Remove obsolete descriptions.txt previously used for fulltext search

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

diff --git a/lib/Packages/Search.pm b/lib/Packages/Search.pm
index 0bc157d735334fd3f516a5bf4687e959546ae11a..671e3401c1f75fe7eadef8f82182e7a1cb2660ae 100644 (file)
--- a/lib/Packages/Search.pm
+++ b/lib/Packages/Search.pm
@@ -46,6 +46,8 @@ use warnings;
 use POSIX;
 use HTML::Entities;
 use DB_File;
+use Lingua::Stem v0.82;
+use Search::Xapian qw(:ops);
 
 use Deb::Versions;
 use Packages::CGI;
@@ -55,7 +57,7 @@ our @ISA = qw( Exporter );
 
 our @EXPORT_OK = qw( read_entry read_entry_all read_entry_simple
                     read_src_entry read_src_entry_all find_binaries
-                    do_names_search do_fulltext_search
+                    do_names_search do_fulltext_search do_xapian_search
                     );
 our %EXPORT_TAGS = ( all => [ @EXPORT_OK ] );
 
@@ -100,6 +102,8 @@ sub read_entry {
 
 #FIXME: make configurable
 my %fallback_suites = (
+                      'oldstable-backports' => 'oldstable',
+                      'oldstable-volatile' => 'oldstable',
                       'stable-backports' => 'stable',
                       'stable-volatile' => 'stable',
                       experimental => 'unstable' );
@@ -154,23 +158,29 @@ sub read_src_entry {
     read_src_entry_all( $hash, $key, $results, \@non_results, $opts );
 }
 sub do_names_search {
-    my ($keyword, $packages, $postfixes, $read_entry, $opts,
+    my ($keywords, $packages, $postfixes, $read_entry, $opts,
        $results, $non_results) = @_;
 
-    $keyword = lc $keyword;
+    my $first_keyword = lc shift @$keywords;
+    @$keywords = map { lc $_ } @$keywords;
         
-    my ($key, $prefixes) = ($keyword, '');
+    my ($key, $prefixes) = ($first_keyword, '');
     my %pkgs;
     $postfixes->seq( $key, $prefixes, R_CURSOR );
-    while (index($key, $keyword) >= 0) {
+    while (index($key, $first_keyword) >= 0) {
        if ($prefixes =~ /^\001(\d+)/o) {
            debug( "$key has too many hits", 2 ) if DEBUG;
            $too_many_hits += $1;
        } else {
+         PREFIX:
            foreach (split /\000/o, $prefixes) {
                $_ = '' if $_ eq '^';
-               debug( "add word $_$key", 2) if DEBUG;
-               $pkgs{$_.$key}++;
+               my $word = "$_$key";
+               foreach my $k (@$keywords) {
+                   next PREFIX unless $word =~ /\Q$k\E/;
+               }
+               debug( "add word $word", 2) if DEBUG;
+               $pkgs{$word}++;
            }
        }
        last if $postfixes->seq( $key, $prefixes, R_NEXT ) != 0;
@@ -180,34 +190,42 @@ sub do_names_search {
     my $no_results = keys %pkgs;
     if ($too_many_hits || ($no_results >= 100)) {
        $too_many_hits += $no_results;
-       %pkgs = ( $keyword => 1 );
+       %pkgs = ( $first_keyword => 1 ) unless @$keywords;
     }
     foreach my $pkg (sort keys %pkgs) {
        &$read_entry( $packages, $pkg, $results, $non_results, $opts );
     }
 }
-sub do_fulltext_search {
-    my ($keyword, $file, $did2pkg, $packages, $read_entry, $opts,
+
+sub do_xapian_search {
+    my ($keywords, $db, $did2pkg, $packages, $read_entry, $opts,
        $results, $non_results) = @_;
 
 # NOTE: this needs to correspond with parse-packages!
-    $keyword =~ tr [A-Z] [a-z];
-    if ($opts->{exact}) {
-       $keyword = " $keyword ";
+    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;
     }
-    $keyword =~ s/[(),.-]+//og;
-    $keyword =~ s#[^a-z0-9_/+]+# #og;
+    my $stemmer = Lingua::Stem->new();
+    $keywords = $stemmer->stem( @tmp );
+
+    my $db = Search::Xapian::Database->new( $db );
+    my $enq = $db->enquire( OP_AND, @$keywords );
+    debug( "Xapian Query was: ".$enq->get_query()->get_description(), 1) if DEBUG;
+    my @matches = $enq->matches(0, 100);
 
     my $numres = 0;
     my %tmp_results;
-    # fgrep is seriously faster than using perl
-    open DESC, '-|', 'fgrep', '-n', '--', $keyword, $file
-       or die "couldn't open $file: $!";
-    while (<DESC>) {
-       /^(\d+)/;
-       my $nr = $1;
-       debug( "Matched line $_", 2) if DEBUG;
-       my $result = $did2pkg->{$nr};
+    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;
@@ -217,14 +235,13 @@ sub do_fulltext_search {
        }
        last if $numres > 100;
     }
-    close DESC;
+    undef $db;
     $too_many_hits++ if $numres > 100;
 
-    my @results;
     foreach my $pkg (keys %tmp_results) {
        &$read_entry( $packages, $pkg, $results, $non_results, $opts );
     }
- }
+}
 
 sub find_binaries {
     my ($pkg, $archive, $suite, $src2bin) = @_;