Description Search: Move to Xapian

Use Xapian as Backend for fulltext search in descriptions.
Introduces new do_xapian_search function to be able to switch
to the old do_fulltext_search in case of problems.

diff --git a/bin/parse-packages b/bin/parse-packages
index 189ea0267ea78e946f1b3def780ea1561b1c149d..fa81fd400c60676140a79c7e8cbd43b42ef87904 100755 (executable)
--- a/bin/parse-packages
+++ b/bin/parse-packages
@@ -33,6 +33,8 @@ use DB_File;
 use Storable;
 use File::Path;
 use Deb::Versions;
+use Lingua::Stem v0.82;
+use Search::Xapian;
 use Packages::Config qw( $TOPDIR $DBDIR @ARCHIVES @SUITES );
 &Packages::Config::init( './' );
 my %packages_small = ();
@@ -52,6 +54,9 @@ my %priorities = ();
 $/ = "";
 
 -d $DBDIR || mkpath( $DBDIR );
+-d "$DBDIR/xapian.new" && rmtree("$DBDIR/xapian.new");
+-d "$DBDIR/xapian.old" && rmtree("$DBDIR/xapian.old");
+mkpath( "$DBDIR/xapian.new" );
 
 for my $suite (@SUITES) {
     my %package_names_suite = ();
@@ -97,10 +102,14 @@ for my $suite (@SUITES) {
                my $src_version = '';
                if ($data{'source'}) {
                        $src = $data{'source'};
-                       $src =~ s/\s+.*//; # strip version info
+                       $src =~ s/\s+.*//o; # strip version info
                }
                $data{'source'} = $src;
-               my $descr = $data{'description'};
+               # we add some additional data here
+               my $descr = "$data{'description'}\000$data{'package'}\000"
+                   .($data{'tag'}||'');
+               my $sdescr = $data{'description'};
+               $sdescr =~ s/\n.*//os;
                my $did = undef;
                if (exists($descriptions{$descr})) {
                        $did  = $descriptions{$descr};
@@ -114,8 +123,6 @@ for my $suite (@SUITES) {
                $descriptions_packages{$did} .=
                        "$data{'package'} $data{'version'} $data{'architecture'}\000";
 
-               my $sdescr = $descr;
-               $sdescr =~ s/\n.*//s;
                my $section = 'main';
                my $subsection = $data{section} || '-';
                if ($data{section} && ($data{section} =~ m=/=o)) {
@@ -178,7 +185,7 @@ while (my ($pkg, $v) = each(%packages_small)) {
                $res3 .= $v3;
            }
        }
-       
+
        if (exists $virtual_packages{$pkg}) {
                while (my ($suite, $v2) = each %{$virtual_packages{$pkg}}) {
                        $res1 .= "$suite\01".(join ' ', keys %$v2)."\01";
@@ -223,13 +230,25 @@ while (my ($k, $v) = each(%descriptions_packages)) {
 }
 untie %descriptions_packages_db;
 
+my $stemmer = Lingua::Stem->new();
+$stemmer->stem_caching({ -level => 2 });
+my $xapian_db;
+eval {
+    $xapian_db = Search::Xapian::WritableDatabase->new("$DBDIR/xapian.new",
+                                                      Search::Xapian::DB_CREATE_OR_OPEN)
+       or die "can't create write-able db object: $!\n";
+};
+die $@ if $@;
 my %descriptions_db;
 tie %descriptions_db, "DB_File", "$DBDIR/descriptions.db.new",
        O_RDWR|O_CREAT, 0666, $DB_BTREE
        or die "Error creating DB: $!";
 open DESCR, ">", "$DBDIR/descriptions.txt" or die "Error creating descriptions textfile";
+print "Index $#descriptions descriptions\n";
 for (my $i=1; $i<= $#descriptions; $i++) {
        my $plain_description = $descriptions[$i];
+       # strip away additional data
+       my ($only_desc) = split /\000/o, $plain_description, 2;
 # WARNING: This needs to correspond with what happens in
 # Packages/Search.pm:do_fulltext_search
        $plain_description =~ tr [A-Z] [a-z];
@@ -238,10 +257,33 @@ for (my $i=1; $i<= $#descriptions; $i++) {
        $plain_description =~ s/[(),.-]+//og;
        $plain_description =~ s#[^a-z0-9_/+]+# #og;
        print DESCR "$plain_description\n";
-       $descriptions_db{$i} = $descriptions[$i];
+
+       #XAPIAN
+       eval {
+           my @words = split /\s+/, $plain_description;
+           my $stem_words = $stemmer->stem( \@words );
+           my $doc = Search::Xapian::Document->new()
+               or die "can't create doc object for $i: $!\n";
+           if ($doc->set_data($i)){
+               warn "can't set_data in doc object for $i: $!\n";
+           }
+           for my $j (0 .. (@$stem_words-1)) {
+               next if $stem_words->[$j] =~ /^\s*$/o;
+               if ($doc->add_posting($stem_words->[$j], $j)) {
+                   warn "can't add word $stem_words->[$j] $j: $!\n";
+               }
+           }
+           $xapian_db->add_document($doc)
+               or warn "failed to add document: $i\n";
+       };
+       die $@ if $@;
+
+       $descriptions_db{$i} = $only_desc;
 }
 close DESCR;
 untie %descriptions_db;
+$xapian_db->flush;
+undef $xapian_db;
 
 # package names stuff:
 for my $pkg (keys %package_names) {
@@ -280,6 +322,9 @@ for my $suite (@SUITES) {
        rename("$DBDIR/package_names_$suite.txt.new",
               "$DBDIR/package_names_$suite.txt");
 }
+rename("$DBDIR/xapian", "$DBDIR/xapian.old");
+rename("$DBDIR/xapian.new","$DBDIR/xapian");
+rmtree("$DBDIR/xapian.old");
 rename("$DBDIR/packages_descriptions.db.new",
        "$DBDIR/packages_descriptions.db");
 rename("$DBDIR/descriptions_packages.db.new",

diff --git a/lib/Packages/DoSearch.pm b/lib/Packages/DoSearch.pm
index 2603ff9a10c848333643f27e98a6f450e4c36611..a94dd7bc33e46fd54415cbc50011d9e45b7128b4 100644 (file)
--- a/lib/Packages/DoSearch.pm
+++ b/lib/Packages/DoSearch.pm
@@ -53,10 +53,22 @@ sub do_search {
            do_names_search( [ @keywords ], \%packages, $p_obj,
                             \&read_entry_all, $opts,
                             \@results, \@non_results );
-           do_fulltext_search( [ @keywords ], "$DBDIR/descriptions.txt",
+#          my $fts0 = new Benchmark;
+#          do_fulltext_search( [ @keywords ], "$DBDIR/descriptions.txt",
+#                              \%did2pkg, \%packages,
+#                              \&read_entry_all, $opts,
+#                              \@results, \@non_results );
+           my $fts1 = new Benchmark;
+           do_xapian_search( [ @keywords ], "$DBDIR/xapian/",
                                \%did2pkg, \%packages,
                                \&read_entry_all, $opts,
                                \@results, \@non_results );
+           my $fts2 = new Benchmark;
+#          my $fts_grep = timediff($fts1,$fts0);
+           my $fts_xapian = timediff($fts2,$fts1);
+#          debug( "Fulltext search took ".timestr($fts_grep)." (grep)" ) if DEBUG;
+           debug( "Fulltext search took ".timestr($fts_xapian)." (Xapian)" )
+               if DEBUG;
        }
     }
 

diff --git a/lib/Packages/Search.pm b/lib/Packages/Search.pm
index c6a521ad65385d2cf11daa6f9ca95e8f00f3eb40..1c369868db94014cc26e8c7bf3186f4da00f8013 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 ] );
 
@@ -242,6 +244,52 @@ sub do_fulltext_search {
     }
  }
 
+sub do_xapian_search {
+    my ($keywords, $db, $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 $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;
+    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;
+#          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;
+    }
+    undef $db;
+    $too_many_hits++ if $numres > 100;
+
+    foreach my $pkg (keys %tmp_results) {
+       &$read_entry( $packages, $pkg, $results, $non_results, $opts );
+    }
+}
+
 sub find_binaries {
     my ($pkg, $archive, $suite, $src2bin) = @_;