Fulltext search: Greatly improve by using a more fuzzy approach

[deb/packages.git] / lib / Packages / Search.pm

diff --git a/lib/Packages/Search.pm b/lib/Packages/Search.pm
index fdd3a3a4beeb5f730b76a9fbd980f509b42f6b83..66b09441ba1ea72d84de6578e200a488971bf099 100644 (file)
--- a/lib/Packages/Search.pm
+++ b/lib/Packages/Search.pm
@@ -196,55 +196,6 @@ sub do_names_search {
        &$read_entry( $packages, $pkg, $results, $non_results, $opts );
     }
 }
-sub do_fulltext_search {
-    my ($keywords, $file, $did2pkg, $packages, $read_entry, $opts,
-       $results, $non_results) = @_;
-
-# NOTE: this needs to correspond with parse-packages!
-    my @tmp;
-    foreach my $keyword (@$keywords) {
-       $keyword =~ tr [A-Z] [a-z];
-       if ($opts->{exact}) {
-           $keyword = " $keyword ";
-       }
-       $keyword =~ s/[(),.-]+//og;
-       $keyword =~ s;[^a-z0-9_/+]+; ;og;
-       push @tmp, $keyword;
-    }
-    my $first_keyword = shift @tmp;
-    @$keywords = @tmp;
-
-    my $numres = 0;
-    my %tmp_results;
-    # fgrep is seriously faster than using perl
-    open DESC, '-|', 'fgrep', '-n', '--', $first_keyword, $file
-       or die "couldn't open $file: $!";
-  LINE:
-    while (<DESC>) {
-       foreach my $k (@$keywords) {
-           next LINE unless /\Q$k\E/;
-       }
-       /^(\d+)/;
-       my $nr = $1;
-       debug( "Matched line $_", 2) if DEBUG;
-       my $result = $did2pkg->{$nr};
-       foreach (split /\000/o, $result) {
-           my @data = split /\s/, $_, 3;
-#          debug ("Considering $data[0], arch = $data[2]", 3) if DEBUG;
-#          next unless $data[2] eq 'all' || $opts->{h_archs}{$data[2]};
-#          debug ("Ok", 3) if DEBUG;
-           $numres++ unless $tmp_results{$data[0]}++;
-       }
-       last if $numres > 100;
-    }
-    close DESC;
-    $too_many_hits++ if $numres > 100;
-
-    my @results;
-    foreach my $pkg (keys %tmp_results) {
-       &$read_entry( $packages, $pkg, $results, $non_results, $opts );
-    }
- }
 
 sub do_xapian_search {
     my ($keywords, $db, $did2pkg, $packages, $read_entry, $opts,
@@ -253,41 +204,38 @@ sub do_xapian_search {
 # NOTE: this needs to correspond with parse-packages!
     my @tmp;
     foreach my $keyword (@$keywords) {
-       $keyword =~ tr [A-Z] [a-z];
-       if ($opts->{exact}) {
-           $keyword = " $keyword ";
-       }
-       $keyword =~ s/[(),.-]+//og;
-       $keyword =~ s;[^a-z0-9_/+]+; ;og;
+       $keyword =~ s;[^\w/+]+; ;og;
        push @tmp, $keyword;
     }
     my $stemmer = Lingua::Stem->new();
-    $keywords = $stemmer->stem( @tmp );
+    my $stemmed_keywords = $stemmer->stem( @tmp );
 
     my $db = Search::Xapian::Database->new( $db );
-    my $enq = $db->enquire( OP_AND, @$keywords );
+    my $enq = $db->enquire( OP_OR, @$keywords, @$stemmed_keywords );
     debug( "Xapian Query was: ".$enq->get_query()->get_description(), 1) if DEBUG;
-    my @matches = $enq->matches(0, 100);
+    my @matches = $enq->matches(0, 999);
 
-    my $numres = 0;
-    my %tmp_results;
+    my (@order, %tmp_results);
     foreach my $match ( @matches ) {
        my $id = $match->get_docid();
        my $result = $did2pkg->{$id};
 
        foreach (split /\000/o, $result) {
            my @data = split /\s/, $_, 3;
-#          debug ("Considering $data[0], arch = $data[2]", 3) if DEBUG;
+           debug ("Considering $data[0], arch = $data[2], relevance=".$match->get_percent(), 3) if DEBUG;
 #          next unless $data[2] eq 'all' || $opts->{h_archs}{$data[2]};
 #          debug ("Ok", 3) if DEBUG;
-           $numres++ unless $tmp_results{$data[0]}++;
+           unless ($tmp_results{$data[0]}++) {
+               push @order, $data[0];
+           }
        }
-       last if $numres > 100;
+       last if @order > 100;
     }
     undef $db;
-    $too_many_hits++ if $numres > 100;
+    $too_many_hits++ if @order > 100;
 
-    foreach my $pkg (keys %tmp_results) {
+    debug ("ORDER: @order", 2) if DEBUG;
+    foreach my $pkg (@order) {
        &$read_entry( $packages, $pkg, $results, $non_results, $opts );
     }
 }