Petal::CodePerl - Make Petal go faster by compiling the expressions


  use Petal::CodePerl;

  # continue as you would normally using Petal


  use Petal;
  $Petal::CodeGenerator = 'Petal::CodePerl::CodeGenerator';
  # continue as you would normally use Petal


This module provides a CodeGenerator for Petal that inherits almost everything from Petal::CodeGenerator but modifies how expressions are dealt with. Petal normally includes code like this

  $hash->get( "not:user" )

in the compiled template. This means the path has to be parsed and interpreted at runtime. Using Petal::CodePerl, Petal will now produce this

  ! ($hash->{"user"})

which will be much faster.

It uses Parse::RecDescent to parse the PETALES expressions which makes it a bit slow to load the module but this won't matter much unless you have turned off caching. It won't matter at all for something like Apache's mod_perl.


You have two choices, you can replace use Petal with use Petal::CodePerl in all your scripts or you can do $Petal::CodeGenerator = 'Petal::CodePerl::CodeGenerator'. Either of these will cause Petal to use the expression compiling version of the CodeGenerator.


Using Parse::RecDescent makes it easier to expand the PETALES grammar. I have made the following enhancements.

  • alternators work as in TAL, so you can do

      petal:content="a/name|b/name|string:no name"
  • you can explicitly ask for hash, array or method in a path

    • user{name} is $hash->{"user"}->{"name"}

    • user[1] is $hash->{"user"}->[1]

    • user/method() is $hash->{"user"}->method()

      using these will make your template even faster although you need to be certain of your data types.

  • method arguments can be any expression for example

      user/purchase cookie{basket}

    will give

  • you can do more complex defines, like the following

      petal:define="a{b}[1] string:hello"

    which will give

      $hash->{"a"}->[1] = "hello"
  • some other stuff that I can't remember just now.


Modifiers can now be compiled, partially compiled or work exactly as they did before. When compiling the expression, Petal::CodePerl will look at the modifier's package to figure what it supports. The order of preference is fully compiled, partially compiled, original style. So you can slowly migrate you modifiers to full compilation.

Note that although original style still works, you cannot use any of the extra features of Petal::CodePerl in the path if you are using an original modifier. So you cannot do mymod:hash{key} until you convert your modifier into one of the newer styles. This is because your original style modifier uses Petal's parser and Petal's parser doesn't accept the {} syntax. It would probably be possible to fix this too but I don't think it's worth it. If you really need it, let me know.

Partially Compiled

Partially compiled modifiers are easiest, in fact they are even easier than Petal's original style. To partially compile a modifier, define a package with a process_value() method, and put that into %Petal::Hash::MODIFIERS just as you would with a normal modifier package.

A simple example

  package Petal::Hash::Length;
  $Petal::Hash::MODIFIERS{"length:"} = "Petal::Hash::Length";

  sub process_value
    my $class = shift;
    my $hash = shift;
    my $value = shift;

   return length($value);

This is a little different to the process() method originally used for modifiers. For a path like true:this/is/a/path, Petal would call process() with 2 arguments - the Petal hash and the string "this/is/a/path", it was then up to the method to parse that string and find the value that it pointed to. For process_value(), this string has already been parsed and compiled, so the value defined by the path is passed in and can be used straight away.

Fully Compiled

The most efficient but a little more complicated is fully compiled. For a fully compiled modifier you need to have an inline() method which will return a Code::Perl object that will produce the compiled code. It's not a hard as it sounds. There is an easy to use Code::Perl object provided that makes this fairly straight-forward. Here's the example above rewritten as a fully compiled modifier.

  package Petal::Hash::Length;
  $Petal::Hash::MODIFIERS{"length:"} = "Petal::Hash::Length";

  use Petal::CodePerl::Expr qw( perlsprintf );

  sub inline
    my $class = shift;
    my $hash_obj = shift;
    my $value_obj = shift;

    return perlsprintf("length(%s)", $value_obj);

The key points to note are that instead of getting the Petal hash and the value of the expression you are modifying, you get Code::Perl objects representing them, you then use these objects to construct another Code::Perl object which will eventually spit out the Perl code for your modifier. It's not in the example above but if your modifier wants to access the Petal hash please don't write '$hash->{blah}', instead write '%s->{blah}' and pass in the $hash_obj object. This allows for the possibility in the future that the hash is not called '$hash'.

Why not just use simple strings? Because using Code::Perl objects instead of strings of Perl everywhere means you don't have to worry about escaping and wrapping things in ()s and it means that modifiers can be inside modifiers, deep inside complicated Petales expressions, which themselves are inside a modifier etc, etc.

Tips for writing modifiers

The easiest way to write a compiled modifier is to write a partially compiled one first and when you have that working perfectly, you can turn it into a fully compiled one. If at any stage you need to go back to using the partially compiled one, rather than commenting out the inline() method you can set $Petal::CodePerl::InlineMod = 0 and no inline() methods routines will be used.

If you are going to use the $value_obj more than once in a compiled modifier then you need to be a little bit careful as you only want to calculate the value once. Say you have a modifier that takes a string and doubles it, you could do this

  perlsprintf("%s.%s", $value_obj, $value_obj);

but then double:thing/method() would be compiled to


which is bad news because it's inefficient to call the method twice and there's no guarantee that the method will actually return the same value for each call. So what you should really do is

  perlsprintf("do{my $v = %s; $v.$v}", $value_obj);


  do{my $v = ($hash->{thing}->method()); $v.$v}

which is efficient and safe.


Any modifiers that are defined in the original Petal style, using a process() method will not be compiled and so will still work as before.


Petal::CodePerl is in development. There are no known bugs and Petal passes it's full test suite using this code generator. However there are probably some differences between it's grammar and Petal's current grammar. Please let me know if you find anything that works differently with Petal::CodePerl.


Your templates may now generate "undefined value" warnings if you try to use an undefined value. Previously, Petal prevented many of these from occurring. As always, the best thing to do is not to avoid using undefined values in your templates. Hopefully this will be fixed shortly.


Written by Fergal Daly <>.


Copyright 2003 by Fergal Daly <>.

This program is free software and comes with no warranty. It is distributed under the LGPL license

See the file LGPL included in this distribution or