]> git.deb.at Git - deb/packages.git/blob - bin/parse-packages
Merge branch 'master' into archive-master
[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 # Copyright (C) 2006  Jeroen van Wolffelaar <jeroen@wolffelaar.nl>
6 # Copyright (C) 2006-2007 Frank Lichtenheld <frank@lichtenheld.de>
7 #
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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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 Digest::MD5;
36 use Deb::Versions;
37 use Lingua::Stem v0.82;
38 use Search::Xapian;
39 use Packages::Config qw( $TOPDIR $DBDIR @ARCHIVES @SUITES );
40 &Packages::Config::init( './' );
41 my %packages_small = ();
42 my %virtual_packages = ();
43 my %package_names = ();
44 my %package_postfixes = ();
45 my %sources_packages = ();
46 my %descriptions = ();
47 my @descriptions = ("we count lines one-based\000");
48 my %packages_descriptions = ();
49 my %descriptions_packages = ();
50
51 my %sections = ();
52 my %subsections = ();
53 my %priorities = ();
54
55 $/ = "";
56
57 -d $DBDIR || mkpath( $DBDIR );
58 -d "$DBDIR/xapian.new" && rmtree("$DBDIR/xapian.new");
59 -d "$DBDIR/xapian.old" && rmtree("$DBDIR/xapian.old");
60 mkpath( "$DBDIR/xapian.new" );
61
62 for my $suite (@SUITES) {
63     my %package_names_suite = ();
64     my %packages_all_db;
65     tie %packages_all_db, "DB_File", "$DBDIR/packages_all_$suite.db.new",
66             O_RDWR|O_CREAT, 0666, $DB_BTREE
67             or die "Error creating DB: $!";
68
69     for my $archive (@ARCHIVES) {
70         print "Reading $archive/$suite...\n";
71         if (!-d "$TOPDIR/archive/$archive/$suite/") {
72                 print "\tseems not to exist, skipping...\n";
73                 next;
74         }
75         open PKG, "zcat $TOPDIR/archive/$archive/$suite/$what/{,debian-installer/}binary-*/Packages.gz|";
76         while (<PKG>) {
77                 next if /^\s*$/;
78                 my $data = "";
79                 my %data = ();
80                 chomp;
81                 s/\n /\377/g;
82                 while (/^(\S+):\s*(.*)\s*$/mg) {
83                         my ($key, $value) = ($1, $2);
84                         $value =~ s/\377/\n /g;
85                         $key =~ tr [A-Z] [a-z];
86                         $data{$key} = $value;
87                 }
88                 # Skip broken debs
89                 next unless $data{architecture};
90                 # Skip double package
91                 next if exists($packages_all_db{"$data{'package'} $data{'architecture'} $data{'version'}"});
92                 # Skip arch:all for amd64 & debports, any non-redundancy is
93                 # usually a bug anyway
94                 next if ($archive eq 'amd64' or $archive eq 'debports')
95                     and $data{architecture} eq 'all';
96
97                 if ($data{'provides'}) {
98                     foreach (split /\s*,\s*/, $data{'provides'}) {
99                         $virtual_packages{$_}{$suite}{$data{'package'}}++;
100                         $packages_small{$_} ||= {};
101                     }
102                 }
103                 $package_names{$data{'package'}} = 1;
104                 $package_names_suite{$data{'package'}} = 1;
105                 my $src = $data{'package'};
106                 my $src_version = '';
107                 if ($data{'source'}) {
108                         $src = $data{'source'};
109                         $src =~ s/\s+.*//o; # strip version info
110                 }
111                 $data{'source'} = $src;
112
113                 # expand tags like devel::{lang:c,lang:c++}
114                 if ($data{'tag'} && $data{'tag'} =~ /\{/) {
115                     my @complete_tags = split(/, /, $data{'tag'});
116                     my @tags;
117                     foreach (@complete_tags) {
118                         my ($facet, $tag) = split( /::/, $_, 2);
119                         if ($tag =~ s/^\{(.+)\}$/$1/) {
120                             warn "pkg=$data{package} tags=$data{tag}\n" unless $facet && $tag;
121                             foreach (split( /,/, $tag )) {
122                                 push @tags, "${facet}::$_";
123                             }
124                         } else {
125                             warn "pkg=$data{package} tags=$data{tag}\n" unless $facet && $tag;
126                             push @tags, "${facet}::$tag";
127                         }
128                     }
129                     my $old = $data{tag};
130                     $data{'tag'} = join ", ", @tags;
131                 }
132
133                 # we add some additional data here
134                 my $descr = "$data{'description'}\000$data{'package'}\000"
135                     .($data{'tag'}||'');
136                 my $sdescr = $data{'description'};
137                 $sdescr =~ s/\n.*//os;
138                 my $did = undef;
139                 if (exists($descriptions{$descr})) {
140                         $did  = $descriptions{$descr};
141                 } else {
142                         $did = 1 + $#descriptions;
143                         $descriptions[$did] = $descr;
144                         $descriptions{$descr} = $did;
145                 }
146                 $data{'description-md5'} = Digest::MD5::md5_hex($data{'description'}, "\n");
147                 $data{'description'} = $did;
148                 $packages_descriptions{"$data{'package'} $data{'version'} $data{'architecture'}"} = $did;
149                 $descriptions_packages{$did} .=
150                         "$data{'package'} $data{'version'} $data{'architecture'}\000";
151
152                 my $section = 'main';
153                 my $subsection = $data{section} || '-';
154                 if ($data{section} && ($data{section} =~ m=/=o)) {
155                     ($section, $subsection) = split m=/=o, $data{section}, 2;
156                 }
157                 $data{'section'} = $section;
158                 $data{'subsection'} = $subsection;
159                 $data{'priority'} ||= '-';
160                 $sections{$suite}{$section}++;
161                 $subsections{$suite}{$subsection}++;
162                 $priorities{$suite}{$data{priority}}++;
163                 my $pkgitem = "$archive $suite $data{'architecture'} ".
164                         "$section $subsection $data{'priority'} $data{'version'} $data{'description-md5'} $sdescr\0";
165                 my $previtem = ($packages_small{$data{'package'}}{$suite}{$data{'architecture'}}
166                     ||= $pkgitem);
167                 $packages_small{$data{'package'}}{$suite}{$data{'architecture'}} = $pkgitem
168                     if version_cmp($data{'version'}, (split /\s/o, $previtem)[6]) > 0;
169                 $previtem = ($packages_small{$data{'package'}}{$suite}{'any'}
170                     ||= $pkgitem);
171                 $packages_small{$data{'package'}}{$suite}{'any'} = $pkgitem
172                     if version_cmp($data{'version'}, (split /\s/o, $previtem)[6]) > 0;
173                 $sources_packages{$src} .=
174                         "$archive $suite $data{'package'} $data{'version'} $data{'architecture'}\000";
175                 $data{archive} = $archive;
176                 while (my ($key, $value) = each (%data)) {
177                         next if $key eq 'package' or $key eq 'version' or $key eq 'architecture'
178                                 or $key eq 'maintainer';
179                         print STDERR "WARN: $key ($suite/$archive/$data{package}/$data{architecture}\n" unless defined $value;
180                         $data .= "$key\00$value\00";
181                 }
182                 $data =~ s/.$//s;
183                 $packages_all_db{"$data{'package'} $data{'architecture'} $data{'version'}"}
184                         = $data;
185         }
186     }
187
188     open NAMES, '>', "$DBDIR/package_names_$suite.txt.new"
189         or die "Error creating package names list: $!";
190     foreach (sort keys %package_names_suite) {
191         print NAMES "$_\n";
192     }
193     close NAMES;
194
195     untie %packages_all_db;
196 }
197
198 print "Writing databases...\n";
199 my %packages_small_db;
200 tie %packages_small_db, "DB_File", "$DBDIR/packages_small.db.new",
201         O_RDWR|O_CREAT, 0666, $DB_BTREE
202         or die "Error creating DB: $!";
203 while (my ($pkg, $v) = each(%packages_small)) {
204         my ($res1, $res2, $res3) = ("", "", "");
205         while (my ($suite, $v2) = each %$v) {
206             $res2 .= $v2->{'any'};
207             while (my ($arch, $v3) = each %$v2) {
208                 next if $arch eq 'any' or $v3 eq $v2->{'any'};
209                 $res3 .= $v3;
210             }
211         }
212
213         if (exists $virtual_packages{$pkg}) {
214                 while (my ($suite, $v2) = each %{$virtual_packages{$pkg}}) {
215                         $res1 .= "$suite\01".(join ' ', keys %$v2)."\01";
216                 }
217         } else {
218                 $res1 .= "-\01-\01";
219         }
220         $res1 =~ s/.$/\0/s;
221
222         my $res = "$res1$res2$res3";
223         $res =~ s/.$//s;
224         $packages_small_db{$pkg} = $res;
225 }
226 untie %packages_small_db;
227
228 my %sources_packages_db;
229 tie %sources_packages_db, "DB_File", "$DBDIR/sources_packages.db.new",
230         O_RDWR|O_CREAT, 0666, $DB_BTREE
231         or die "Error creating DB: $!";
232 while (my ($k, $v) = each(%sources_packages)) {
233         $v =~ s/.$//s;
234         $sources_packages_db{$k} = $v;
235 }
236 untie %sources_packages_db;
237
238 my %packages_descriptions_db;
239 tie %packages_descriptions_db, "DB_File", "$DBDIR/packages_descriptions.db.new",
240         O_RDWR|O_CREAT, 0666, $DB_BTREE
241         or die "Error creating DB: $!";
242 while (my ($k, $v) = each(%packages_descriptions)) {
243         $packages_descriptions_db{$k} = $v;
244 }
245 untie %packages_descriptions_db;
246
247 my %descriptions_packages_db;
248 tie %descriptions_packages_db, "DB_File", "$DBDIR/descriptions_packages.db.new",
249         O_RDWR|O_CREAT, 0666, $DB_BTREE
250         or die "Error creating DB: $!";
251 while (my ($k, $v) = each(%descriptions_packages)) {
252         $v =~ s/.$//s;
253         $descriptions_packages_db{$k} = $v;
254 }
255 untie %descriptions_packages_db;
256
257 my $stemmer = Lingua::Stem->new();
258 $stemmer->stem_caching({ -level => 2 });
259 my $xapian_db;
260 eval {
261     $xapian_db = Search::Xapian::WritableDatabase->new("$DBDIR/xapian.new",
262                                                        Search::Xapian::DB_CREATE_OR_OPEN)
263         or die "can't create write-able db object: $!\n";
264 };
265 die $@ if $@;
266 my %descriptions_db;
267 tie %descriptions_db, "DB_File", "$DBDIR/descriptions.db.new",
268         O_RDWR|O_CREAT, 0666, $DB_BTREE
269         or die "Error creating DB: $!";
270 print "Index $#descriptions descriptions\n";
271 for (my $i=1; $i<= $#descriptions; $i++) {
272         # strip away additional data
273         my ($only_desc, $pkg, $tags) = split /\000/o, $descriptions[$i], 3;
274         my $orig_desc = $only_desc;
275 # WARNING: This needs to correspond with what happens in
276 # Packages/Search.pm:do_xapian_search
277         $only_desc =~ s#[^\w/+]+# #og;
278
279         #XAPIAN
280         eval {
281             my @words = split /\s+/, $only_desc;
282             unshift @words, $pkg;
283
284             my $doc = Search::Xapian::Document->new()
285                 or die "can't create doc object for $i: $!\n";
286             if ($doc->set_data($i)){
287                 warn "can't set_data in doc object for $i: $!\n";
288             }
289
290             # package with prefix
291             if ($doc->add_term("P$pkg")) {
292                 warn "can't add term P$pkg: $!\n";
293             }
294             # description, unstemmed with positional info
295             for my $j (0 .. (@words-1)) {
296                 next if $words[$j] =~ /^\s*$/o;
297                 if ($doc->add_posting($words[$j], $j)) {
298                     warn "can't add posting $words[$j] at $j: $!\n";
299                 }
300             }
301             # description, stemmed
302             my $stem_words = $stemmer->stem( \@words );
303             foreach my $w (@$stem_words) {
304                 next if $w =~ /^\s*$/o;
305                 if ($doc->add_term($w)) {
306                     warn "can't add term $w: $!\n";
307                 }
308             }
309             if ($tags) {
310                 foreach my $t (split m/, /, $tags) {
311                     if ($doc->add_term($t)) {
312                         warn "can't add term $t: $!\n";
313                     }
314                 }
315             }
316
317             $xapian_db->add_document($doc)
318                 or warn "failed to add document: $i\n";
319         };
320         die $@ if $@;
321
322         $descriptions_db{$i} = $orig_desc;
323 }
324 untie %descriptions_db;
325 $xapian_db->flush;
326 undef $xapian_db;
327
328 # package names stuff:
329 for my $pkg (keys %package_names) {
330         for (my $i=0;$i<length($pkg)-1;$i++) {
331                 my $before = substr($pkg, 0, $i);
332                 my $after = substr($pkg, $i);
333                 $before = "^" if $before eq ""; # otherwise split doesn't work properly
334                 $package_postfixes{$after} .= "$before\0";
335         }
336 }
337 my %package_postfixes_db;
338 tie %package_postfixes_db, "DB_File", "$DBDIR/package_postfixes.db.new",
339         O_RDWR|O_CREAT, 0666, $DB_BTREE
340         or die "Error creating DB: $!";
341 while (my ($k, $v) = each(%package_postfixes)) {
342         $v =~ s/.$//s;
343         my $nr = ($v =~ tr/\000/\000/) + 1;
344         if ($nr > $MAX_PACKAGE_POSTFIXES) {
345             $v = ($v =~ /\^/) ? "^\001" . $nr
346                 : "\001" . $nr;
347         }
348         $package_postfixes_db{$k} = $v;
349 }
350 untie %package_postfixes_db;
351
352 store \%sections, "$DBDIR/sections.info";
353 store \%subsections, "$DBDIR/subsections.info";
354 store \%priorities, "$DBDIR/priorities.info";
355
356 rename("$DBDIR/packages_small.db.new", "$DBDIR/packages_small.db");
357 rename("$DBDIR/sources_packages.db.new", "$DBDIR/sources_packages.db");
358 for my $suite (@SUITES) {
359         rename("$DBDIR/packages_all_$suite.db.new",
360                "$DBDIR/packages_all_$suite.db");
361         rename("$DBDIR/package_names_$suite.txt.new",
362                "$DBDIR/package_names_$suite.txt");
363 }
364 rename("$DBDIR/xapian", "$DBDIR/xapian.old");
365 rename("$DBDIR/xapian.new","$DBDIR/xapian");
366 rmtree("$DBDIR/xapian.old");
367 rename("$DBDIR/packages_descriptions.db.new",
368        "$DBDIR/packages_descriptions.db");
369 rename("$DBDIR/descriptions_packages.db.new",
370        "$DBDIR/descriptions_packages.db");
371 rename("$DBDIR/descriptions.db.new", "$DBDIR/descriptions.db");
372 rename("$DBDIR/package_postfixes.db.new", "$DBDIR/package_postfixes.db");