package HTTP::Negotiate;

$VERSION = "6.01";
sub Version { $VERSION; }

require Exporter;
@ISA = qw(Exporter);
@EXPORT = qw(choose);

require HTTP::Headers;

$DEBUG = 0;

sub choose ($;$)
{
    my($variants, $request) = @_;
    my(%accept);

    unless (defined $request) {
	# Create a request object from the CGI environment variables
	$request = HTTP::Headers->new;
	$request->header('Accept', $ENV{HTTP_ACCEPT})
	  if $ENV{HTTP_ACCEPT};
	$request->header('Accept-Charset', $ENV{HTTP_ACCEPT_CHARSET})
	  if $ENV{HTTP_ACCEPT_CHARSET};
	$request->header('Accept-Encoding', $ENV{HTTP_ACCEPT_ENCODING})
	  if $ENV{HTTP_ACCEPT_ENCODING};
	$request->header('Accept-Language', $ENV{HTTP_ACCEPT_LANGUAGE})
	  if $ENV{HTTP_ACCEPT_LANGUAGE};
    }

    # Get all Accept values from the request.  Build a hash initialized
    # like this:
    #
    #   %accept = ( type =>     { 'audio/*'     => { q => 0.2, mbx => 20000 },
    #                             'audio/basic' => { q => 1 },
    #                           },
    #               language => { 'no'          => { q => 1 },
    #                           }
    #             );

    $request->scan(sub {
	my($key, $val) = @_;

	my $type;
	if ($key =~ s/^Accept-//) {
	    $type = lc($key);
	}
	elsif ($key eq "Accept") {
	    $type = "type";
	}
	else {
	    return;
	}

	$val =~ s/\s+//g;
	my $default_q = 1;
	for my $name (split(/,/, $val)) {
	    my(%param, $param);
	    if ($name =~ s/;(.*)//) {
		for $param (split(/;/, $1)) {
		    my ($pk, $pv) = split(/=/, $param, 2);
		    $param{lc $pk} = $pv;
		}
	    }
	    $name = lc $name;
	    if (defined $param{'q'}) {
		$param{'q'} = 1 if $param{'q'} > 1;
		$param{'q'} = 0 if $param{'q'} < 0;
	    }
	    else {
		$param{'q'} = $default_q;

		# This makes sure that the first ones are slightly better off
		# and therefore more likely to be chosen.
		$default_q -= 0.0001;
	    }
	    $accept{$type}{$name} = \%param;
	}
    });

    # Check if any of the variants specify a language.  We do this
    # because it influences how we treat those without (they default to
    # 0.5 instead of 1).
    my $any_lang = 0;
    for $var (@$variants) {
	if ($var->[5]) {
	    $any_lang = 1;
	    last;
	}
    }

    if ($DEBUG) {
	print "Negotiation parameters in the request\n";
	for $type (keys %accept) {
	    print " $type:\n";
	    for $name (keys %{$accept{$type}}) {
		print "    $name\n";
		for $pv (keys %{$accept{$type}{$name}}) {
		    print "      $pv = $accept{$type}{$name}{$pv}\n";
		}
	    }
	}
    }

    my @Q = ();  # This is where we collect the results of the
		 # quality calculations

    # Calculate quality for all the variants that are available.
    for (@$variants) {
	my($id, $qs, $ct, $enc, $cs, $lang, $bs) = @$_;
	$qs = 1 unless defined $qs;
        $ct = '' unless defined $ct;
	$bs = 0 unless defined $bs;
	$lang = lc($lang) if $lang; # lg tags are always case-insensitive
	if ($DEBUG) {
	    print "\nEvaluating $id (ct='$ct')\n";
	    printf "  qs   = %.3f\n", $qs;
	    print  "  enc  = $enc\n"  if $enc && !ref($enc);
	    print  "  enc  = @$enc\n" if $enc && ref($enc);
	    print  "  cs   = $cs\n"   if $cs;
	    print  "  lang = $lang\n" if $lang;
	    print  "  bs   = $bs\n"   if $bs;
	}

	# Calculate encoding quality
	my $qe = 1;
	# If the variant has no assigned Content-Encoding, or if no
	# Accept-Encoding field is present, then the value assigned
	# is "qe=1".  If *all* of the variant's content encodings
	# are listed in the Accept-Encoding field, then the value
	# assigned is "qw=1".  If *any* of the variant's content
	# encodings are not listed in the provided Accept-Encoding
	# field, then the value assigned is "qe=0"
	if (exists $accept{'encoding'} && $enc) {
	    my @enc = ref($enc) ? @$enc : ($enc);
	    for (@enc) {
		print "Is encoding $_ accepted? " if $DEBUG;
		unless(exists $accept{'encoding'}{$_}) {
		    print "no\n" if $DEBUG;
		    $qe = 0;
		    last;
		}
		else {
		    print "yes\n" if $DEBUG;
		}
	    }
	}

	# Calculate charset quality
	my $qc  = 1;
	# If the variant's media-type has no charset parameter,
	# or the variant's charset is US-ASCII, or if no Accept-Charset
	# field is present, then the value assigned is "qc=1".  If the
	# variant's charset is listed in the Accept-Charset field,
	# then the value assigned is "qc=1.  Otherwise, if the variant's
	# charset is not listed in the provided Accept-Encoding field,
	# then the value assigned is "qc=0".
	if (exists $accept{'charset'} && $cs && $cs ne 'us-ascii' ) {
	    $qc = 0 unless $accept{'charset'}{$cs};
	}

	# Calculate language quality
	my $ql  = 1;
	if ($lang && exists $accept{'language'}) {
	    my @lang = ref($lang) ? @$lang : ($lang);
	    # If any of the variant's content languages are listed
	    # in the Accept-Language field, the the value assigned is
	    # the largest of the "q" parameter values for those language
	    # tags.
	    my $q = undef;
	    for (@lang) {
		next unless exists $accept{'language'}{$_};
		my $this_q = $accept{'language'}{$_}{'q'};
		$q = $this_q unless defined $q;
		$q = $this_q if $this_q > $q;
	    }
	    if(defined $q) {
	        $DEBUG and print " -- Exact language match at q=$q\n";
	    }
	    else {
		# If there was no exact match and at least one of
		# the Accept-Language field values is a complete
		# subtag prefix of the content language tag(s), then
		# the "q" parameter value of the largest matching
		# prefix is used.
		$DEBUG and print " -- No exact language match\n";
		my $selected = undef;
		for $al (keys %{ $accept{'language'} }) {
		    if (index($al, "$lang-") == 0) {
		        # $lang starting with $al isn't enough, or else
		        #  Accept-Language: hu (Hungarian) would seem
		        #  to accept a document in hup (Hupa)
		        $DEBUG and print " -- $al ISA $lang\n";
			$selected = $al unless defined $selected;
			$selected = $al if length($al) > length($selected);
		    }
		    else {
		        $DEBUG and print " -- $lang  isn't a $al\n";
		    }
		}
		$q = $accept{'language'}{$selected}{'q'} if $selected;

		# If none of the variant's content language tags or
		# tag prefixes are listed in the provided
		# Accept-Language field, then the value assigned
		# is "ql=0.001"
		$q = 0.001 unless defined $q;
	    }
	    $ql = $q;
	}
	else {
	    $ql = 0.5 if $any_lang && exists $accept{'language'};
	}

	my $q   = 1;
	my $mbx = undef;
	# If no Accept field is given, then the value assigned is "q=1".
	# If at least one listed media range matches the variant's media
	# type, then the "q" parameter value assigned to the most specific
	# of those matched is used (e.g. "text/html;version=3.0" is more
	# specific than "text/html", which is more specific than "text/*",
	# which in turn is more specific than "*/*"). If not media range
	# in the provided Accept field matches the variant's media type,
	# then the value assigned is "q=0".
	if (exists $accept{'type'} && $ct) {
	    # First we clean up our content-type
	    $ct =~ s/\s+//g;
	    my $params = "";
	    $params = $1 if $ct =~ s/;(.*)//;
	    my($type, $subtype) = split("/", $ct, 2);
	    my %param = ();
	    for $param (split(/;/, $params)) {
		my($pk,$pv) = split(/=/, $param, 2);
		$param{$pk} = $pv;
	    }

	    my $sel_q = undef;
	    my $sel_mbx = undef;
	    my $sel_specificness = 0;

	    ACCEPT_TYPE:
	    for $at (keys %{ $accept{'type'} }) {
		print "Consider $at...\n" if $DEBUG;
		my($at_type, $at_subtype) = split("/", $at, 2);
		# Is it a match on the type
		next if $at_type    ne '*' && $at_type    ne $type;
		next if $at_subtype ne '*' && $at_subtype ne $subtype;
		my $specificness = 0;
		$specificness++ if $at_type ne '*';
		$specificness++ if $at_subtype ne '*';
		# Let's see if content-type parameters also match
		while (($pk, $pv) = each %param) {
		    print "Check if $pk = $pv is true\n" if $DEBUG;
		    next unless exists $accept{'type'}{$at}{$pk};
		    next ACCEPT_TYPE
		      unless $accept{'type'}{$at}{$pk} eq $pv;
		    print "yes it is!!\n" if $DEBUG;
		    $specificness++;
		}
		print "Hurray, type match with specificness = $specificness\n"
		  if $DEBUG;

		if (!defined($sel_q) || $sel_specificness < $specificness) {
		    $sel_q   = $accept{'type'}{$at}{'q'};
		    $sel_mbx = $accept{'type'}{$at}{'mbx'};
		    $sel_specificness = $specificness;
		}
	    }
	    $q   = $sel_q || 0;
	    $mbx = $sel_mbx;
	}

	my $Q;
	if (!defined($mbx) || $mbx >= $bs) {
	    $Q = $qs * $qe * $qc * $ql * $q;
	}
	else {
	    $Q = 0;
	    print "Variant's size is too large ==> Q=0\n" if $DEBUG;
	}

	if ($DEBUG) {
	    $mbx = "undef" unless defined $mbx;
	    printf "Q=%.4f", $Q;
	    print "  (q=$q, mbx=$mbx, qe=$qe, qc=$qc, ql=$ql, qs=$qs)\n";
	}

	push(@Q, [$id, $Q, $bs]);
    }


    @Q = sort { $b->[1] <=> $a->[1] || $a->[2] <=> $b->[2] } @Q;

    return @Q if wantarray;
    return undef unless @Q;
    return undef if $Q[0][1] == 0;
    $Q[0][0];
}

1;

__END__


=head1 NAME

HTTP::Negotiate - choose a variant to serve

=head1 SYNOPSIS

 use HTTP::Negotiate qw(choose);

 #  ID       QS     Content-Type   Encoding Char-Set        Lang   Size
 $variants =
  [['var1',  1.000, 'text/html',   undef,   'iso-8859-1',   'en',   3000],
   ['var2',  0.950, 'text/plain',  'gzip',  'us-ascii',     'no',    400],
   ['var3',  0.3,   'image/gif',   undef,   undef,          undef, 43555],
  ];

 @preferred = choose($variants, $request_headers);
 $the_one   = choose($variants);

=head1 DESCRIPTION

This module provides a complete implementation of the HTTP content
negotiation algorithm specified in F<draft-ietf-http-v11-spec-00.ps>
chapter 12.  Content negotiation allows for the selection of a
preferred content representation based upon attributes of the
negotiable variants and the value of the various Accept* header fields
in the request.

The variants are ordered by preference by calling the function
choose().

The first parameter is reference to an array of the variants to
choose among.
Each element in this array is an array with the values [$id, $qs,
$content_type, $content_encoding, $charset, $content_language,
$content_length] whose meanings are described
below. The $content_encoding and $content_language can be either a
single scalar value or an array reference if there are several values.

The second optional parameter is either a HTTP::Headers or a HTTP::Request
object which is searched for "Accept*" headers.  If this
parameter is missing, then the accept specification is initialized
from the CGI environment variables HTTP_ACCEPT, HTTP_ACCEPT_CHARSET,
HTTP_ACCEPT_ENCODING and HTTP_ACCEPT_LANGUAGE.

In an array context, choose() returns a list of [variant
identifier, calculated quality, size] tuples.  The values are sorted by
quality, highest quality first.  If the calculated quality is the same
for two variants, then they are sorted by size (smallest first). I<E.g.>:

  (['var1', 1, 2000], ['var2', 0.3, 512], ['var3', 0.3, 1024]);

Note that also zero quality variants are included in the return list
even if these should never be served to the client.

In a scalar context, it returns the identifier of the variant with the
highest score or C<undef> if none have non-zero quality.

If the $HTTP::Negotiate::DEBUG variable is set to TRUE, then a lot of
noise is generated on STDOUT during evaluation of choose().

=head1 VARIANTS

A variant is described by a list of the following values.  If the
attribute does not make sense or is unknown for a variant, then use
C<undef> instead.

=over 3

=item identifier

This is a string that you use as the name for the variant.  This
identifier for the preferred variants returned by choose().

=item qs

This is a number between 0.000 and 1.000 that describes the "source
quality".  This is what F<draft-ietf-http-v11-spec-00.ps> says about this
value:

Source quality is measured by the content provider as representing the
amount of degradation from the original source.  For example, a
picture in JPEG form would have a lower qs when translated to the XBM
format, and much lower qs when translated to an ASCII-art
representation.  Note, however, that this is a function of the source
- an original piece of ASCII-art may degrade in quality if it is
captured in JPEG form.  The qs values should be assigned to each
variant by the content provider; if no qs value has been assigned, the
default is generally "qs=1".

=item content-type

This is the media type of the variant.  The media type does not
include a charset attribute, but might contain other parameters.
Examples are:

  text/html
  text/html;version=2.0
  text/plain
  image/gif
  image/jpg

=item content-encoding

This is one or more content encodings that has been applied to the
variant.  The content encoding is generally used as a modifier to the
content media type.  The most common content encodings are:

  gzip
  compress

=item content-charset

This is the character set used when the variant contains text.
The charset value should generally be C<undef> or one of these:

  us-ascii
  iso-8859-1 ... iso-8859-9
  iso-2022-jp
  iso-2022-jp-2
  iso-2022-kr
  unicode-1-1
  unicode-1-1-utf-7
  unicode-1-1-utf-8

=item content-language

This describes one or more languages that are used in the variant.
Language is described like this in F<draft-ietf-http-v11-spec-00.ps>: A
language is in this context a natural language spoken, written, or
otherwise conveyed by human beings for communication of information to
other human beings.  Computer languages are explicitly excluded.

The language tags are defined by RFC 3066.  Examples
are:

  no               Norwegian
  en               International English
  en-US            US English
  en-cockney

=item content-length

This is the number of bytes used to represent the content.

=back

=head1 ACCEPT HEADERS

The following Accept* headers can be used for describing content
preferences in a request (This description is an edited extract from
F<draft-ietf-http-v11-spec-00.ps>):

=over 3

=item Accept

This header can be used to indicate a list of media ranges which are
acceptable as a response to the request.  The "*" character is used to
group media types into ranges, with "*/*" indicating all media types
and "type/*" indicating all subtypes of that type.

The parameter q is used to indicate the quality factor, which
represents the user's preference for that range of media types.  The
parameter mbx gives the maximum acceptable size of the response
content. The default values are: q=1 and mbx=infinity. If no Accept
header is present, then the client accepts all media types with q=1.

For example:

  Accept: audio/*;q=0.2;mbx=200000, audio/basic

would mean: "I prefer audio/basic (of any size), but send me any audio
type if it is the best available after an 80% mark-down in quality and
its size is less than 200000 bytes"


=item Accept-Charset

Used to indicate what character sets are acceptable for the response.
The "us-ascii" character set is assumed to be acceptable for all user
agents.  If no Accept-Charset field is given, the default is that any
charset is acceptable.  Example:

  Accept-Charset: iso-8859-1, unicode-1-1


=item Accept-Encoding

Restricts the Content-Encoding values which are acceptable in the
response.  If no Accept-Encoding field is present, the server may
assume that the client will accept any content encoding.  An empty
Accept-Encoding means that no content encoding is acceptable.  Example:

  Accept-Encoding: compress, gzip


=item Accept-Language

This field is similar to Accept, but restricts the set of natural
languages that are preferred in a response.  Each language may be
given an associated quality value which represents an estimate of the
user's comprehension of that language.  For example:

  Accept-Language: no, en-gb;q=0.8, de;q=0.55

would mean: "I prefer Norwegian, but will accept British English (with
80% comprehension) or German (with 55% comprehension).

=back


=head1 COPYRIGHT

Copyright 1996,2001 Gisle Aas.

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

=head1 AUTHOR

Gisle Aas <gisle@aas.no>

=cut