--- trunk/spider/swishspider	2003/11/19 12:07:07	45
+++ trunk/spider/swishspider	2004/01/17 23:57:55	46
@@ -5,6 +5,10 @@
 use HTTP::Status;
 use HTML::LinkExtor;
 
+my $basedir = $0;
+$basedir =~ s,/[^/]+$,/,;
+require "$basedir/filter.pm";
+
 if (scalar(@ARGV) != 2) {
     print STDERR "Usage: SwishSpider localpath url\n";
     exit(1);
@@ -65,42 +69,7 @@
     open( CONTENTS, ">$localpath.contents" ) || die( "Could not open contents file $localpath.contents\n" );
     # fixup just HTML files
     if ($response->header("content-type") =~ "text/html") {
-	# if you don't want content to be indexed, include it in
-	# <noindex> foobar </noindex> tags or surround it with comments
-	# <!-- noindex --> foobar <!-- /noindex -->
-	# <!-- noindex --> foobar <!-- index --> (also supported by swish)
-	$contents =~ s,<noindex>.+?</noindex>,,isg;
-	$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*/noindex\s*-->,,isg;
-	$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*index\s*-->,,isg;
-	# this will remove all script from indexing content
-	$contents =~ s,<script>.+?</script>,,isg;
-	# remap Windows charset to ISO-8859-2
-	$contents =~ tr/氿炶鎶袔绕/桂捐姗挟绕/;	# 1250 -> iso8859-2
-	# this will fix badly formatted html in form:
-	# <head><title>some text</title
-	# ></head>
-	# which will confuse indexer (or libxml2?)
-	$contents =~ s/[\n\r]^(>)/$1\n/msg;
-	# remove comments between <html> and <head> texi2html inserts them
-	# there and swish can't find document title then (libxml or swish bug?)
-	while ($contents =~ s/(<html>.*)<!--.*?-->(.*<head>)/$1$2/msi) { };
-
-	# remote TPJ left column
-	if ($contents =~ s,<!-- BEGIN LEFT SIDE BAR CELL -->.+?<!-- END LEFT SIDE BAR CELL -->,,isg) {
-		my $title;
-		# extract title and add to title
-		if ($contents =~ m,<!-- the article goes here -->\s*<h2[^>]*>(.+?)</h2>,si) {
-			$title = $1;
-		} elsif ($contents =~ m,<h1[^>]*>(.+?)</h1>,is) {
-			$title = $1;
-		} elsif ($contents =~ m,<h2[^>]*>(.+?)</h2>,is) {
-			$title = $1;
-		} else {
-			$title = "no detail title";
-		}
-		$contents =~ s,(<title>)([^<]+)(</title>),$1$2: $title$3,gsi if ($title);
-
-	}
+    	$contents = filter($contents);
     }
     print CONTENTS $contents;
     close( CONTENTS );
