]> git.deb.at Git - deb/packages.git/blob - bin/parse-packages
Description Search: Move to Xapian
[deb/packages.git] / bin / parse-packages
1 #!/usr/bin/perl -w
2 # Convert Packages.gz files into Sleepycat db files for efficient usage of
3 # data
4 #
5 # $Id$
6 #
7 # Copyright (C) 2006  Jeroen van Wolffelaar <jeroen@wolffelaar.nl>
8 # This program is free software; you can redistribute it and/or modify
9 # it under the terms of the GNU General Public License as published by
10 # the Free Software Foundation; either version 2 of the License, or
11 # (at your option) any later version.
12
13 # This program is distributed in the hope that it will be useful,
14 # but WITHOUT ANY WARRANTY; without even the implied warranty of
15 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16 # GNU General Public License for more details.
17
18 # You should have received a copy of the GNU General Public License
19 # along with this program; if not, write to the Free Software
20 # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
21
22 use strict;
23 use warnings;
24 use lib './lib';
25
26 $| = 1;
27
28 my $what = $ARGV[0] ? "non-free" : "*";
29 # max. distinct results for a given package postfix
30 my $MAX_PACKAGE_POSTFIXES = 100;
31
32 use DB_File;
33 use Storable;
34 use File::Path;
35 use Deb::Versions;
36 use Lingua::Stem v0.82;
37 use Search::Xapian;
38 use Packages::Config qw( $TOPDIR $DBDIR @ARCHIVES @SUITES );
39 &Packages::Config::init( './' );
40 my %packages_small = ();
41 my %virtual_packages = ();
42 my %package_names = ();
43 my %package_postfixes = ();
44 my %sources_packages = ();
45 my %descriptions = ();
46 my @descriptions = ("we count lines one-based\000");
47 my %packages_descriptions = ();
48 my %descriptions_packages = ();
49
50 my %sections = ();
51 my %subsections = ();
52 my %priorities = ();
53
54 $/ = "";
55
56 -d $DBDIR || mkpath( $DBDIR );
57 -d "$DBDIR/xapian.new" && rmtree("$DBDIR/xapian.new");
58 -d "$DBDIR/xapian.old" && rmtree("$DBDIR/xapian.old");
59 mkpath( "$DBDIR/xapian.new" );
60
61 for my $suite (@SUITES) {
62     my %package_names_suite = ();
63     my %packages_all_db;
64     tie %packages_all_db, "DB_File", "$DBDIR/packages_all_$suite.db.new",
65             O_RDWR|O_CREAT, 0666, $DB_BTREE
66             or die "Error creating DB: $!";
67
68     for my $archive (@ARCHIVES) {
69         print "Reading $archive/$suite...\n";
70         if (!-d "$TOPDIR/archive/$archive/$suite/") {
71                 print "\tseems not to exist, skipping...\n";
72                 next;
73         }
74         open PKG, "zcat $TOPDIR/archive/$archive/$suite/$what/{,debian-installer/}binary-*/Packages.gz|";
75         while (<PKG>) {
76                 next if /^\s*$/;
77                 my $data = "";
78                 my %data = ();
79                 chomp;
80                 s/\n /\377/g;
81                 while (/^(\S+):\s*(.*)\s*$/mg) {
82                         my ($key, $value) = ($1, $2);
83                         $value =~ s/\377/\n /g;
84                         $key =~ tr [A-Z] [a-z];
85                         $data{$key} = $value;
86                 }
87                 # Skip double package
88                 next if exists($packages_all_db{"$data{'package'} $data{'architecture'} $data{'version'}"});
89                 # Skip arch:all for amd64&kfreebsd, too often broken
90                 next if ($archive eq 'amd64' or $archive eq 'kfreebsd')
91                     and $data{architecture} eq 'all';
92
93                 if ($data{'provides'}) {
94                     foreach (split /\s*,\s*/, $data{'provides'}) {
95                         $virtual_packages{$_}{$suite}{$data{'package'}}++;
96                         $packages_small{$_} ||= {};
97                     }
98                 }
99                 $package_names{$data{'package'}} = 1;
100                 $package_names_suite{$data{'package'}} = 1;
101                 my $src = $data{'package'};
102                 my $src_version = '';
103                 if ($data{'source'}) {
104                         $src = $data{'source'};
105                         $src =~ s/\s+.*//o; # strip version info
106                 }
107                 $data{'source'} = $src;
108                 # we add some additional data here
109                 my $descr = "$data{'description'}\000$data{'package'}\000"
110                     .($data{'tag'}||'');
111                 my $sdescr = $data{'description'};
112                 $sdescr =~ s/\n.*//os;
113                 my $did = undef;
114                 if (exists($descriptions{$descr})) {
115                         $did  = $descriptions{$descr};
116                 } else {
117                         $did = 1 + $#descriptions;
118                         $descriptions[$did] = $descr;
119                         $descriptions{$descr} = $did;
120                 }
121                 $data{'description'} = $did;
122                 $packages_descriptions{"$data{'package'} $data{'version'} $data{'architecture'}"} = $did;
123                 $descriptions_packages{$did} .=
124                         "$data{'package'} $data{'version'} $data{'architecture'}\000";
125
126                 my $section = 'main';
127                 my $subsection = $data{section} || '-';
128                 if ($data{section} && ($data{section} =~ m=/=o)) {
129                     ($section, $subsection) = split m=/=o, $data{section}, 2;
130                     ($subsection, $section) = split m=/=o, $data{section}, 2
131                         if $section eq 'non-US';
132                 }
133                 $data{'section'} = $section;
134                 $data{'subsection'} = $subsection;
135                 $data{'priority'} ||= '-';
136                 $sections{$suite}{$section}++;
137                 $subsections{$suite}{$subsection}++;
138                 $priorities{$suite}{$data{priority}}++;
139                 my $pkgitem = "$archive $suite $data{'architecture'} ".
140                         "$section $subsection $data{'priority'} $data{'version'} $sdescr\0";
141                 my $previtem = ($packages_small{$data{'package'}}{$suite}{$data{'architecture'}}
142                     ||= $pkgitem);
143                 $packages_small{$data{'package'}}{$suite}{$data{'architecture'}} = $pkgitem
144                     if version_cmp($data{'version'}, (split /\s/o, $previtem)[6]) > 0;
145                 $previtem = ($packages_small{$data{'package'}}{$suite}{'any'}
146                     ||= $pkgitem);
147                 $packages_small{$data{'package'}}{$suite}{'any'} = $pkgitem
148                     if version_cmp($data{'version'}, (split /\s/o, $previtem)[6]) > 0;
149                 $sources_packages{$src} .=
150                         "$archive $suite $data{'package'} $data{'version'} $data{'architecture'}\000";
151                 $data{archive} = $archive;
152                 while (my ($key, $value) = each (%data)) {
153                         next if $key eq 'package' or $key eq 'version' or $key eq 'architecture'
154                                 or $key eq 'maintainer';
155                         print STDERR "WARN: $key ($suite/$archive/$data{package}/$data{architecture}\n" unless defined $value;
156                         $data .= "$key\00$value\00";
157                 }
158                 $data =~ s/.$//s;
159                 $packages_all_db{"$data{'package'} $data{'architecture'} $data{'version'}"}
160                         = $data;
161         }
162     }
163
164     open NAMES, '>', "$DBDIR/package_names_$suite.txt.new"
165         or die "Error creating package names list: $!";
166     foreach (sort keys %package_names_suite) {
167         print NAMES "$_\n";
168     }
169     close NAMES;
170
171     untie %packages_all_db;
172 }
173
174 print "Writing databases...\n";
175 my %packages_small_db;
176 tie %packages_small_db, "DB_File", "$DBDIR/packages_small.db.new",
177         O_RDWR|O_CREAT, 0666, $DB_BTREE
178         or die "Error creating DB: $!";
179 while (my ($pkg, $v) = each(%packages_small)) {
180         my ($res1, $res2, $res3) = ("", "", "");
181         while (my ($suite, $v2) = each %$v) {
182             $res2 .= $v2->{'any'};
183             while (my ($arch, $v3) = each %$v2) {
184                 next if $arch eq 'any' or $v3 eq $v2->{'any'};
185                 $res3 .= $v3;
186             }
187         }
188
189         if (exists $virtual_packages{$pkg}) {
190                 while (my ($suite, $v2) = each %{$virtual_packages{$pkg}}) {
191                         $res1 .= "$suite\01".(join ' ', keys %$v2)."\01";
192                 }
193         } else {
194                 $res1 .= "-\01-\01";
195         }
196         $res1 =~ s/.$/\0/s;
197
198         my $res = "$res1$res2$res3";
199         $res =~ s/.$//s;
200         $packages_small_db{$pkg} = $res;
201 }
202 untie %packages_small_db;
203
204 my %sources_packages_db;
205 tie %sources_packages_db, "DB_File", "$DBDIR/sources_packages.db.new",
206         O_RDWR|O_CREAT, 0666, $DB_BTREE
207         or die "Error creating DB: $!";
208 while (my ($k, $v) = each(%sources_packages)) {
209         $v =~ s/.$//s;
210         $sources_packages_db{$k} = $v;
211 }
212 untie %sources_packages_db;
213
214 my %packages_descriptions_db;
215 tie %packages_descriptions_db, "DB_File", "$DBDIR/packages_descriptions.db.new",
216         O_RDWR|O_CREAT, 0666, $DB_BTREE
217         or die "Error creating DB: $!";
218 while (my ($k, $v) = each(%packages_descriptions)) {
219         $packages_descriptions_db{$k} = $v;
220 }
221 untie %packages_descriptions_db;
222
223 my %descriptions_packages_db;
224 tie %descriptions_packages_db, "DB_File", "$DBDIR/descriptions_packages.db.new",
225         O_RDWR|O_CREAT, 0666, $DB_BTREE
226         or die "Error creating DB: $!";
227 while (my ($k, $v) = each(%descriptions_packages)) {
228         $v =~ s/.$//s;
229         $descriptions_packages_db{$k} = $v;
230 }
231 untie %descriptions_packages_db;
232
233 my $stemmer = Lingua::Stem->new();
234 $stemmer->stem_caching({ -level => 2 });
235 my $xapian_db;
236 eval {
237     $xapian_db = Search::Xapian::WritableDatabase->new("$DBDIR/xapian.new",
238                                                        Search::Xapian::DB_CREATE_OR_OPEN)
239         or die "can't create write-able db object: $!\n";
240 };
241 die $@ if $@;
242 my %descriptions_db;
243 tie %descriptions_db, "DB_File", "$DBDIR/descriptions.db.new",
244         O_RDWR|O_CREAT, 0666, $DB_BTREE
245         or die "Error creating DB: $!";
246 open DESCR, ">", "$DBDIR/descriptions.txt" or die "Error creating descriptions textfile";
247 print "Index $#descriptions descriptions\n";
248 for (my $i=1; $i<= $#descriptions; $i++) {
249         my $plain_description = $descriptions[$i];
250         # strip away additional data
251         my ($only_desc) = split /\000/o, $plain_description, 2;
252 # WARNING: This needs to correspond with what happens in
253 # Packages/Search.pm:do_fulltext_search
254         $plain_description =~ tr [A-Z] [a-z];
255         # ensure one space on both ends
256         $plain_description = " $plain_description ";
257         $plain_description =~ s/[(),.-]+//og;
258         $plain_description =~ s#[^a-z0-9_/+]+# #og;
259         print DESCR "$plain_description\n";
260
261         #XAPIAN
262         eval {
263             my @words = split /\s+/, $plain_description;
264             my $stem_words = $stemmer->stem( \@words );
265             my $doc = Search::Xapian::Document->new()
266                 or die "can't create doc object for $i: $!\n";
267             if ($doc->set_data($i)){
268                 warn "can't set_data in doc object for $i: $!\n";
269             }
270             for my $j (0 .. (@$stem_words-1)) {
271                 next if $stem_words->[$j] =~ /^\s*$/o;
272                 if ($doc->add_posting($stem_words->[$j], $j)) {
273                     warn "can't add word $stem_words->[$j] $j: $!\n";
274                 }
275             }
276             $xapian_db->add_document($doc)
277                 or warn "failed to add document: $i\n";
278         };
279         die $@ if $@;
280
281         $descriptions_db{$i} = $only_desc;
282 }
283 close DESCR;
284 untie %descriptions_db;
285 $xapian_db->flush;
286 undef $xapian_db;
287
288 # package names stuff:
289 for my $pkg (keys %package_names) {
290         for (my $i=0;$i<length($pkg)-1;$i++) {
291                 my $before = substr($pkg, 0, $i);
292                 my $after = substr($pkg, $i);
293                 $before = "^" if $before eq ""; # otherwise split doesn't work properly
294                 $package_postfixes{$after} .= "$before\0";
295         }
296 }
297 my %package_postfixes_db;
298 tie %package_postfixes_db, "DB_File", "$DBDIR/package_postfixes.db.new",
299         O_RDWR|O_CREAT, 0666, $DB_BTREE
300         or die "Error creating DB: $!";
301 while (my ($k, $v) = each(%package_postfixes)) {
302         $v =~ s/.$//s;
303         my $nr = $v;
304         $nr =~ s/[^\000]//g;
305         $nr = length($nr) + 1; # < number of hits
306         if ($nr > $MAX_PACKAGE_POSTFIXES) {
307                 $v = "\001" . $nr;
308         }
309         $package_postfixes_db{$k} = $v;
310 }
311 untie %package_postfixes_db;
312
313 store \%sections, "$DBDIR/sections.info";
314 store \%subsections, "$DBDIR/subsections.info";
315 store \%priorities, "$DBDIR/priorities.info";
316
317 rename("$DBDIR/packages_small.db.new", "$DBDIR/packages_small.db");
318 rename("$DBDIR/sources_packages.db.new", "$DBDIR/sources_packages.db");
319 for my $suite (@SUITES) {
320         rename("$DBDIR/packages_all_$suite.db.new",
321                "$DBDIR/packages_all_$suite.db");
322         rename("$DBDIR/package_names_$suite.txt.new",
323                "$DBDIR/package_names_$suite.txt");
324 }
325 rename("$DBDIR/xapian", "$DBDIR/xapian.old");
326 rename("$DBDIR/xapian.new","$DBDIR/xapian");
327 rmtree("$DBDIR/xapian.old");
328 rename("$DBDIR/packages_descriptions.db.new",
329        "$DBDIR/packages_descriptions.db");
330 rename("$DBDIR/descriptions_packages.db.new",
331        "$DBDIR/descriptions_packages.db");
332 rename("$DBDIR/descriptions.txt.new", "$DBDIR/descriptions.txt");
333 rename("$DBDIR/descriptions.db.new", "$DBDIR/descriptions.db");
334 rename("$DBDIR/package_postfixes.db.new", "$DBDIR/package_postfixes.db");