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