X-Git-Url: https://sipb.mit.edu/gitweb.cgi/ikiwiki.git/blobdiff_plain/44fde6cbff2aaa8d647084721320111e8807ea74..ac1226031eb5c5f9b8fe528420483f4d8828c73d:/IkiWiki/Plugin/search.pm

diff --git a/IkiWiki/Plugin/search.pm b/IkiWiki/Plugin/search.pm
index a28d58a1d..eedfa6924 100644
--- a/IkiWiki/Plugin/search.pm
+++ b/IkiWiki/Plugin/search.pm
@@ -9,10 +9,7 @@ use IkiWiki 2.00;
 sub import { #{{{
 	hook(type => "checkconfig", id => "search", call => \&checkconfig);
 	hook(type => "pagetemplate", id => "search", call => \&pagetemplate);
-	# run last so other needsbuild hooks can modify the list
-	hook(type => "needsbuild", id => "search", call => \&needsbuild,
-		last => 1);
-	hook(type => "filter", id => "search", call => \&filter);
+	hook(type => "postscan", id => "search", call => \&index);
 	hook(type => "delete", id => "search", call => \&delete);
 	hook(type => "cgi", id => "search", call => \&cgi);
 } # }}}
@@ -23,19 +20,10 @@ sub checkconfig () { #{{{
 			error(sprintf(gettext("Must specify %s when using the search plugin"), $required));
 		}
 	}
-
+	
 	if (! exists $config{omega_cgi}) {
 		$config{omega_cgi}="/usr/lib/cgi-bin/omega/omega";
 	}
-	
-	if (! -e $config{wikistatedir}."/xapian" || $config{rebuild}) {
-		writefile("omega.conf", $config{wikistatedir}."/xapian",
-			"database_dir .\n".
-			"template_dir ./templates\n");
-		writefile("query", $config{wikistatedir}."/xapian/templates",
-			IkiWiki::misctemplate(gettext("search"),
-				readfile(IkiWiki::template_file("searchquery.tmpl"))));
-	}
 } #}}}
 
 my $form;
@@ -56,52 +44,99 @@ sub pagetemplate (@) { #{{{
 	}
 } #}}}
 
-my %toindex;
-sub needsbuild ($) { #{{{
-	%toindex = map { $_ => 1 } @{shift()};
-} #}}}
-
-sub filter (@) { #{{{
+my $scrubber;
+my $stemmer;
+sub index (@) { #{{{
 	my %params=@_;
+
+	setupfiles();
+
+	# A unique pageterm is used to identify the document for a page.
+	my $pageterm=pageterm($params{page});
+	return $params{content} unless defined $pageterm;
 	
-	if ($params{page} eq $params{destpage} && $toindex{$params{page}}) {
-		# index page
-		my $db=xapiandb();
-		my $doc=Search::Xapian::Document->new();
-		my $title=$params{page};
-		if (exists $pagestate{$params{page}}{meta} &&
-		    exists $pagestate{$params{page}}{meta}{title}) {
-			$title=$pagestate{$params{page}}{meta}{title};
-		}
+	my $db=xapiandb();
+	my $doc=Search::Xapian::Document->new();
+	my $caption=IkiWiki::pagetitle($params{page});
+	my $title;
+	if (exists $pagestate{$params{page}}{meta} &&
+		exists $pagestate{$params{page}}{meta}{title}) {
+		$title=$pagestate{$params{page}}{meta}{title};
+	}
+	else {
+		$title=$caption;
+	}
 
-		# data used by omega
-		$doc->set_data(
-			"url=".urlto($params{page}, "")."\n".
-			"sample=\n". # TODO
-			"caption=$title\n".
-			"modtime=$IkiWiki::pagemtime{$params{page}}\n".
-			"size=".length($params{content})."\n"
-		);
-
-		my $tg = Search::Xapian::TermGenerator->new();
-		$tg->set_stemmer(new Search::Xapian::Stem("english"));
-		$tg->set_document($doc);
-		$tg->index_text($params{page}, 2);
-		$tg->index_text($title, 2);
-		$tg->index_text($params{content}); # TODO html strip; preprocessor too
-
-		my $pageterm=pageterm($params{page});
-		$doc->add_term($pageterm);
-		$db->replace_document_by_term($pageterm, $doc);
+	# Remove html from text to be indexed.
+	if (! defined $scrubber) {
+		eval q{use HTML::Scrubber};
+		if (! $@) {
+			$scrubber=HTML::Scrubber->new(allow => []);
+		}
+	}
+	my $toindex = defined $scrubber ? $scrubber->scrub($params{content}) : $params{content};
+	
+	# Take 512 characters for a sample, then extend it out
+	# if it stopped in the middle of a word.
+	my $size=512;
+	my ($sample)=substr($toindex, 0, $size);
+	if (length($sample) == $size) {
+		my $max=length($toindex);
+		my $next;
+		while ($size < $max &&
+		       ($next=substr($toindex, $size++, 1)) !~ /\s/) {
+			$sample.=$next;
+		}
+	}
+	$sample=~s/\n/ /g;
+	
+	# data used by omega
+	# Decode html entities in it, since omega re-encodes them.
+	eval q{use HTML::Entities};
+	$doc->set_data(
+		"url=".urlto($params{page}, "")."\n".
+		"sample=".decode_entities($sample)."\n".
+		"caption=".decode_entities($caption)."\n".
+		"modtime=$IkiWiki::pagemtime{$params{page}}\n".
+		"size=".length($params{content})."\n"
+	);
+
+	# Index document and add terms for other metadata.
+	my $tg = Search::Xapian::TermGenerator->new();
+	if (! $stemmer) {
+		my $langcode=$ENV{LANG} || "en";
+		$langcode=~s/_.*//;
+
+		# This whitelist is here to work around a xapian bug (#486138)
+		my @whitelist=qw{da de en es fi fr hu it no pt ru ro sv tr};
+
+		if (grep { $_ eq $langcode } @whitelist) {
+			$stemmer=Search::Xapian::Stem->new($langcode);
+		}
+		else {
+			$stemmer=Search::Xapian::Stem->new("english");
+		}
+	}
+	$tg->set_stemmer($stemmer);
+	$tg->set_document($doc);
+	$tg->index_text($params{page}, 2);
+	$tg->index_text($caption, 2);
+	$tg->index_text($title, 2) if $title ne $caption;
+	$tg->index_text($toindex);
+	$tg->index_text(lc($title), 1, "S"); # for title:foo
+	foreach my $link (@{$links{$params{page}}}) {
+		$tg->index_text(lc($link), 1, "XLINK"); # for link:bar
 	}
 
-	return $params{content};
+	$doc->add_term($pageterm);
+	$db->replace_document_by_term($pageterm, $doc);
 } #}}}
 
 sub delete (@) { #{{{
 	my $db=xapiandb();
 	foreach my $page (@_) {
-		$db->delete_document_by_term(pageterm($page));
+		my $pageterm=pageterm(pagename($page));
+		$db->delete_document_by_term($pageterm) if defined $pageterm;
 	}
 } #}}}
 
@@ -113,6 +148,9 @@ sub cgi ($) { #{{{
 		chdir("$config{wikistatedir}/xapian") || error("chdir: $!");
 		$ENV{OMEGA_CONFIG_FILE}="./omega.conf";
 		$ENV{CGIURL}=$config{cgiurl},
+		IkiWiki::loadindex();
+		$ENV{HELPLINK}=htmllink("", "", "ikiwiki/searching",
+			noimageinline => 1, linktext => "Help");
 		exec($config{omega_cgi}) || error("$config{omega_cgi} failed: $!");
 	}
 } #}}}
@@ -120,9 +158,22 @@ sub cgi ($) { #{{{
 sub pageterm ($) { #{{{
 	my $page=shift;
 
-	# TODO: check if > 255 char page names overflow term
-	# length; use sha1 if so?
-	return "P".$page;
+	# 240 is the number used by omindex to decide when to hash an
+	# overlong term. This does not use a compatible hash method though.
+	if (length $page > 240) {
+		eval q{use Digest::SHA1};
+		if ($@) {
+			debug("search: ".sprintf(gettext("need Digest::SHA1 to index %s"), $page)) if $@;
+			return undef;
+		}
+
+		# Note no colon, therefore it's guaranteed to not overlap
+		# with a page with the same name as the hash..
+		return "U".lc(Digest::SHA1::sha1_hex($page));
+	}
+	else {
+		return "U:".$page;
+	}
 } #}}}
 
 my $db;
@@ -139,4 +190,19 @@ sub xapiandb () { #{{{
 	return $db;
 } #}}}
 
+{
+my $setup=0;
+sub setupfiles () { #{{{
+	if (! $setup and (! -e $config{wikistatedir}."/xapian" || $config{rebuild})) {
+		writefile("omega.conf", $config{wikistatedir}."/xapian",
+			"database_dir .\n".
+			"template_dir ./templates\n");
+		writefile("query", $config{wikistatedir}."/xapian/templates",
+			IkiWiki::misctemplate(gettext("search"),
+				readfile(IkiWiki::template_file("searchquery.tmpl"))));
+		$setup=1;
+	}
+} #}}}
+}
+
 1