1 # Copyright (C) 2015, all contributors <meta@public-inbox.org>
2 # License: AGPLv3 or later (https://www.gnu.org/licenses/agpl-3.0.txt)
3 # based on notmuch, but with no concept of folders, files or flags
4 package PublicInbox::Search;
7 use PublicInbox::SearchMsg;
8 use Search::Xapian qw/:standard/;
9 require PublicInbox::View;
11 use PublicInbox::MID qw/mid_clean mid_compressed/;
17 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
21 my %bool_pfx_internal = (
22 type => 'T', # "mail" or "ghost"
23 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
26 my %bool_pfx_external = (
28 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
29 references => 'XREFS',
37 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
39 sub xpfx { $all_pfx{$_[0]} }
42 while (my ($k, $v) = each %all_pfx) {
43 next if $prob_prefix{$k};
44 $PFX2TERM_RMAP{$v} = $k;
47 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
50 my ($class, $git_dir, $writable) = @_;
51 # allow concurrent versions for easier rollback:
52 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
55 if ($writable) { # not used by the WWW interface
56 require Search::Xapian::WritableDatabase;
58 File::Path::mkpath($dir);
59 $db = Search::Xapian::WritableDatabase->new($dir,
60 Search::Xapian::DB_CREATE_OR_OPEN);
62 $db = Search::Xapian::Database->new($dir);
64 bless { xdb => $db, git_dir => $git_dir }, $class;
67 sub reopen { $_[0]->{xdb}->reopen }
70 my ($self, $mime) = @_; # mime = Email::MIME object
71 my $db = $self->{xdb};
74 my $mid = mid_clean($mime->header('Message-ID'));
75 $mid = mid_compressed($mid);
77 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
78 my $enc_msg = PublicInbox::View::enc_for($ct_msg);
80 $db->begin_transaction;
82 my $smsg = $self->lookup_message($mid);
86 $smsg->ensure_metadata;
87 # convert a ghost to a regular message
88 # it will also clobber any existing regular message
92 my $type = xpfx('type');
94 $doc->remove_term($type . 'ghost');
98 # probably does not exist:
99 eval { $doc->remove_term($type . 'mail') };
100 $doc->add_term($type . 'mail');
102 $smsg = PublicInbox::SearchMsg->new($mime);
104 $doc->add_term(xpfx('mid') . $mid);
107 my $subj = $smsg->subject;
110 $doc->add_term(xpfx('subject') . $subj);
112 my $path = subject_path($subj);
113 $doc->add_term(xpfx('path') . $path);
116 my $from = $smsg->from_name;
117 my $date = $smsg->date;
118 my $ts = Search::Xapian::sortable_serialise($smsg->ts);
119 $doc->add_value(PublicInbox::Search::TS, $ts);
121 $doc->set_data($smsg->to_doc_data);
123 my $tg = $self->term_generator;
125 $tg->set_document($doc);
126 $tg->index_text($subj, 1, 'S') if $subj;
127 $tg->increase_termpos;
128 $tg->index_text($subj) if $subj;
129 $tg->increase_termpos;
131 $tg->index_text($smsg->from->format);
132 $tg->increase_termpos;
134 $mime->walk_parts(sub {
136 return if $part->subparts; # walk_parts already recurses
137 my $ct = $part->content_type || $ct_msg;
139 # account for filter bugs...
140 $ct =~ m!\btext/plain\b!i or return;
142 my $enc = PublicInbox::View::enc_for($ct, $enc_msg);
144 foreach my $l (split(/\n/, $enc->decode($part->body))) {
152 $tg->index_text(join("\n", @quot), 0);
153 $tg->increase_termpos;
156 $tg->index_text(join("\n", @orig));
157 $tg->increase_termpos;
162 $doc_id = $smsg->doc_id;
163 $self->link_message($smsg, 0);
164 $db->replace_document($doc_id, $doc);
166 $self->link_message($smsg, 0);
167 $doc_id = $db->add_document($doc);
172 warn "failed to index message <$mid>: $@\n";
173 $db->cancel_transaction;
175 $db->commit_transaction;
180 # returns deleted doc_id on success, undef on missing
182 my ($self, $mid) = @_;
183 my $db = $self->{xdb};
185 $mid = mid_clean($mid);
186 $mid = mid_compressed($mid);
188 $db->begin_transaction;
190 $doc_id = $self->find_unique_doc_id('mid', $mid);
191 $db->delete_document($doc_id) if defined $doc_id;
195 warn "failed to remove message <$mid>: $@\n";
196 $db->cancel_transaction;
198 $db->commit_transaction;
205 my ($self, $query_string, $opts) = @_;
206 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
208 $query = Search::Xapian::Query->new(OP_AND, $mail_query, $query);
209 $self->do_enquire($query, $opts);
212 # given a message ID, get replies to a message
214 my ($self, $mid, $opts) = @_;
215 $mid = mid_clean($mid);
216 $mid = mid_compressed($mid);
218 my $irt = $qp->parse_query("inreplyto:$mid", 0);
219 my $ref = $qp->parse_query("references:$mid", 0);
220 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
222 $self->do_enquire($query);
226 my ($self, $mid, $opts) = @_;
227 my $smsg = eval { $self->lookup_message($mid) };
229 return { count => 0, msgs => [] } unless $smsg;
231 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
232 my $qsub = $qp->parse_query('path:'.$smsg->path);
233 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
234 $self->do_enquire($query);
240 my ($self, $query, $opts) = @_;
241 my $enquire = $self->enquire;
243 $enquire->set_query($query);
244 $enquire->set_sort_by_relevance_then_value(TS, 0);
246 my $offset = $opts->{offset} || 0;
247 my $limit = $opts->{limit} || 50;
248 my $mset = $enquire->get_mset($offset, $limit);
250 PublicInbox::SearchMsg->load_doc($_->get_document);
253 { count => $mset->get_matches_estimated, msgs => \@msgs }
257 sub stemmer { Search::Xapian::Stem->new(LANG) }
263 my $qp = $self->{query_parser};
267 $qp = Search::Xapian::QueryParser->new;
268 $qp->set_default_op(OP_AND);
269 $qp->set_database($self->{xdb});
270 $qp->set_stemmer($self->stemmer);
271 $qp->set_stemming_strategy(STEM_SOME);
272 $qp->add_valuerangeprocessor($self->ts_range_processor);
273 $qp->add_valuerangeprocessor($self->date_range_processor);
275 while (my ($name, $prefix) = each %bool_pfx_external) {
276 $qp->add_boolean_prefix($name, $prefix);
279 while (my ($name, $prefix) = each %prob_prefix) {
280 $qp->add_prefix($name, $prefix);
283 $self->{query_parser} = $qp;
286 sub term_generator { # write-only
289 my $tg = $self->{term_generator};
292 $tg = Search::Xapian::TermGenerator->new;
293 $tg->set_stemmer($self->stemmer);
295 $self->{term_generator} = $tg;
298 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
300 # increments last_thread_id counter
301 # returns a 64-bit integer represented as a hex string
304 my $db = $self->{xdb};
305 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
307 $db->set_metadata('last_thread_id', ++$last_thread_id);
312 sub ts_range_processor {
313 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
316 sub date_range_processor {
317 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
321 my ($self, $smsg, $is_ghost) = @_;
324 $smsg->ensure_metadata;
326 $self->link_message_to_parents($smsg);
330 sub link_message_to_parents {
331 my ($self, $smsg) = @_;
332 my $doc = $smsg->{doc};
333 my $mid = mid_compressed($smsg->mid);
334 my $mime = $smsg->mime;
335 my $refs = $mime->header('References');
336 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
337 my $irt = $mime->header('In-Reply-To');
339 if ($irt =~ /<([^>]+)>/) {
347 @refs = map { mid_compressed($_) } @refs;
349 @refs = grep { !$uniq{$_}++ } @refs; # uniq
351 $doc->add_term(xpfx('inreplyto') . $refs[-1]);
353 my $ref_pfx = xpfx('references');
355 # first ref *should* be the thread root,
356 # but we can never trust clients to do the right thing
357 my $ref = shift @refs;
358 $doc->add_term($ref_pfx . $ref);
359 $tid = $self->_resolve_mid_to_tid($ref);
361 # the rest of the refs should point to this tid:
362 foreach $ref (@refs) {
363 $doc->add_term($ref_pfx . $ref);
364 my $ptid = $self->_resolve_mid_to_tid($ref);
366 $self->merge_threads($tid, $ptid);
370 $tid = $self->next_thread_id;
372 $doc->add_term(xpfx('thread') . $tid);
376 my ($self, $mid) = @_;
377 $mid = mid_clean($mid);
378 $mid = mid_compressed($mid);
380 my $doc_id = $self->find_unique_doc_id('mid', $mid);
382 if (defined $doc_id) {
384 my $doc = $self->{xdb}->get_document($doc_id);
385 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
386 $smsg->doc_id($doc_id);
391 sub find_unique_doc_id {
392 my ($self, $term, $value) = @_;
394 my ($begin, $end) = $self->find_doc_ids($term, $value);
396 return undef if $begin->equal($end); # not found
398 my $rv = $begin->get_docid;
402 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
406 # returns begin and end PostingIterator
408 my ($self, $term, $value) = @_;
410 $self->find_doc_ids_for_term(xpfx($term) . $value);
413 # returns begin and end PostingIterator
414 sub find_doc_ids_for_term {
415 my ($self, $term) = @_;
416 my $db = $self->{xdb};
418 ($db->postlist_begin($term), $db->postlist_end($term));
421 # this will create a ghost as necessary
422 sub _resolve_mid_to_tid {
423 my ($self, $mid) = @_;
425 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
430 my ($self, $mid, $tid) = @_;
432 $mid = mid_compressed($mid);
433 $tid = $self->next_thread_id unless defined $tid;
435 my $doc = Search::Xapian::Document->new;
436 $doc->add_term(xpfx('mid') . $mid);
437 $doc->add_term(xpfx('thread') . $tid);
438 $doc->add_term(xpfx('type') . 'ghost');
440 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
441 $self->link_message($smsg, 1);
442 $self->{xdb}->add_document($doc);
448 my ($self, $winner_tid, $loser_tid) = @_;
449 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
450 my $thread_pfx = xpfx('thread');
451 my $db = $self->{xdb};
453 for (; $head != $tail; $head->inc) {
454 my $docid = $head->get_docid;
455 my $doc = $db->get_document($docid);
456 $doc->remove_term($thread_pfx . $loser_tid);
457 $doc->add_term($thread_pfx . $winner_tid);
458 $db->replace_document($docid, $doc);
462 # normalize subjects so they are suitable as pathnames for URLs
468 $subj =~ s/^(?:re|aw):\s*//i; # remove reply prefix (aw: German)
469 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
474 my ($git, $blob) = @_;
476 my $str = $git->cat_file($blob);
477 Email::MIME->new($str);
483 my ($self, $git, $blob) = @_;
484 my $mime = do_cat_mail($git, $blob) or return;
485 eval { $self->add_message($mime) };
486 warn "W: index_blob $blob: $@\n" if $@;
490 my ($self, $git, $blob) = @_;
491 my $mime = do_cat_mail($git, $blob) or return;
492 my $mid = $mime->header('Message-ID');
493 eval { $self->remove_message($mid) } if defined $mid;
494 warn "W: unindex_blob $blob: $@\n" if $@;
499 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
502 # indexes all unindexed messages
504 my ($self, $git) = @_;
505 my $db = $self->{xdb};
506 my $latest = $db->get_metadata('last_commit');
507 my $range = length $latest ? "$latest..HEAD" : 'HEAD';
510 my $hex = '[a-f0-9]';
511 my $h40 = $hex .'{40}';
512 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
513 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
515 # get indexed messages
516 my @cmd = ('git', "--git-dir=$git->{git_dir}", "log",
517 qw/--reverse --no-notes --no-color --raw -r --no-abbrev/,
520 my $pid = open(my $log, '-|', @cmd) or
521 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
523 while (my $line = <$log>) {
524 if ($line =~ /$addmsg/o) {
525 $self->index_blob($git, $1);
526 } elsif ($line =~ /$delmsg/o) {
527 $self->unindex_blob($git, $1);
528 } elsif ($line =~ /^commit ($h40)/o) {
530 if (defined $latest) {
531 $db->set_metadata('last_commit', $latest)
537 $db->set_metadata('last_commit', $latest) if defined $latest;