Changeset 2715 for trunk/gsdl

Timestamp:

2001-08-23T18:00:26+12:00 (23 years ago)

Author:

jrm21

Message:

we now escape _ characters to protect against macro expansion, unless they
occur within html tags -- eg <img src="foo_bar.png">.

File:

• trunk/gsdl/bin/script/pdftohtml.pl (modified) (4 diffs)

trunk/gsdl/bin/script/pdftohtml.pl

@@ -4 +4 @@
 ###########################################################################
 #
-# pdftohtml.pl -- convert documents to HTML or TEXT format
+# pdftohtml.pl -- convert PDF documents to HTML format
 #
 # A component of the Greenstone digital library software
@@ -10 +10 @@
 # University of Waikato, New Zealand.
 #
-# Copyright (C) 1999 New Zealand Digital Library Project
+# Copyright (C) 2001 New Zealand Digital Library Project
 #
 # This program is free software; you can redistribute it and/or modify
@@ -119 +119 @@
     $cmd = "pdftohtml" if ($ENV{'GSDLOS'} =~ /^windows$/);
 
-    if ($timeout) {$cmd = "ulimit -t $timeout; $cmd";}
     $cmd .= " -noframes \"$input_filename\" \"$output_filestem.html\"";
     $cmd .= " > \"$output_filestem.out\"";
@@ -152 +151 @@
     $line =~ s#</b><b>##g;
     $line =~ s#</i><i>##g;
-    $line =~ s#\\#\\\\#g;
+    $line =~ s#\\#\\\\#g; # until macro language parsing is fixed...
+# escape underscores, but not if they're inside tags (eg img/href names)
+    my $inatag = 0; # allow multi-line tags
+    if ($line =~ /_/) {
+        my @parts=split('_',$line);
+        my $lastpart=pop @parts;
+        foreach my $part (@parts) {
+        if ($part =~ /<[^>]*$/) { # if we're starting a tag...
+            $inatag=1;
+        } elsif ($part =~ />[^<]*$/) { # closing a tag
+            $inatag=0;
+        }
+        if ($inatag) {
+            $part.='_';
+        } else {
+            $part.="&#95;";
+        }
+        }
+        $line=join('',@parts,$lastpart);
+    }
+
     print OUTFILE $line;
     }