Changeset 3411 for trunk/gsdl/perllib/plugins/PDFPlug.pm

Timestamp:

2002-08-29T16:37:54+12:00 (22 years ago)

Author:

jrm21

Message:

Now takes a "-use_sections" option to make a section per page.

File:

• trunk/gsdl/perllib/plugins/PDFPlug.pm (modified) (2 diffs)

trunk/gsdl/perllib/plugins/PDFPlug.pm

@@ -39 +39 @@
     # if your document title actually starts with "1 " - is there a better way?
 
-    my $self = new ConvertToPlug ($class, @_, "-title_sub", '^(Page\s+\d+)?(\s*1\s+)?');
+    my @args=();
+    my $use_sections=0;
+    while ($#_ > -1) {
+    my $arg=shift(@_);
+    if ($arg =~ /^\-use_sections/) {
+        $use_sections=1;
+        # in case someone put "-use_sections 1"
+        $arg=shift(@_);
+        if (defined($arg) && $arg ne "1") {
+        push(@args, $arg);
+        }
+        push(@args, "-description_tags"); # so HTML uses sections...
+        push(@args, @_); # any other args to PDFPlug...
+        @_=();
+        last;
+    } else {
+        push(@args, $arg);
+    }
+    }
+    my $self = new ConvertToPlug ($class, @args, "-title_sub", '^(Page\s+\d+)?(\s*1\s+)?');
+    if ($use_sections) {
+    $self->{'use_sections'}=1;
+    }
     
     return bless $self, $class;
+}
+
+
+sub print_usage {
+    print STDERR "\n  usage: plugin PDFPlug [options]\n\n";
+    print STDERR "  options:\n";
+    print STDERR "   -convert_to (html|text) plugin converts to TEXT or HTML\n";
+    print STDERR "                           (default html)\n";
+    print STDERR "   -use_sections     create a separate section for each page of the PDF file.\n\n";
 }
 
@@ -62 +93 @@
     my $self = shift (@_);
 
+    if ($self->{'use_sections'}
+    && $self->{'converted_to'} eq "HTML") {
+
+    print STDERR "PDFPlug: Calculating sections...\n";
+    my $textref=$_[0];
+
+    # This might be specific to the version of pdftohtml in <=  gsdl-2.38
+    my @sections = split('<p><a name=\d+>', $$textref);
+
+    shift @sections; # don't need HTML header, etc
+    # handle first section specially for title? Or all use first 100...
+    
+    my $title = $sections[0];
+    $title =~ s/^\d+>//; # specific for pdftohtml...
+    $title =~ s/<\/([^>]+)><\1>//g; # (eg) </b><b> - no space
+    $title =~ s/<[^>]*>/ /g;
+    $title =~ s/(?:&nbsp;|\xc2\xa0)/ /g; # utf-8 for nbsp...
+    $title =~ s/^\s+//s;
+    $title =~ s/\s+$//;
+    $title =~ s/\s+/ /gs;
+    $title =~ s/^$self->{'title_sub'}// if ($self->{'title_sub'});
+    $title =~ s/^\s+//s; # in case title_sub introduced any...
+    $title = substr ($title, 0, 100);
+    $title =~ s/\s\S*$/.../;
+
+    my $top_section = "<!--<Section>\n<Metadata name=\"Title\">$title</Metadata>\n-->\n <!--</Section>-->\n";
+
+    # add metadata per section...
+    foreach my $section (@sections) {
+        $section =~ m@^<b>Page (\d+)</b>@;
+        $title = $1; # Greenstone does magic if sections are titled digits
+        if (! defined($title) ) {
+        print STDERR "no title: $section\n";
+        }
+        my $newsection = "<!-- from PDFPlug -->\n<!-- <Section>\n";
+        $newsection .= "<Metadata name=\"Title\">" . $title
+        . "</Metadata>\n--><p>\n";
+        $newsection .= $section;
+        $newsection .= "<!--</Section>-->\n";
+        $section = $newsection;
+    }
+
+    $$textref=join('', ($top_section, @sections));
+    }
+
     my $outhandle = $self->{'outhandle'};
     print $outhandle "PDFPlug: passing $_[3] on to $self->{'converted_to'}Plug\n"