support ";\s*" separater and valuless keys

[kazeburo]

Added suppoting ";" separator. W3C recommends that all web servers support ; separators in addition to & separators. This p-r also supports space after separator, like this RT.
https://rt.cpan.org/Public/Bug/Display.html?id=31055

And I added valueless keys support. When foo&bar=baz is given, foo parameter is not accessible. Plack::Request support this.

[chansen]

Before going further we really need to define the application/www-form-urlencoded format. I propose that we implement the draft written by @hoehrmann. That way we would get consistent behavior between implementations.

I noticed that Plack recently changed it's decoder, plack/Plack#445.

@miyagawa and @yanik would you care to join the discussion so we can settle this for once?

#!/usr/bin/perl
use strict;
use warnings;

sub decode {
    my ($encoded) = @_;

    my @params;
    foreach my $pair (split /[&;]/, $encoded, -1) {
        my ($name, $value) = split /[=]/, $pair, 2;
        $name = '' unless defined $name;
        for ($name, $value) {
            next unless defined;
            y/+/\x20/;
            s/%([0-9a-fA-F]{2})/chr(hex($1))/ge;
        }
        push @params, $name, $value;
    }
    return \@params;
}

# Tests based on application/www-form-urlencoded draft written by B. Hoehrmann
# <http://tools.ietf.org/html/draft-hoehrmann-urlencoded-01>

my @tests = (
    [ 'a;b',                => [ 'a'    => undef, 'b'  => undef ] ],
    [ 'a ; b',              => [ 'a '   => undef, ' b' => undef ] ],
    [ 'a==1;b==2',          => [ 'a'    => '=1',  'b'  => '=2'  ] ],
    [ 'Fo%2=',              => [ 'Fo%2' => ''                   ] ],
    [ ' a = 1 '             => [ ' a '  => ' 1 '                ] ],
    [ '+a+=+1+'             => [ ' a '  => ' 1 '                ] ],
    [ '%20a%20=%201%20'     => [ ' a '  => ' 1 '                ] ],
    [ ';'                   => [ ''     => undef, ''   => undef ] ],
    [ ';='                  => [ ''     => undef, ''   => ''    ] ],
    [ '=;'                  => [ ''     => '',    ''   => undef ] ],
    [ '=;='                 => [ ''     => '',    ''   => ''    ] ],
    [ '=',                  => [ ''     => '',                  ] ],
    [ '',                   => [                                ] ],
);

use Test::More;
foreach my $test (@tests) {
    my ($string, $expected) = @$test;
    is_deeply(decode($string), $expected, qq["$string"]);
}

done_testing();

[kazeburo]

I found another document about application/x-www-form-urlencoded
http://www.w3.org/TR/html5/forms.html#url-encoded-form-data
ruby's uri module refers this
http://docs.ruby-lang.org/en/2.0.0/URI.html

[chansen]

W3's definition of application/www-form-urlencoded is mostly useless as it focus on encoding of perfect data. I would like to bring on @kraih to this discussion, I'm sure he has some valuable POV's.

[chansen]

Considering that the current implementation process millions of requests every day (POST and GET requests) and is mostly in agreement with W3 and IETF I'm going to reject this pull request.

[chansen]

By popular demand, I have implemented support for ";" as a separator and added support for parameters without value (i.e. without '=' sign).