source: trunk/gsdl/perllib/plugins/HTMLPlug.pm@ 10513

###########################################################################
#
# HTMLPlug.pm -- basic html plugin
#
# A component of the Greenstone digital library software
# from the New Zealand Digital Library Project at the 
# University of Waikato, New Zealand.
#
# Copyright (C) 1999 New Zealand Digital Library Project
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
#
###########################################################################

# 
# Note that this plugin handles frames only in a very simple way
# i.e. each frame is treated as a separate document. This means
# search results will contain links to individual frames rather
# than linking to the top level frameset.
# There may also be some problems caused by the _parent target
# (it's removed by this plugin)
#

package HTMLPlug;

use BasPlug;
use ghtml;
use unicode;
use util;
use XMLParser;

sub BEGIN {
    @HTMLPlug::ISA = ('BasPlug');
}

use strict; # every perl program should have this!
no strict 'refs'; # make an exception so we can use variables as filehandles

my $arguments =
    [ { 'name' => "process_exp",
    'desc' => "{BasPlug.process_exp}",
    'type' => "regexp",
    'deft' =>  &get_default_process_exp() },
      { 'name' => "block_exp",
    'desc' => "{BasPlug.block_exp}",
    'type' => 'regexp',
    'deft' =>  &get_default_block_exp() },
      { 'name' => "nolinks",
    'desc' => "{HTMLPlug.nolinks}",
    'type' => "flag" },
      { 'name' => "keep_head",
    'desc' => "{HTMLPlug.keep_head}",
    'type' => "flag" },
      { 'name' => "no_metadata",
    'desc' => "{HTMLPlug.no_metadata}",
    'type' => "flag" },
      { 'name' => "metadata_fields",
    'desc' => "{HTMLPlug.metadata_fields}",
    'type' => "string",
    'deft' => "Title" },
      { 'name' => "hunt_creator_metadata",
    'desc' => "{HTMLPlug.hunt_creator_metadata}",
    'type' => "flag" },
      { 'name' => "file_is_url",
    'desc' => "{HTMLPlug.file_is_url}",
    'type' => "flag" },
      { 'name' => "assoc_files",
    'desc' => "{HTMLPlug.assoc_files}",
    'type' => "regexp",
    'deft' => &get_default_block_exp() },
      { 'name' => "rename_assoc_files",
    'desc' => "{HTMLPlug.rename_assoc_files}",
    'type' => "flag" },
      { 'name' => "title_sub",
    'desc' => "{HTMLPlug.title_sub}",
    'type' => "string", 
    'deft' => "" },
      { 'name' => "description_tags",
    'desc' => "{HTMLPlug.description_tags}",
    'type' => "flag" },
      # retain this for backward compatibility (w3mir option was replaced by
      # file_is_url)
      { 'name' => "w3mir",
#   'desc' => "{HTMLPlug.w3mir}",
    'type' => "flag",
    'hiddengli' => "yes"},
      { 'name' => "no_strip_metadata_html",
    'desc' => "{HTMLPlug.no_strip_metadata_html}",
    'type' => "string",
    'deft' => "",
    'reqd' => "no"},
      { 'name' => "sectionalise_using_h_tags",
    'desc' => "{HTMLPlug.sectionalise_using_h_tags}",
    'type' => "flag" }
      ];

my $options = { 'name'     => "HTMLPlug",
        'desc'     => "{HTMLPlug.desc}",
        'abstract' => "no",
        'inherits' => "yes",
        'args'     => $arguments };

sub new {
    my ($class) = shift (@_);
    my ($pluginlist,$inputargs,$hashArgOptLists) = @_;
    push(@$pluginlist, $class);
    
    if(defined $arguments){ push(@{$hashArgOptLists->{"ArgList"}},@{$arguments});}
    if(defined $options) { push(@{$hashArgOptLists->{"OptList"}},$options)};
   

    my $self = (defined $hashArgOptLists)? new BasPlug($pluginlist,$inputargs,$hashArgOptLists): new BasPlug($pluginlist,$inputargs);
    
    if ($self->{'w3mir'}) {
    $self->{'file_is_url'} = 1;
    }
    $self->{'aux_files'} = {};
    $self->{'dir_num'} = 0;
    $self->{'file_num'} = 0;
    
    return bless $self, $class;
}

# may want to use (?i)\.(gif|jpe?g|jpe|png|css|js(?:@.*)?)$
# if have eg <script language="javascript" src="img/lib.js@123">
sub get_default_block_exp {
    my $self = shift (@_);
    
    return q^(?i)\.(gif|jpe?g|jpe|jpg|png|css)$^;
}

sub get_default_process_exp {
    my $self = shift (@_);
    
    # the last option is an attempt to encode the concept of an html query ...
    return q^(?i)(\.html?|\.shtml|\.shm|\.asp|\.php\d?|\.cgi|.+\?.+=.*)$^;
}

sub store_block_files
{
    my $self =shift (@_);
    my ($filename) = @_;
    my $html_fname = $filename;
    my @file_blocks;
    
    my ($language, $encoding) = $self->textcat_get_language_encoding ($filename);

    # read in file ($text will be in utf8)
    my $text = "";
    $self->read_file ($filename, $encoding, $language, \$text);
    my $textref = \$text;
    my $opencom = '(?:<!--|&lt;!(?:&mdash;|&#151;|--))';
    my $closecom = '(?:-->|(?:&mdash;|&#151;|--)&gt;)';
    $$textref =~ s/$opencom(.*?)$closecom//gs;

    my $attval = "\\\"[^\\\"]+\\\"|[^\\s>]+";
    my @img_matches = ($$textref =~ m/<img[^>]*?src\s*=\s*($attval)[^>]*>/igs);
    my @usemap_matches = ($$textref =~ m/<img[^>]*?usemap\s*=\s*($attval)[^>]*>/igs);
    my @link_matches = ($$textref =~ m/<link[^>]*?href\s*=\s*($attval)[^>]*>/igs);
    my @embed_matches = ($$textref =~ m/<embed[^>]*?src\s*=\s*($attval)[^>]*>/igs);
    my @tabbg_matches = ($$textref =~ m/<(?:table|tr|td)[^>]*?background\s*=\s*($attval)[^>]*>/igs);

    foreach my $link (@img_matches, @usemap_matches, @link_matches, @embed_matches, @tabbg_matches) {

    # remove quotes from link at start and end if necessary
    if ($link=~/^\"/) {
        $link=~s/^\"//;
        $link=~s/\"$//;
    }

    $link =~ s/\#.*$//s; # remove any anchor names, e.g. foo.html#name becomes foo.html

    if ($link !~ m@^/@ && $link !~ m/^([A-Z]:?)\\/) {
        # Turn relative file path into full path
        my $dirname = &File::Basename::dirname($filename);
        $link = &util::filename_cat($dirname, $link);
    }
    $link = $self->eval_dir_dots($link);
    
    $self->{'file_blocks'}->{$link} = 1;
    }
}

  
# do plugin specific processing of doc_obj
sub process {
    my $self = shift (@_);
    my ($textref, $pluginfo, $base_dir, $file, $metadata, $doc_obj, $gli) = @_;
    my $outhandle = $self->{'outhandle'};

    print STDERR "<Processing n='$file' p='HTMLPlug'>\n" if ($gli);

    print $outhandle "HTMLPlug: processing $file\n"
    if $self->{'verbosity'} > 1;

    if ($ENV{'GSDLOS'} =~ /^windows/i) {
      # this makes life so much easier... perl can cope with unix-style '/'s.
      $base_dir =~ s@(\\)+@/@g;
      $file =~ s@(\\)+@/@g; 
    }
    
    # reset per-doc stuff...
    $self->{'aux_files'} = {};
    $self->{'dir_num'} = 0;
    $self->{'file_num'} = 0;

    # process an HTML file where sections are divided by headings tags (H1, H2 ...)
    # you can also include metadata in the format (X can be any number)
    # <hX>Title<!--gsdl-metadata
    #   <Metadata name="name1">value1</Metadata>
    #   ...
    #   <Metadata name="nameN">valueN</Metadata>
    #--></hX>
    if ($self->{'sectionalise_using_h_tags'}) {
    # description_tags should allways be activated because we convert headings to description tags
    $self->{'description_tags'} = 1;

    my $arrSections = [];
    $$textref =~ s/<h([0-9]+)[^>]*>(.*?)<\/h[0-9]+>/$self->process_heading($1, $2, $arrSections, $file)/isge;

    if (scalar(@$arrSections)) {
        my $strMetadata = $self->update_section_data($arrSections, -1);
        if (length($strMetadata)) {
        $strMetadata = '<!--' . $strMetadata . "\n-->\n</body>";
        $$textref =~ s/<\/body>/$strMetadata/ig;
        }
    }
    }

    my $cursection = $doc_obj->get_top_section();

    $self->extract_metadata ($textref, $metadata, $doc_obj, $cursection)
    unless $self->{'no_metadata'} || $self->{'description_tags'};

    # Store URL for page as metadata - this can be used for an
    # altavista style search interface. The URL won't be valid
    # unless the file structure contains the domain name (i.e.
    # like when w3mir is used to download a website).

    # URL metadata (even invalid ones) are used to support internal
    # links, so even if 'file_is_url' is off, still need to store info

    my $web_url = "http://$file";
    $doc_obj->add_metadata($cursection, "URL", $web_url);

    if ($self->{'file_is_url'}) {
    $doc_obj->add_metadata($cursection, "weblink", "<a href=\"$web_url\">");
    $doc_obj->add_metadata($cursection, "webicon", "_iconworld_");
    $doc_obj->add_metadata($cursection, "/weblink", "</a>");
    }

    if ($self->{'description_tags'}) {
    # remove the html header - note that doing this here means any
    # sections defined within the header will be lost (so all <Section>
    # tags must appear within the body of the HTML)
    my ($head_keep) = ($$textref =~ m/^(.*?)<body[^>]*>/is);

    $$textref =~ s/^.*?<body[^>]*>//is;
    $$textref =~ s/(<\/body[^>]*>|<\/html[^>]*>)//isg;

    my $opencom = '(?:<!--|&lt;!(?:&mdash;|&#151;|--))';
    my $closecom = '(?:-->|(?:&mdash;|&#151;|--)&gt;)';

    my $lt = '(?:<|&lt;)';
    my $gt = '(?:>|&gt;)';
    my $quot = '(?:"|&quot;|&rdquo;|&ldquo;)';

    my $dont_strip = '';
    if ($self->{'no_strip_metadata_html'}) {
        ($dont_strip = $self->{'no_strip_metadata_html'}) =~ s{,}{|}g;
    }

    my $found_something = 0; my $top = 1;
    while ($$textref =~ s/^(.*?)$opencom(.*?)$closecom//s) {
        my $text = $1;
        my $comment = $2;
        if (defined $text) {
        # text before a comment - note that getting to here
        # doesn't necessarily mean there are Section tags in
        # the document
        $self->process_section(\$text, $base_dir, $file, $doc_obj, $cursection);
        }
        while ($comment =~ s/$lt(.*?)$gt//s) {
        my $tag = $1;
        if ($tag eq "Section") {
            $found_something = 1;
            $cursection = $doc_obj->insert_section($doc_obj->get_end_child($cursection)) unless $top;
            $top = 0;
        } elsif ($tag eq "/Section") {
            $found_something = 1;
            $cursection = $doc_obj->get_parent_section ($cursection);
        } elsif ($tag =~ /^Metadata name=$quot(.*?)$quot/s) {
            my $metaname = $1;
            my $accumulate = $tag =~ /mode=${quot}accumulate${quot}/ ? 1 : 0;
            $comment =~ s/^(.*?)$lt\/Metadata$gt//s;
            my $metavalue = $1;
            $metavalue =~ s/^\s+//;
            $metavalue =~ s/\s+$//;
                    # assume that no metadata value intentionally includes
                    # carriage returns or HTML tags (if they're there they
                    # were probably introduced when converting to HTML from
                    # some other format).
            # actually some people want to have html tags in their
            # metadata.
            $metavalue =~ s/[\cJ\cM]/ /sg;
            $metavalue =~ s/<[^>]+>//sg
            unless $dont_strip && ($dont_strip eq 'all' || $metaname =~ /^($dont_strip)$/);
            $metavalue =~ s/\s+/ /sg;
            if ($accumulate) {
            $doc_obj->add_utf8_metadata($cursection, $metaname, $metavalue);
            } else {
            $doc_obj->set_utf8_metadata_element($cursection, $metaname, $metavalue);    
            }
        } elsif ($tag eq "Description" || $tag eq "/Description") {
            # do nothing with containing Description tags
        } else {
            # simple HTML tag (probably created by the conversion
            # to HTML from some other format) - we'll ignore it and
            # hope for the best ;-)
        }
        }
    }
    if ($cursection ne "") {
        print $outhandle "HTMLPlug: WARNING: $file contains unmatched <Section></Section> tags\n";
    }

    $$textref =~ s/^.*?<body[^>]*>//is;
    $$textref =~ s/(<\/body[^>]*>|<\/html[^>]*>)//isg;
    if ($$textref =~ /\S/) {
        if (!$found_something) {
        if ($self->{'verbosity'} > 2) {
            print $outhandle "HTMLPlug: WARNING: $file appears to contain no Section tags so\n";
            print $outhandle "          will be processed as a single section document\n";
        }

        # go ahead and process single-section document
        $self->process_section($textref, $base_dir, $file, $doc_obj, $cursection);

        # if document contains no Section tags we'll go ahead
        # and extract metadata (this won't have been done
        # above as the -description_tags option prevents it)
        my $complete_text = $head_keep.$doc_obj->get_text($cursection);
        $self->extract_metadata (\$complete_text, $metadata, $doc_obj, $cursection) 
            unless $self->{'no_metadata'};

        } else {
        print $outhandle "HTMLPlug: WARNING: $file contains the following text outside\n";
        print $outhandle "          of the final closing </Section> tag. This text will\n";
        print $outhandle "          be ignored.";

        my ($text);
        if (length($$textref) > 30) {
            $text = substr($$textref, 0, 30) . "...";
        } else {
            $text = $$textref;
        }
        $text =~ s/\n/ /isg;
        print $outhandle " ($text)\n";
        }
    } elsif (!$found_something) {

        if ($self->{'verbosity'} > 2) {
        # may get to here if document contained no valid Section
        # tags but did contain some comments. The text will have
        # been processed already but we should print the warning
        # as above and extract metadata
        print $outhandle "HTMLPlug: WARNING: $file appears to contain no Section tags and\n";
        print $outhandle "          is blank or empty.  Metadata will be assigned if present.\n";
        }

        my $complete_text = $head_keep.$doc_obj->get_text($cursection);
        $self->extract_metadata (\$complete_text, $metadata, $doc_obj, $cursection) 
        unless $self->{'no_metadata'};
    }
    
    } else {

    # remove header and footer
    if (!$self->{'keep_head'} || $self->{'description_tags'}) {
        $$textref =~ s/^.*?<body[^>]*>//is;
        $$textref =~ s/(<\/body[^>]*>|<\/html[^>]*>)//isg;
    }

    # single section document
    $self->process_section($textref, $base_dir, $file, $doc_obj, $cursection);
    }
    return 1;
}


sub process_heading
{
    my ($self, $nHeadNo, $strHeadingText, $arrSections, $file) = @_;
    $strHeadingText = '' if (!defined($strHeadingText));

    my $strMetadata = $self->update_section_data($arrSections, int($nHeadNo));

    my $strSecMetadata = '';
    while ($strHeadingText =~ s/<!--gsdl-metadata(.*?)-->//is)
    {
    $strSecMetadata .= $1;
    }

    $strHeadingText =~ s/^\s+//g;
    $strHeadingText =~ s/\s+$//g;
    $strSecMetadata =~ s/^\s+//g;
    $strSecMetadata =~ s/\s+$//g;

    $strMetadata .= "\n<Section>\n\t<Description>\n\t\t<Metadata name=\"Title\">" . $strHeadingText . "</Metadata>\n";

    if (length($strSecMetadata)) {
    $strMetadata .= "\t\t" . $strSecMetadata . "\n";
    }

    $strMetadata .= "\t</Description>\n";

    return "<!--" . $strMetadata . "-->";
}


sub update_section_data
{
    my ($self, $arrSections, $nCurTocNo) = @_;
    my ($strBuffer, $nLast, $nSections) = ('', 0, scalar(@$arrSections));

    if ($nSections == 0) {
    push @$arrSections, $nCurTocNo;
    return $strBuffer;
    }
    $nLast = $arrSections->[$nSections - 1];
    if ($nCurTocNo > $nLast) {
    push @$arrSections, $nCurTocNo;
    return $strBuffer;
    }
    for(my $i = $nSections - 1; $i >= 0; $i--) {
    if ($nCurTocNo <= $arrSections->[$i]) {
        $strBuffer .= "\n</Section>";
        pop @$arrSections;
    }
    }
    push @$arrSections, $nCurTocNo;
    return $strBuffer;
}


# note that process_section may be called multiple times for a single
# section (relying on the fact that add_utf8_text appends the text to any
# that may exist already).
sub process_section {
    my $self = shift (@_);
    my ($textref, $base_dir, $file, $doc_obj, $cursection) = @_;
    # trap links
    if (!$self->{'nolinks'}) {

    # usemap="./#index" not handled correctly => change to "#index"
    $$textref =~ s/(<img[^>]*?usemap\s*=\s*\"?)([^\">\s]+)(\"?[^>]*>)/
        $self->replace_usemap_links($1, $2, $3)/isge;

    $$textref =~ s/(<(?:a|area|frame|link|script)\s+[^>]*?\s*(?:href|src)\s*=\s*\"?)([^\">\s]+)(\"?[^>]*>)/
        $self->replace_href_links ($1, $2, $3, $base_dir, $file, $doc_obj, $cursection)/isge;
    }

    # trap images

    # allow spaces if inside quotes - jrm21
    $$textref =~ s/(<(?:img|embed|table|tr|td)[^>]*?(?:src|background)\s*=\s*)(\"[^\"]+\"|[^\s>]+)([^>]*>)/
    $self->replace_images ($1, $2, $3, $base_dir, $file, $doc_obj, $cursection)/isge;

    # add text to document object
    # turn \ into \\ so that the rest of greenstone doesn't think there
    # is an escape code following. (Macro parsing loses them...)
    $$textref =~ s/\\/\\\\/go;
    
    $doc_obj->add_utf8_text($cursection, $$textref);
}

sub replace_images {
    my $self = shift (@_);
    my ($front, $link, $back, $base_dir, 
    $file, $doc_obj, $section) = @_;

    # remove quotes from link at start and end if necessary
    if ($link=~/^\"/) {
    $link=~s/^\"//;$link=~s/\"$//;
    $front.='"';
    $back="\"$back";
    }

    $link =~ s/\n/ /g;

    # Hack to overcome Windows wv 0.7.1 bug that causes embedded images to be broken
    # If the Word file path has spaces in it, wv messes up and you end up with
    #   absolute paths for the images, and without the "file://" prefix
    # So check for this special case and massage the data to be correct
    if ($ENV{'GSDLOS'} =~ /^windows/i && $self->{'plugin_type'} eq "WordPlug" && $link =~ /^[A-Za-z]\:\\/) {
    $link =~ s/^.*\\([^\\]+)$/$1/;
    }

    my ($href, $hash_part, $rl) = $self->format_link ($link, $base_dir, $file);

    my $img_file =  $self->add_file ($href, $rl, $hash_part, $base_dir, $doc_obj, $section);

    my $anchor_name = $img_file;
    $anchor_name =~ s/^.*\///;
    $anchor_name = "<a name=\"$anchor_name\">";

    return $front . $img_file . $back . $anchor_name;
}

sub replace_href_links {
    my $self = shift (@_);
    my ($front, $link, $back, $base_dir, $file, $doc_obj, $section) = @_;

    # attempt to sort out targets - frames are not handled 
    # well in this plugin and some cases will screw things
    # up - e.g. the _parent target (so we'll just remove 
    # them all ;-)
    $front =~ s/(target=\"?)_top(\"?)/$1_gsdltop_$2/is;
    $back =~ s/(target=\"?)_top(\"?)/$1_gsdltop_$2/is;
    $front =~ s/target=\"?_parent\"?//is;
    $back =~ s/target=\"?_parent\"?//is;

    return $front . $link . $back if $link =~ /^\#/s;
    $link =~ s/\n/ /g;

    my ($href, $hash_part, $rl) = $self->format_link ($link, $base_dir, $file);
    # href may use '\'s where '/'s should be on Windows
    $href =~ s/\\/\//g;

    my ($filename) = $href =~ /^(?:.*?):(?:\/\/)?(.*)/;
    

    ##### leave all these links alone (they won't be picked up by intermediate 
    ##### pages). I think that's safest when dealing with frames, targets etc.
    ##### (at least until I think of a better way to do it). Problems occur with
    ##### mailto links from within small frames, the intermediate page is displayed
    ##### within that frame and can't be seen. There is still potential for this to
    ##### happen even with html pages - the solution seems to be to somehow tell
    ##### the browser from the server side to display the page being sent (i.e. 
    ##### the intermediate page) in the top level window - I'm not sure if that's 
    ##### possible - the following line should probably be deleted if that can be done
    return $front . $link . $back if $href =~ /^(mailto|news|gopher|nntp|telnet|javascript):/is;


    if (($rl == 0) || ($filename =~ /$self->{'process_exp'}/) || 
    ($href =~ /\/$/) || ($href =~ /^(mailto|news|gopher|nntp|telnet|javascript):/i)) {
    &ghtml::urlsafe ($href);
    return $front . "_httpextlink_&rl=" . $rl . "&href=" . $href . $hash_part . $back;
    } else {
    # link is to some other type of file (eg image) so we'll
    # need to associate that file
    return $front . $self->add_file ($href, $rl, $hash_part, $base_dir, $doc_obj, $section) . $back;
    }
}

sub add_file {
    my $self = shift (@_);
    my ($href, $rl, $hash_part, $base_dir, $doc_obj, $section) = @_;
    my ($newname);

    my $filename = $href;
    if ($base_dir eq "") {
    # remove http:/ thereby leaving one slash at the start
    $filename =~ s/^[^:]*:\///;
    }
    else {
    # remove http://
    $filename =~ s/^[^:]*:\/\///;
    }

    $filename = &util::filename_cat($base_dir, $filename);
       
    # Replace %20's in URL with a space if required. Note that the filename
    # may include the %20 in some situations
    if ($filename =~ /\%20/) {
    if (!-e $filename) {
        $filename =~ s/\%20/ /g;
    }
    }

    my ($ext) = $filename =~ /(\.[^\.]*)$/;

    if ($rl == 0) {
    if ((!defined $ext) || ($ext !~ /$self->{'assoc_files'}/)) {
        return "_httpextlink_&rl=0&el=prompt&href=" . $href . $hash_part;
    }
    else {
        return "_httpextlink_&rl=0&el=direct&href=" . $href . $hash_part;
    }
    }

    if ((!defined $ext) || ($ext !~ /$self->{'assoc_files'}/)) {
    return "_httpextlink_&rl=" . $rl . "&href=" . $href . $hash_part;
    }
    if ($self->{'rename_assoc_files'}) {
    if (defined $self->{'aux_files'}->{$href}) {
        $newname = $self->{'aux_files'}->{$href}->{'dir_num'} . "/" .
        $self->{'aux_files'}->{$href}->{'file_num'} . $ext;
    } else {
        $newname = $self->{'dir_num'} . "/" . $self->{'file_num'} . $ext;
        $self->{'aux_files'}->{$href} = {'dir_num' => $self->{'dir_num'}, 'file_num' => $self->{'file_num'}};
        $self->inc_filecount ();
    }
    $doc_obj->associate_file($filename, $newname, undef, $section);
    return "_httpdocimg_/$newname";
    } else {
    ($newname) = $filename =~ /([^\/\\]*)$/;
    $doc_obj->associate_file($filename, $newname, undef, $section);
    return "_httpdocimg_/$newname";
    }
}


sub format_link {
    my $self = shift (@_);
    my ($link, $base_dir, $file) = @_;

    my ($before_hash, $hash_part) = $link =~ /^([^\#]*)(\#?.*)$/;
    
    $hash_part = "" if !defined $hash_part;
    if (!defined $before_hash || $before_hash !~ /[\w\.\/]/) {
    my $outhandle = $self->{'outhandle'};
    print $outhandle "HTMLPlug: ERROR - badly formatted tag ignored ($link)\n"
        if $self->{'verbosity'};
    return ($link, "", 0);
    }

    if ($before_hash =~ s@^((?:http|ftp|file)://)@@i) {
    my $type = $1;

    if ($link =~ /^(http|ftp):/i) {
        # Turn url (using /) into file name (possibly using \ on windows)
        my @http_dir_split = split('/', $before_hash);
        $before_hash = &util::filename_cat(@http_dir_split);
    }

    $before_hash = $self->eval_dir_dots($before_hash);
                     
    my $linkfilename = &util::filename_cat ($base_dir, $before_hash);
    
    my $rl = 0;
    $rl = 1 if (-e $linkfilename);

    # make sure there's a slash on the end if it's a directory
    if ($before_hash !~ /\/$/) {
        $before_hash .= "/" if (-d $linkfilename);
    }

    return ($type . $before_hash, $hash_part, $rl);
    
    } elsif ($link !~ /^(mailto|news|gopher|nntp|telnet|javascript):/i && $link !~ /^\//) {
    if ($before_hash =~ s@^/@@ || $before_hash =~ /\\/) {

        # the first directory will be the domain name if file_is_url
        # to generate archives, otherwise we'll assume all files are
        # from the same site and base_dir is the root

        if ($self->{'file_is_url'}) {
        my @dirs = split /[\/\\]/, $file;
        my $domname = shift (@dirs);
        $before_hash = &util::filename_cat($domname, $before_hash);
        $before_hash =~ s@\\@/@g; # for windows
        }
        else
        {
        # see if link shares directory with source document
        # => turn into relative link if this is so!
        
        if ($ENV{'GSDLOS'} =~ /^windows/i) {
            # too difficult doing a pattern match with embedded '\'s...
            my $win_before_hash=$before_hash;
            $win_before_hash =~ s@(\\)+@/@g;
            # $base_dir is already similarly "converted" on windows.
            if ($win_before_hash =~ s@^$base_dir/@@o) {
              # if this is true, we removed a prefix
              $before_hash=$win_before_hash;
            }
        }
        else {
            # before_hash has lost leading slash by this point,
            # -> add back in prior to substitution with $base_dir
            $before_hash = "/$before_hash"; 

            $before_hash = &util::filename_cat("",$before_hash);
            $before_hash =~ s@^$base_dir/@@;
        }
        }
    } else {
        # Turn relative file path into full path
        my $dirname = &File::Basename::dirname($file);
        $before_hash = &util::filename_cat($dirname, $before_hash);
        $before_hash = $self->eval_dir_dots($before_hash);
    }

    my $linkfilename = &util::filename_cat ($base_dir, $before_hash); 
    # make sure there's a slash on the end if it's a directory
    if ($before_hash !~ /\/$/) {
        $before_hash .= "/" if (-d $linkfilename);
    }
    return ("http://" . $before_hash, $hash_part, 1);
    } else {
    # mailto, news, nntp, telnet, javascript or gopher link
    return ($before_hash, "", 0);
    }
}

sub extract_first_NNNN_characters {
    my $self = shift (@_);
    my ($textref, $doc_obj, $thissection) = @_;
    
    foreach my $size (split /,/, $self->{'first'}) {
    my $tmptext =  $$textref;
    # skip to the body
    $tmptext =~ s/.*<body[^>]*>//i;
    # remove javascript
    $tmptext =~ s@<script.*?</script>@ @sig;
    $tmptext =~ s/<[^>]*>/ /g;
    $tmptext =~ s/&nbsp;/ /g;
    $tmptext =~ s/^\s+//;
    $tmptext =~ s/\s+$//;
    $tmptext =~ s/\s+/ /gs;
    $tmptext = &unicode::substr ($tmptext, 0, $size);
    $tmptext =~ s/\s\S*$/&#8230;/; # adds an ellipse (...)
    $doc_obj->add_utf8_metadata ($thissection, "First$size", $tmptext);
    }
}


sub extract_metadata {
    my $self = shift (@_);
    my ($textref, $metadata, $doc_obj, $section) = @_;
    my $outhandle = $self->{'outhandle'};
    # if we don't want metadata, we may as well not be here ...
    return if (!defined $self->{'metadata_fields'});

    # metadata fields to extract/save. 'key' is the (lowercase) name of the
    # html meta, 'value' is the metadata name for greenstone to use
    my %find_fields = ();

    my %creator_fields = (); # short-cut for lookups


    foreach my $field (split /,/, $self->{'metadata_fields'}) {
    # support tag<tagname>
    if ($field =~ /^(.*?)<(.*?)>$/) {
        # "$2" is the user's preferred gs metadata name
        $find_fields{lc($1)}=$2; # lc = lowercase
    } else { # no <tagname> for mapping
        # "$field" is the user's preferred gs metadata name
        $find_fields{lc($field)}=$field; # lc = lowercase
    }
    }

    if (defined $self->{'hunt_creator_metadata'} &&
    $self->{'hunt_creator_metadata'} == 1 ) {
    my @extra_fields =
        (
         'author',
         'author.email',
         'creator',
         'dc.creator',
         'dc.creator.corporatename',
         );

    # add the creator_metadata fields to search for
    foreach my $field (@extra_fields) {
        $creator_fields{$field}=0; # add to lookup hash
    }
    }


    # find the header in the html file, which has the meta tags
    $$textref =~ m@<head>(.*?)</head>@si;

    my $html_header=$1;
    # go through every <meta... tag defined in the html and see if it is
    # one of the tags we want to match.

    # special case for title - we want to remember if its been found
    my $found_title = 0;
    # this assumes that ">" won't appear. (I don't think it's allowed to...)
    $html_header =~ /^/; # match the start of the string, for \G assertion
 
    while ($html_header =~ m/\G.*?<meta(.*?)>/sig) {
    my $metatag=$1;
    my ($tag, $value);

    # find the tag name
    $metatag =~ /(?:name|http-equiv)\s*=\s*([\"\'])?(.*?)\1/is;
    $tag=$2;
    # in case they're not using " or ', but they should...
    if (! $tag) {
        $metatag =~ /(?:name|http-equiv)\s*=\s*([^\s\>]+)/is;
        $tag=$1;
    }

    if (!defined $tag) {
        print $outhandle "HTMLPlug: can't find NAME in \"$metatag\"\n";
        next;
    }

    # don't need to assign this field if it was passed in from a previous 
    # (recursive) plugin
    if (defined $metadata->{$tag}) {next}

    # find the tag content
    $metatag =~ /content\s*=\s*([\"\'])?(.*?)\1/is;
    $value=$2;
    if (! $value) {
        $metatag =~ /(?:name|http-equiv)\s*=\s*([^\s\>]+)/is;
        $value=$1;
    }
    if (!defined $value) {
        print $outhandle "HTMLPlug: can't find VALUE in \"$metatag\"\n";
        next;
    }

    # clean up and add
    $value =~ s/\s+/ /gs;
    chomp($value); # remove trailing \n, if any
    if (exists $creator_fields{lc($tag)}) {
        # map this value onto greenstone's "Creator" metadata
        $tag='Creator';
    } elsif (!exists $find_fields{lc($tag)}) {
        next; # don't want this tag
    } else {
        # get the user's preferred capitalisation
        $tag = $find_fields{lc($tag)};
    }
    if (lc($tag) eq "title") {
        $found_title = 1;
    }
    print $outhandle " extracted \"$tag\" metadata \"$value\"\n" 
        if ($self->{'verbosity'} > 2);
    $doc_obj->add_utf8_metadata($section, $tag, $value);

    }
    
    # TITLE: extract the document title
    if (exists $find_fields{'title'} && !$found_title) {
    # we want a title, and didn't find one in the meta tags
    # see if there's a <title> tag
    my $title;
    my $from = ""; # for debugging output only
    if ($html_header =~ /<title[^>]*>([^<]+)<\/title[^>]*>/is) {
        $title = $1;
        $from = "<title> tags";
    }

    if (!defined $title) {
        $from = "first 100 chars";
        # if no title use first 100 or so characters
        $title = $$textref;
        $title =~ s/^\xFE\xFF//; # Remove unicode byte order mark       
        $title =~ s/^.*?<body>//si;
        # ignore javascript!
        $title =~ s@<script.*?</script>@ @sig;
        $title =~ s/<\/([^>]+)><\1>//g; # (eg) </b><b> - no space
        $title =~ s/<[^>]*>/ /g; # remove all HTML tags
        $title = substr ($title, 0, 100);
        $title =~ s/\s\S*$/.../;
    }
    $title =~ s/<[^>]*>/ /g; # remove html tags
    $title =~ s/&nbsp;/ /g;
    $title =~ s/(?:&nbsp;|\xc2\xa0)/ /g; # utf-8 for nbsp...
    $title =~ s/\s+/ /gs; # collapse multiple spaces
    $title =~ s/^\s*//;   # remove leading spaces
    $title =~ s/\s*$//;   # remove trailing spaces

    $title =~ s/^$self->{'title_sub'}// if ($self->{'title_sub'});
    $title =~ s/^\s+//s; # in case title_sub introduced any...
    $doc_obj->add_utf8_metadata ($section, 'Title', $title);
    print $outhandle " extracted Title metadata \"$title\" from $from\n" 
        if ($self->{'verbosity'} > 2);
    } 
    
    # add FileFormat metadata
    $doc_obj->add_metadata($section,"FileFormat", "HTML");

    # Special, for metadata names such as tagH1 - extracts
    # the text between the first <H1> and </H1> tags into "H1" metadata.

    foreach my $field (keys %find_fields) {
    if ($field !~ /^tag([a-z0-9]+)$/i) {next}
    my $tag = $1;
    if ($$textref =~ m@<$tag[^>]*>(.*?)</$tag[^>]*>@g) {
        my $content = $1;
        $content =~ s/&nbsp;/ /g;
        $content =~ s/<[^>]*>/ /g;
        $content =~ s/^\s+//;
        $content =~ s/\s+$//;
        $content =~ s/\s+/ /gs;
        if ($content) {
        $tag=$find_fields{"tag$tag"}; # get the user's capitalisation
        $tag =~ s/^tag//i;
        $doc_obj->add_utf8_metadata ($section, $tag, $content);
        print $outhandle " extracted \"$tag\" metadata \"$content\"\n" 
            if ($self->{'verbosity'} > 2);
        }
    }
    }    
}


# evaluate any "../" to next directory up
# evaluate any "./" as here
sub eval_dir_dots {
    my $self = shift (@_);
    my ($filename) = @_;
    my $dirsep_os = &util::get_os_dirsep();
    my @dirsep = split(/$dirsep_os/,$filename);

    my @eval_dirs = ();
    foreach my $d (@dirsep) {
    if ($d eq "..") {
        pop(@eval_dirs);
        
    } elsif ($d eq ".") {
        # do nothing!

    } else {
        push(@eval_dirs,$d);
    }
    }

    # Need to fiddle with number of elements in @eval_dirs if the
    # first one is the empty string.  This is because of a
    # modification to util::filename_cat that supresses the addition
    # of a leading '/' character (or \ if windows) (intended to help
    # filename cat with relative paths) if the first entry in the
    # array is the empty string.  Making the array start with *two*
    # empty strings is a way to defeat this "smart" option.
    #
    if (scalar(@eval_dirs) > 0) {
    if ($eval_dirs[0] eq ""){
        unshift(@eval_dirs,"");
    }
    }
    return &util::filename_cat(@eval_dirs);
}

sub replace_usemap_links {
    my $self = shift (@_);
    my ($front, $link, $back) = @_;

    $link =~ s/^\.\///;
    return $front . $link . $back;
}

sub inc_filecount {
    my $self = shift (@_);

    if ($self->{'file_num'} == 1000) {
    $self->{'dir_num'} ++;
    $self->{'file_num'} = 0;
    } else {
    $self->{'file_num'} ++;
    }
}


# Extend the BasPlug read_file so that strings like &eacute; are
# converted to UTF8 internally.  
#
# We don't convert &lt; or &gt; or &amp; or &quot; in case
# they interfere with the GML files

sub read_file {
    my ($self, $filename, $encoding, $language, $textref) = @_;

    &BasPlug::read_file($self, $filename, $encoding, $language, $textref);

    # Convert entities to their UTF8 equivalents
    $$textref =~ s/&(lt|gt|amp|quot|nbsp);/&z$1;/go;
    $$textref =~ s/&([^;]+);/&ghtml::getcharequiv($1,1)/gseo;
    $$textref =~ s/&z(lt|gt|amp|quot|nbsp);/&$1;/go;
}

1;