_CHOMP settings internally, so extraction could fail silently on extra linebreaks.

It is somewhat awkward to use global variables to control EXACT and DEBUG behaviour; patches welcome to promote them into per-instance options.


This module's companion class, Template::Generate, is still in early experimental stages; it can take data structures and rendered documents, then automagically generates templates to do the transformation. If you are into related research, please mail any ideas to me.


Template::Extract::Compile, Template::Extract::Run, Template::Extract::Parser

Template, Template::Generate

Simon Cozens's introduction to this module, in O'Reilly's Spidering Hacks:

Mark Fowler's introduction to this module, in The 2003 Perl Advent Calendar:


Autrijus Tang <>


Copyright 2001, 2002, 2003, 2004, 2005 by Autrijus Tang <>.

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