package PLP::Fields;

use strict;
use warnings;

our $VERSION = '1.00';

# Has only one function: doit(), which ties the hashes %get, %post, %fields
# and %header in PLP::Script. Also generates %cookie immediately.
sub doit {

	# %get
	
	my $get = \%PLP::Script::get;
	if (defined $ENV{QUERY_STRING} and length $ENV{QUERY_STRING}){
		for (split /[&;]/, $ENV{QUERY_STRING}) {
			my @keyval = split /=/, $_, 2;
			PLP::Functions::DecodeURI(@keyval);
			$get->{$keyval[0]} = $keyval[1] unless $keyval[0] =~ /^\@/;
			push @{ $get->{ '@' . $keyval[0] } }, $keyval[1];
		}
	}

	# %post

	tie %PLP::Script::post, 'PLP::Tie::Delay', 'PLP::Script::post', sub {
		my %post;
		return \%post unless $ENV{CONTENT_TYPE} and $ENV{CONTENT_LENGTH} and
			$ENV{CONTENT_TYPE} =~ m!^(?:application/x-www-form-urlencoded|$)!;
		
		my $post = $PLP::read->($ENV{CONTENT_LENGTH});
		return \%post unless defined $post and length $post;
		
		for (split /&/, $post) {
			my @keyval = split /=/, $_, 2;
			PLP::Functions::DecodeURI(@keyval);
			$post{$keyval[0]} = $keyval[1] unless $keyval[0] =~ /^\@/;
			push @{ $post{ '@' . $keyval[0] } }, $keyval[1];
		}
		
		return \%post;
	};

	# %fields

	tie %PLP::Script::fields, 'PLP::Tie::Delay', 'PLP::Script::fields', sub {
		return { %PLP::Script::get, %PLP::Script::post };
	};

	# %header

	tie %PLP::Script::header, 'PLP::Tie::Headers';

	# %cookie

	if (defined $ENV{HTTP_COOKIE} and length $ENV{HTTP_COOKIE}) {
		for (split /; ?/, $ENV{HTTP_COOKIE}) {
			my @keyval = split /=/, $_, 2;
			$PLP::Script::cookie{$keyval[0]} ||= $keyval[1];
		}
	}
}

1;

=head1 NAME

PLP::Fields - Special hashes for PLP

=head1 DESCRIPTION

For your convenience, PLP uses hashes to put things in. Some of these are tied
hashes, so they contain a bit magic. For example, building the hash can be
delayed until you actually use the hash.

=over 10

=item C<%get> and C<%post>

These are built from the C<key=value&key=value> (or C<key=value;key=value>
strings in query string and post content. C<%post> is not built if the content
type is not C<application/x-www-form-urlencoded>. In post content, the
semi-colon is not a valid separator.

%post isn't built until it is used, to speed up your script if you
don't use it. Because POST content can only be read once, you can C<use CGI;>
and just never access C<%post> to avoid its building.

With a query string of C<key=firstvalue&key=secondvalue>, C<$get{key}> will
contain only C<secondvalue>. You can access both elements by using the array
reference C<$get{'@key'}>, which will contain C<[ 'firstvalue', 'secondvalue'
]>.

=item C<%fields>

This hash combines %get and %post, and triggers creation of %post. POST gets
precedence over GET (note: not even the C<@>-keys contain both values).

This hash is built on first use, just like %post.

=item C<%cookie>, C<%cookies>

This is built immediately, because cookies are usually short in length. Cookies
are B<not> automatically url-decoded.

=item C<%header>, C<%headers>

This is a hash of HTTP headers to accompany the first output.
By default it will contain C<X-PLP-Version> to identify the serving module,
and C<Content-Type> set to C<text/html>.

If STDOUT has been opened as :utf8,
a C<charset=utf-8> attribute will automatically be added.
This will not be possible with FastCGI
because FCGI (as of version 0.74) does not support output layers.

Headers can be added and/or changed as long as they have not yet been sent.
Underscores in key names are converted to normal minus signs,
so you can leave out quotes.  The hash is case insensitive: the case used
when sending the headers is the one you used first. The following are equal:

    $header{CONTENT_TYPE}
    $header{'Content-Type'}
    $header{Content_Type}
    $headers{CONTENT_type}

If a value contains newlines, the header is repeated for each line:

	$header{Allow} = "HEAD\nGET";  # equivalent to HEAD,GET

For example, to send out a non-HTML text instead of the default HTML:

    <:
    $header{content_type} = 'text/plain';
    use open ':std', ':utf8';
    :>
    This text should be prefixed by the following header:
        Content-Type: text/plain; charset=utf-8

=back

=head1 AUTHOR

Juerd Waalboer <juerd@cpan.org>

Current maintainer: Mischa POSLAWSKY <shiar@cpan.org>

=cut