Compiler.pm

# Copyright (c) 1998-2005 by Jonathan Swartz. All rights reserved.
# This program is free software; you can redistribute it and/or modify it
# under the same terms as Perl itself.

package HTML::Mason::Compiler;

use strict;
use warnings;
use Data::Dumper;
use HTML::Mason::Component::FileBased;
use HTML::Mason::Component::Subcomponent;
use HTML::Mason::Exceptions( abbr => [qw(param_error compiler_error syntax_error)] );
use HTML::Mason::Lexer;
use HTML::Mason::Tools qw(checksum);
use Params::Validate qw(:all);
Params::Validate::validation_options( on_fail => sub { param_error join '', @_ } );

use Class::Container;
use base qw(Class::Container);

BEGIN
{
    __PACKAGE__->valid_params
        (
         allow_globals =>
         { parse => 'list', type => ARRAYREF, default => [],
           descr => "An array of names of Perl variables that are allowed globally within components" },

         default_escape_flags =>
         { parse => 'string', type => SCALAR|ARRAYREF, default => [],
           descr => "Escape flags that will apply by default to all Mason tag output" },

         enable_autoflush =>
         { parse => 'boolean', type => SCALAR, default => 1,
           descr => "Whether to include support for autoflush when compiling components" },

         lexer =>
         { isa => 'HTML::Mason::Lexer',
           descr => "A Lexer object that will scan component text during compilation" },

         preprocess =>
         { parse => 'code', type => CODEREF, optional => 1,
           descr => "A subroutine through which all component text will be sent during compilation" },

         postprocess_perl =>
         { parse => 'code', type => CODEREF, optional => 1,
           descr => "A subroutine through which all Perl code will be sent during compilation" },

         postprocess_text =>
         { parse => 'code', type => CODEREF, optional => 1,
           descr => "A subroutine through which all plain text will be sent during compilation" },

         use_source_line_numbers =>
         { parse => 'boolean', type => SCALAR, default => 1,
           descr => "Whether to use source line numbers in errors and debugger" },
        );

    __PACKAGE__->contained_objects
        ( lexer => { class => 'HTML::Mason::Lexer',
                     descr => "This class generates compiler events based on the components source" },
        );

    # Define an IN_PERL_DB compile-time constant indicating whether we are
    # in the Perl debugger. This is used in the object file to
    # determine whether to call $m->debug_hook.
    #
    if (defined($DB::sub)) {
        *IN_PERL_DB = sub () { 1 };
    } else {
        *IN_PERL_DB = sub () { 0 };
    }
}

use HTML::Mason::MethodMaker
    ( read_only => [qw(
                       enable_autoflush
                       lexer
                       object_id
                       preprocess
                       postprocess_perl
                       postprocess_text
                       use_source_line_numbers
                       )
                    ],
      );

my $old_escape_re = qr/^[hnu]+$/;

sub new
{
    my $class = shift;
    my $self = $class->SUPER::new(@_);

    $self->default_escape_flags( $self->{default_escape_flags} )
        if defined $self->{default_escape_flags};

    # Verify the validity of the global names
    $self->allow_globals( @{$self->{allow_globals}} );

    # Compute object_id once, on the assumption that all of compiler's
    # and lexer's parameters are read-only.
    $self->compute_object_id;
    
    return $self;
}

sub compute_object_id
{
    my $self = shift;

    # Can't use object keys because they stringify differently every
    # time the program is loaded, whether they are a reference to the
    # same object or not.
    my $spec = $self->validation_spec;
    my @id_keys =
        ( grep { ! exists $spec->{$_}{isa} && ! exists $spec->{$_}{can} }
          grep { $_ ne 'container' } keys %$spec );

    my @vals = ('HTML::Mason::VERSION', $HTML::Mason::VERSION);
    foreach my $k ( @id_keys ) {
        push @vals, $k, $self->{$k};
    }
    my $dumped_vals = Data::Dumper->new(\@vals)->Indent(0)->Dump;
    $self->{object_id} = checksum($dumped_vals);
}

my %top_level_only_block = map { $_ => 1 } qw( cleanup once shared );
my %valid_comp_flag = map { $_ => 1 } qw( inherit );

sub add_allowed_globals
{
    my $self = shift;
    my @globals = @_;

    if ( my @bad = grep { ! /^[\$@%]/ } @globals )
    {
        param_error "add_allowed_globals: bad parameters '@bad', must begin with one of \$, \@, %\n";
    }

    $self->{allow_globals} = [ keys %{ { map { $_ => 1 } @globals, @{ $self->{allow_globals} } } } ];
    return @{ $self->{allow_globals} };
}

sub allow_globals
{
    my $self = shift;

    if (@_)
    {
        $self->{allow_globals} = [];
        return if @_ == 1 and not defined $_[0]; # @_ is (undef)
        $self->add_allowed_globals(@_);
    }

    return @{ $self->{allow_globals} };
}

sub default_escape_flags
{
    my $self = shift;

    return $self->{default_escape_flags} unless @_;

    my $flags = shift;

    unless ( defined $flags )
    {
        $self->{default_escape_flags} = [];
        return;
    }

    # make sure this is always an arrayref
    unless ( ref $flags )
    {
        if ( $flags =~ /^[hu]+$/ )
        {
            $self->{default_escape_flags} = [ split //, $flags ];
        }
        else
        {
            $self->{default_escape_flags} = [ $flags ];
        }
    }

    return $self->{default_escape_flags};
}

sub compile
{
    my $self = shift;
    my %p = validate( @_, { comp_source => { type => SCALAR|SCALARREF },
                            name => { type => SCALAR },
                            comp_path => { type => SCALAR },
                            fh => { type => HANDLE, optional => 1 },
                          } );
    my $src = ref($p{comp_source}) ? $p{comp_source} : \$p{comp_source};

    # The current compile - initially the main component, then each subcomponent/method
    local $self->{current_compile} = {};
    
    # Useful for implementing features that affect both main body and methods/subcomps
    local $self->{main_compile} = $self->{current_compile};

    # So we're re-entrant in subcomps
    local $self->{paused_compiles} = [];

    local $self->{comp_path} = $p{comp_path};

    # Preprocess the source.  The preprocessor routine is handed a
    # reference to the entire source.
    if ($self->preprocess)
    {
        eval { $self->preprocess->( $src ) };
        compiler_error "Error during custom preprocess step: $@" if $@;
    }

    $self->lexer->lex( comp_source => $src, name => $p{name}, compiler => $self );

    return $self->compiled_component( exists($p{fh}) ? (fh => $p{fh}) : () );
}

sub start_component
{
    my $self = shift;
    my $c = $self->{current_compile};

    $c->{in_main} = 1;

    $c->{in_block} = undef;

    $self->_init_comp_data($c);
}

sub _init_comp_data
{
    my $self = shift;
    my $data = shift;

    $data->{body} = '';
    $data->{last_body_code_type} = '';

    foreach ( qw( def method ) )
    {
        $data->{$_} = {};
    }

    $data->{args} = [];
    $data->{flags} = {};
    $data->{attr} = {};

    $data->{comp_with_content_stack} = [];

    foreach ( qw( cleanup filter init once shared ) )
    {
        $data->{blocks}{$_} = [];
    }
}

sub end_component
{
    my $self = shift;
    my $c = $self->{current_compile};

    $self->lexer->throw_syntax_error("Not enough component-with-content ending tags found")
        if @{ $c->{comp_with_content_stack} };
}

sub start_block
{
    my $self = shift;
    my $c = $self->{current_compile};
    my %p = @_;

    $self->lexer->throw_syntax_error("Cannot define a $p{block_type} section inside a method or subcomponent")
         if $top_level_only_block{ $p{block_type} } && ! $c->{in_main};

    $self->lexer->throw_syntax_error("Cannot nest a $p{block_type} inside a $c->{in_block} block")
         if $c->{in_block};

    $c->{in_block} = $p{block_type};
}

sub raw_block
{
    # These blocks contain Perl code - so don't include <%text> and so on.

    my $self = shift;
    my $c = $self->{current_compile};
    my %p = @_;

    eval { $self->postprocess_perl->( \$p{block} ) if $self->postprocess_perl };
    compiler_error $@ if $@;

    my $method = "$p{block_type}_block";
    return $self->$method(%p) if $self->can($method);

    my $comment = '';
    if ( $self->lexer->line_number && $self->use_source_line_numbers )
    {
        my $line = $self->lexer->line_number;
        my $file = $self->_escape_filename( $self->lexer->name );
        $comment = qq{#line $line "$file"\n};
    }

    push @{ $self->{current_compile}{blocks}{ $p{block_type} } }, "$comment$p{block}";
}

sub doc_block
{
    # Don't do anything - just discard the comment.
}

sub perl_block
{
    my $self = shift;
    my %p = @_;

    $self->_add_body_code( $p{block} );

    $self->{current_compile}{last_body_code_type} = 'perl_block';
}

sub text
{
    my ($self, %p) = @_;
    my $tref = ref($p{text}) ? $p{text} : \$p{text};  # Allow a reference

    eval { $self->postprocess_text->($tref) } if $self->postprocess_text;
    compiler_error $@ if $@;

    $$tref =~ s,([\'\\]),\\$1,g;

    if ($self->enable_autoflush) {
        $self->_add_body_code("\$m->print( '", $$tref, "' );\n");
    } else {
        $self->_add_body_code("\$\$_outbuf .= '", $$tref, "';\n");
    }

    $self->{current_compile}{last_body_code_type} = 'text';
}

sub text_block
{
    my $self = shift;
    my %p = @_;
    $self->text(text => \$p{block});
}

sub end_block
{
    my $self = shift;
    my $c = $self->{current_compile};
    my %p = @_;

    $self->lexer->throw_syntax_error("End of $p{block_type} encountered while in $c->{in_block} block")
        unless $c->{in_block} eq $p{block_type};

    $c->{in_block} = undef;
}

sub variable_declaration
{
    my $self = shift;
    my %p = @_;

    $self->lexer->throw_syntax_error("variable_declaration called inside a $p{block_type} block")
        unless $p{block_type} eq 'args';

    my $arg = "$p{type}$p{name}";

    $self->lexer->throw_syntax_error("$arg already defined")
        if grep { "$_->{type}$_->{name}" eq $arg } @{ $self->{current_compile}{args} };

    push @{ $self->{current_compile}{args} }, { type => $p{type},
                                             name => $p{name},
                                             default => $p{default},
                                             line => $self->lexer->line_number,
                                             file => $self->lexer->name,
                                           };
}

sub key_value_pair
{
    my $self = shift;
    my %p = @_;

    compiler_error "key_value_pair called inside a $p{block_type} block"
        unless $p{block_type} eq 'flags' || $p{block_type} eq 'attr';

    my $type = $p{block_type} eq 'flags' ? 'flag' : 'attribute';
    $self->lexer->throw_syntax_error("$p{key} $type already defined")
        if exists $self->{current_compile}{ $p{block_type} }{ $p{key} };

    $self->{current_compile}{ $p{block_type} }{ $p{key} } = $p{value}
}

sub start_named_block
{
    my $self = shift;
    my $c = $self->{current_compile};
    my %p = @_;

    # Error if defining one def or method inside another
    $self->lexer->throw_syntax_error
        ("Cannot define a $p{block_type} block inside a method or subcomponent")
            unless $c->{in_main};

    # Error for invalid character in name
    $self->lexer->throw_syntax_error("Invalid $p{block_type} name: $p{name}")
        if $p{name} =~ /[^.\w-]/;

    # Error if two defs or two methods defined with same name
    $self->lexer->throw_syntax_error
        (sprintf("Duplicate definition of %s '%s'",
                 $p{block_type} eq 'def' ? 'subcomponent' : 'method', $p{name}))
            if exists $c->{$p{block_type}}{ $p{name} };
    
    # Error if def and method defined with same name
    my $other_type = $p{block_type} eq 'def' ? 'method' : 'def';
    $self->lexer->throw_syntax_error
        ("Cannot define a method and subcomponent with the same name ($p{name})")
            if exists $c->{$other_type}{ $p{name} };

    $c->{in_main}--;

    $c->{ $p{block_type} }{ $p{name} } = {};
    $self->_init_comp_data( $c->{ $p{block_type} }{ $p{name} } );
    push @{$self->{paused_compiles}}, $c;
    $self->{current_compile} = $c->{ $p{block_type} }{ $p{name} };
    $self->{current_compile}->{in_named_block} = {block_type => $p{block_type}, name => $p{name}};
}

sub end_named_block
{
    my $self = shift;

    delete $self->{current_compile}->{in_named_block};
    $self->{current_compile} = pop @{$self->{paused_compiles}};
    $self->{current_compile}{in_main}++;
}

sub substitution
{
    my $self = shift;
    my %p = @_;

    my $text = $p{substitution};

    # This is a comment tag if all lines of text contain only whitespace
    # or start with whitespace and a comment marker, e.g.
    #
    #   <%
    #     #
    #     # foo
    #   %>
    #
    my @lines = split(/\n/, $text);
    unless (grep { /^\s*[^\s\#]/ } @lines) {
        $self->{current_compile}{last_body_code_type} = 'substitution';
        return;
    }

    if ( ( exists $p{escape} && defined $p{escape} ) ||
         @{ $self->{default_escape_flags} }
       )
    {
        my @flags;
        if ( defined $p{escape} )
        {
            $p{escape} =~ s/\s+$//;

            if ( $p{escape} =~ /$old_escape_re/ )
            {
                @flags = split //, $p{escape};
            }
            else
            {
                @flags = split /\s*,\s*/, $p{escape};
            }
        }

        # is there any way to check the flags for validity and still
        # allow them to be dynamically set from components?

        unshift @flags, @{ $self->default_escape_flags }
            unless grep { $_ eq 'n' } @flags;

        my %seen;
        my $flags =
            ( join ', ',
              map { $seen{$_}++ ? () : "'$_'" }
              grep { $_ ne 'n' } @flags
            );

        $text = "\$m->interp->apply_escapes( (join '', ($text)), $flags )" if $flags;
    }

    my $code;

    # Make sure to allow lists within <% %> tags.
    #
    if ($self->enable_autoflush) {
        $code = "\$m->print( $text );\n";
    } else {
        # more efficient output form when autoflush is disabled. only
        # output defined bits, which is what $m->print does internally
        # as well. use 'if defined' for maximum efficiency; grep
        # creates a list.
        $code = "for ( $text ) { \$\$_outbuf .= \$_ if defined }\n";
    }

    eval { $self->postprocess_perl->(\$code) } if $self->postprocess_perl;
    compiler_error $@ if $@;

    $self->_add_body_code($code);

    $self->{current_compile}{last_body_code_type} = 'substitution';
}

sub component_call
{
    my $self = shift;
    my %p = @_;

    my ($prespace, $call, $postspace) = ($p{call} =~ /(\s*)(.*)(\s*)/s);
    if ( $call =~ m,^[\w/.],)
    {
        my $comma = index($call, ',');
        $comma = length $call if $comma == -1;
        (my $comp = substr($call, 0, $comma)) =~ s/\s+$//;
        $call = "'$comp'" . substr($call, $comma);
    }
    my $code = "\$m->comp( $prespace $call $postspace \n); ";
    eval { $self->postprocess_perl->(\$code) } if $self->postprocess_perl;
    compiler_error $@ if $@;

    $self->_add_body_code($code);

    $self->{current_compile}{last_body_code_type} = 'component_call';
}

sub component_content_call
{
    my $self = shift;
    my $c = $self->{current_compile};
    my %p = @_;

    my $call = $p{call};
    for ($call) { s/^\s+//; s/\s+$//; }
    push @{ $c->{comp_with_content_stack} }, $call;

    my $code = "\$m->comp( { content => sub {\n";
    $code .= $self->_set_buffer();

    eval { $self->postprocess_perl->(\$code) } if $self->postprocess_perl;
    compiler_error $@ if $@;

    $self->_add_body_code($code);

    $c->{last_body_code_type} = 'component_content_call';
}

sub component_content_call_end
{
    my $self = shift;
    my $c = $self->{current_compile};
    my %p = @_;

    $self->lexer->throw_syntax_error("Found component with content ending tag but no beginning tag")
        unless @{ $c->{comp_with_content_stack} };

    my $call = pop @{ $c->{comp_with_content_stack} };
    my $call_end = $p{call_end};
    for ($call_end) { s/^\s+//; s/\s+$//; }

    my $comp = undef;
    if ( $call =~ m,^[\w/.],)
    {
        my $comma = index($call, ',');
        $comma = length $call if $comma == -1;
        ($comp = substr($call, 0, $comma)) =~ s/\s+$//;
        $call = "'$comp'" . substr($call, $comma);
    }
    if ($call_end) {
        if ($call_end !~ m,^[\w/.],) {
            $self->lexer->throw_syntax_error("Cannot use an expression inside component with content ending tag; use a bare component name or </&> instead");
        }
        if (!defined($comp)) {
            $self->lexer->throw_syntax_error("Cannot match an expression as a component name; use </&> instead");
        }
        if ($call_end ne $comp) {
            $self->lexer->throw_syntax_error("Component name in ending tag ($call_end) does not match component name in beginning tag ($comp)");
        }
    }

    my $code = "} }, $call\n );\n";

    eval { $self->postprocess_perl->(\$code) } if $self->postprocess_perl;
    compiler_error $@ if $@;

    $self->_add_body_code($code);

    $c->{last_body_code_type} = 'component_content_call_end';
}

sub perl_line
{
    my $self = shift;
    my %p = @_;

    my $code = "$p{line}\n";

    eval { $self->postprocess_perl->(\$code) } if $self->postprocess_perl;
    compiler_error $@ if $@;

    $self->_add_body_code($code);

    $self->{current_compile}{last_body_code_type} = 'perl_line';
}

sub _add_body_code
{
    my $self = shift;

    # We know a perl-line is always _one_ line, so we know that the
    # line numbers are going to match up as long as the first line in
    # a series has a line number comment before it.  Adding a comment
    # can break certain constructs like qw() list that spans multiple
    # perl-lines.
    if ( $self->lexer->line_number &&
         $self->{current_compile}{last_body_code_type} ne 'perl_line' &&
         $self->use_source_line_numbers )
    {
        my $line = $self->lexer->line_number;
        my $file = $self->_escape_filename( $self->lexer->name );
        $self->{current_compile}{body} .= qq{#line $line "$file"\n};
    }

    $self->{current_compile}{body} .= $_ foreach @_;
}

sub _escape_filename
{
    my $self = shift;
    my $file = shift;

    $file =~ s/\"//g;

    return $file;
}

sub dump
{
    my $self = shift;
    my $c = $self->{current_compile};

    warn "Main component\n";

    $self->_dump_data( $c );

    foreach ( keys %{ $c->{def} } )
    {
        warn "  Subcomponent $_\n";
        $self->_dump_data( $c->{def}{$_}, '  ' );
    }

    foreach ( keys %{ $c->{method} } )
    {
        warn "  Methods $_\n";
        $self->_dump_data( $c->{method}{$_}, '  ');
    }
}

sub _dump_data
{
    my $self = shift;
    my $data = shift;
    my $indent = shift || '';

    if ( @{ $data->{args} } )
    {
        warn "$indent  args\n";
        foreach ( @{ $data->{args} } )
        {
            warn "$indent    $_->{type}$_->{name}";
            warn " => $_->{default}" if defined $_->{default};
            warn "\n";
        }
    }

    warn "\n$indent  body\n";
    warn $data->{body}, "\n";
}

sub _blocks
{
    my $self = shift;

    return @{ $self->{current_compile}{blocks}{ shift() } };
}

sub HTML::Mason::Parser::new
{
    die "The Parser module is no longer a part of HTML::Mason.  Please see ".
        "the Lexer and Compiler modules, its replacements.\n";
}

1;

__END__

=head1 NAME

HTML::Mason::Compiler - Compile Mason component source

=head1 SYNOPSIS

  package My::Funky::Compiler;

  use base qw(HTML::Mason::Compiler);

=head1 DESCRIPTION

The compiler starts the compilation process by calling its lexer's
C<lex> method and passing itself as the C<compiler> parameter.  The
lexer then calls various methods in the compiler as it parses the
component source.

=head1 PARAMETERS TO THE new() CONSTRUCTOR

=over 4

=item allow_globals

List of variable names, complete with prefix (C<$@%>), that you intend
to use as globals in components.  Normally global variables are
forbidden by C<strict>, but any variable mentioned in this list is
granted a reprieve via a "use vars" statement. For example:

    allow_globals => [qw($DBH %session)]

In a mod_perl environment, C<$r> (the request object) is automatically
added to this list.

=item default_escape_flags

Escape flags to apply to all <% %> expressions by default. The current
valid flags are

    h - escape for HTML ('<' => '&lt;', etc.)
    u - escape for URL (':' => '%3A', etc.)

The developer can override default escape flags on a per-expression
basis; see the L<escaping expressions|HTML::Mason::Devel/escaping expressions> section of the developer's manual.

If you want to set I<multiple> flags as the default, this should be
given as a reference to an array of flags.

=item enable_autoflush

True or false, default is true. Indicates whether components are
compiled with support for L<autoflush|HTML::Mason::Params/autoflush>. The component can be compiled
to a more efficient form if it does not have to check for autoflush
mode, so you should set this to 0 if you can.

=item lexer

The Lexer object to associate with this Compiler. By default a new
object of class L<lexer_class|HTML::Mason::Params/lexer_class> will be created.

=item lexer_class

The class to use when creating a lexer. Defaults to L<HTML::Mason::Lexer|HTML::Mason::Lexer>.

=item preprocess

Sub reference that is called to preprocess each component before the compiler does
it's magic.  The sub is called with a single parameter, a scalar reference
to the script.  The sub is expected to process the script in-place.   This is
one way to extend the HTML::Mason syntax with new tags, etc., although a much
more flexible way is to subclass the Lexer or Compiler class. See also
L<postprocess_text|HTML::Mason::Params/postprocess_text> and L<postprocess_perl|HTML::Mason::Params/postprocess_perl>.

=item postprocess_text

Sub reference that is called to postprocess the text portion of a
compiled component, just before it is assembled into its final
subroutine form.  The sub is called with a single parameter, a scalar
reference to the text portion of the component.  The sub is expected
to process the string in-place. See also
L<preprocess|HTML::Mason::Params/preprocess> and L<postprocess_perl|HTML::Mason::Params/postprocess_perl>.

=item postprocess_perl

Sub reference that is called to postprocess the Perl portion of a
compiled component, just before it is assembled into its final
subroutine form.  The sub is called with a single parameter, a scalar
reference to the Perl portion of the component.  The sub is expected
to process the string in-place. See also
L<preprocess|HTML::Mason::Params/preprocess> and L<postprocess_text|HTML::Mason::Params/postprocess_text>.

=item use_source_line_numbers

True or false, default is true. Indicates whether component line
numbers that appear in error messages, stack traces, etc. are in terms
of the source file instead of the object file. Mason does this by
inserting '#line' directives into compiled components.  While source
line numbers are more immediately helpful, object file line numbers
may be more appropriate for in-depth debugging sessions.

=back

=head1 ACCESSOR METHODS

All of the above properties have read-only accessor methods of the
same name.

You cannot change any property of a compiler after it has been created
- among other things, this would potentially invalidate any existing
cached component objects or object files. Your best bet is to create
different compiler objects and load them into different interpreters.

=head1 METHODS

There are several methods besides the compilation callbacks below that
a Compiler subclass needs to implement.

=over 4

=item compile(...)

This method has several parameters:

=over 8

=item * comp_source (required)

Either a scalar or reference to a scalar containing the component source.

=item * name (required)

The name of the component. This should be the filename of the
component if it is file-based, or some other clear identifier of the
component source.

=item * comp_path (required)

This should be the component's path.

=item * fh (optional)

If this is given then the output of the compiler will be sent directly
to this handle, rather than being buffered in memory. This is an
optimization to avoid memory usage.

=back

=item object_id

This method should return a unique id for the given compiler object.
This is used by the interpreter when determining the object directory,
for example.

=back

=head2 Compilation Callbacks

These are methods called by the Lexer while processing a component
source.  You may wish to override some of these methods if you're
implementing your own custom Compiler class.

=over 4

=item start_component()

This method is called by the Lexer when it starts processing a
component.

=item end_component()

This method is called by the Lexer when it finishes processing a
component.

=item start_block(block_type => <string>)

This method is called by the Lexer when it encounters an opening Mason
block tag like C<< <%perl> >> or C<< <%args> >>.  Its main purpose is
to keep track of the nesting of different kinds of blocks within each
other.  The type of block ("init", "once", etc.) is passed via the
"block_type" parameter.

=item end_block(block_type => <string>)

This method is called by the Lexer when it encounters a closing Mason
block tag like C<< </%perl> >> or C<< </%args> >>.  Like
C<start_block()>, its main purpose is to help maintain syntactic
integrity.

=item *_block(block => <string>, [ block_type => <string> ])

Several compiler methods like C<doc_block()>, C<text_block()>, and
C<raw_block()> are called by the Lexer after C<start_block()> when it
encounters blocks of certain types.  These methods actually do the
work of putting the body of a block into the compiled data structure.

The methods that follow this pattern are C<init_block()>,
C<perl_block()>, C<doc_block()>, C<text_block()>, and C<raw_block()>.
The last method is called for all C<< <%once> >>, C<< <%cleanup> >>,
C<< <%filter> >>, C<< <%init> >>, C<< <%perl> >>, and C<< <%shared> >>
blocks.

=item text(text => <string>)

Inserts the text contained in a C<text> parameter into the component
for verbatim output.

This is called when the lexer finds plain text in a component.

=item variable_declaration( type => <string>, name => <string>, default => <string> )

Inserts a variable declaration from the C<< <%args> >> section into
the component.

The type will be either "$", "@", or "%", indicating a scalar, array,
or hash.  The name is the variable name without the leading sigil.
The default is everything found after the first "=>" on an C<< <%args> >>
block line, and may include a comment.

=item key_value_pair(block_type => <string>, key => <string>, value => <string>)

Inserts a key-value pair from a C<< <%flags> >> or C<< <%attr> >>
section into the component.

The "block_type" parameter will be either "flags" or "attr".

=item start_named_block(block_type => <string>, name => <name>)

Analogous to L<item_start_block|start_block>, but starts a "named" block 
(C<< <%method> >> or C<< <%def> >>).

=item end_named_block()

Called by the Lexer to end a "named" block.

=item substitution(substitution => <string>, escape => <string>)

Called by the Lexer when it encounters a substitution tag 
(C<< <% ... %> >>).

The value of the "escape" parameter will be everything found after the
pipe (|) in the substitution tag, and may be more than one character
such as "nh".

=item component_call(call => <string>)

Called by the Lexer when it encounters a component call tag without
embedded content (C<< <& ... &> >>).

The "call" parameter contains the entire contents of the tag.

=item component_content_call(call => <string>)

Called by the Lexer when it encounters a component call tag with
embedded content (C<< <&| ... &> >>).

=item component_content_call_end()

Called by the Lexer when it encounters an ending tag for a component
call with content (C<< </&> >>).  Note that there is no corresponding
C<component_call_end()> method for component calls without content,
because these calls don't have ending tags.

=item perl_line(line => <string>)

Called by the Lexer when it encounters a C<%>-line.

=back

=head1 SUBCLASSING

We recommend that any parameters you add to Compiler be read-only,
because the compiler object_id is only computed once on creation
and would not reflect any changes to Lexer parameters.

=cut