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;
9 use Search::Xapian qw/:standard/;
10 require PublicInbox::View;
12 use PublicInbox::MID qw/mid_clean mid_compressed/;
14 our @EXPORT = qw/xpfx/;
20 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
24 my %bool_pfx_internal = (
25 type => 'T', # "mail" or "ghost"
26 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
29 my %bool_pfx_external = (
31 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
32 references => 'XREFS',
40 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
42 sub xpfx { $all_pfx{$_[0]} }
45 while (my ($k, $v) = each %all_pfx) {
46 next if $prob_prefix{$k};
47 $PFX2TERM_RMAP{$v} = $k;
50 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
53 my ($class, $git_dir, $writable) = @_;
54 # allow concurrent versions for easier rollback:
55 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
58 if ($writable) { # not used by the WWW interface
59 require Search::Xapian::WritableDatabase;
61 File::Path::mkpath($dir);
62 $db = Search::Xapian::WritableDatabase->new($dir,
63 Search::Xapian::DB_CREATE_OR_OPEN);
65 $db = Search::Xapian::Database->new($dir);
67 bless { xdb => $db, git_dir => $git_dir }, $class;
70 sub reopen { $_[0]->{xdb}->reopen }
73 my ($self, $mime) = @_; # mime = Email::MIME object
74 my $db = $self->{xdb};
77 my $mid = mid_clean($mime->header('Message-ID'));
78 $mid = mid_compressed($mid);
80 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
81 my $enc_msg = PublicInbox::View::enc_for($ct_msg);
83 $db->begin_transaction;
85 my $smsg = $self->lookup_message($mid);
89 $smsg->ensure_metadata;
90 # convert a ghost to a regular message
91 # it will also clobber any existing regular message
95 my $type = xpfx('type');
97 $doc->remove_term($type . 'ghost');
101 # probably does not exist:
102 eval { $doc->remove_term($type . 'mail') };
103 $doc->add_term($type . 'mail');
105 $smsg = PublicInbox::SearchMsg->new($mime);
107 $doc->add_term(xpfx('mid') . $mid);
110 my $subj = $smsg->subject;
113 $doc->add_term(xpfx('subject') . $subj);
115 my $path = subject_path($subj);
116 $doc->add_term(xpfx('path') . $path);
119 my $from = $smsg->from_name;
120 my $date = $smsg->date;
121 my $ts = Search::Xapian::sortable_serialise($smsg->ts);
122 $doc->add_value(PublicInbox::Search::TS, $ts);
124 $doc->set_data($smsg->to_doc_data);
126 my $tg = $self->term_generator;
128 $tg->set_document($doc);
129 $tg->index_text($subj, 1, 'S') if $subj;
130 $tg->increase_termpos;
131 $tg->index_text($subj) if $subj;
132 $tg->increase_termpos;
134 $tg->index_text($smsg->from->format);
135 $tg->increase_termpos;
137 $mime->walk_parts(sub {
139 return if $part->subparts; # walk_parts already recurses
140 my $ct = $part->content_type || $ct_msg;
142 # account for filter bugs...
143 $ct =~ m!\btext/plain\b!i or return;
145 my $enc = PublicInbox::View::enc_for($ct, $enc_msg);
147 foreach my $l (split(/\n/, $enc->decode($part->body))) {
155 $tg->index_text(join("\n", @quot), 0);
156 $tg->increase_termpos;
159 $tg->index_text(join("\n", @orig));
160 $tg->increase_termpos;
165 $doc_id = $smsg->doc_id;
166 $self->link_message($smsg, 0);
167 $db->replace_document($doc_id, $doc);
169 $self->link_message($smsg, 0);
170 $doc_id = $db->add_document($doc);
175 warn "failed to index message <$mid>: $@\n";
176 $db->cancel_transaction;
178 $db->commit_transaction;
183 # returns deleted doc_id on success, undef on missing
185 my ($self, $mid) = @_;
186 my $db = $self->{xdb};
188 $mid = mid_clean($mid);
189 $mid = mid_compressed($mid);
191 $db->begin_transaction;
193 $doc_id = $self->find_unique_doc_id('mid', $mid);
194 $db->delete_document($doc_id) if defined $doc_id;
198 warn "failed to remove message <$mid>: $@\n";
199 $db->cancel_transaction;
201 $db->commit_transaction;
208 my ($self, $query_string, $opts) = @_;
209 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
211 $query = Search::Xapian::Query->new(OP_AND, $mail_query, $query);
212 $self->do_enquire($query, $opts);
215 # given a message ID, get replies to a message
217 my ($self, $mid, $opts) = @_;
218 $mid = mid_clean($mid);
219 $mid = mid_compressed($mid);
221 my $irt = $qp->parse_query("inreplyto:$mid", 0);
222 my $ref = $qp->parse_query("references:$mid", 0);
223 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
225 $self->do_enquire($query);
229 my ($self, $mid, $opts) = @_;
230 my $smsg = eval { $self->lookup_message($mid) };
232 return { count => 0, msgs => [] } unless $smsg;
234 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
235 my $qsub = $qp->parse_query('path:'.$smsg->path);
236 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
237 $self->do_enquire($query);
243 my ($self, $query, $opts) = @_;
244 my $enquire = $self->enquire;
246 $enquire->set_query($query);
247 $enquire->set_sort_by_relevance_then_value(TS, 0);
249 my $offset = $opts->{offset} || 0;
250 my $limit = $opts->{limit} || 50;
251 my $mset = $enquire->get_mset($offset, $limit);
253 PublicInbox::SearchMsg->load_doc($_->get_document);
256 { count => $mset->get_matches_estimated, msgs => \@msgs }
260 sub stemmer { Search::Xapian::Stem->new(LANG) }
266 my $qp = $self->{query_parser};
270 $qp = Search::Xapian::QueryParser->new;
271 $qp->set_default_op(OP_AND);
272 $qp->set_database($self->{xdb});
273 $qp->set_stemmer($self->stemmer);
274 $qp->set_stemming_strategy(STEM_SOME);
275 $qp->add_valuerangeprocessor($self->ts_range_processor);
276 $qp->add_valuerangeprocessor($self->date_range_processor);
278 while (my ($name, $prefix) = each %bool_pfx_external) {
279 $qp->add_boolean_prefix($name, $prefix);
282 while (my ($name, $prefix) = each %prob_prefix) {
283 $qp->add_prefix($name, $prefix);
286 $self->{query_parser} = $qp;
289 sub term_generator { # write-only
292 my $tg = $self->{term_generator};
295 $tg = Search::Xapian::TermGenerator->new;
296 $tg->set_stemmer($self->stemmer);
298 $self->{term_generator} = $tg;
301 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
303 # increments last_thread_id counter
304 # returns a 64-bit integer represented as a hex string
307 my $db = $self->{xdb};
308 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
310 $db->set_metadata('last_thread_id', ++$last_thread_id);
315 sub ts_range_processor {
316 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
319 sub date_range_processor {
320 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
324 my ($self, $smsg, $is_ghost) = @_;
327 $smsg->ensure_metadata;
329 $self->link_message_to_parents($smsg);
333 sub link_message_to_parents {
334 my ($self, $smsg) = @_;
335 my $doc = $smsg->{doc};
336 my $mid = mid_compressed($smsg->mid);
337 my $mime = $smsg->mime;
338 my $refs = $mime->header('References');
339 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
340 my $irt = $mime->header('In-Reply-To');
342 if ($irt =~ /<([^>]+)>/) {
350 @refs = map { mid_compressed($_) } @refs;
352 @refs = grep { !$uniq{$_}++ } @refs; # uniq
354 $doc->add_term(xpfx('inreplyto') . $refs[-1]);
356 my $ref_pfx = xpfx('references');
358 # first ref *should* be the thread root,
359 # but we can never trust clients to do the right thing
360 my $ref = shift @refs;
361 $doc->add_term($ref_pfx . $ref);
362 $tid = $self->_resolve_mid_to_tid($ref);
364 # the rest of the refs should point to this tid:
365 foreach $ref (@refs) {
366 $doc->add_term($ref_pfx . $ref);
367 my $ptid = $self->_resolve_mid_to_tid($ref);
369 $self->merge_threads($tid, $ptid);
373 $tid = $self->next_thread_id;
375 $doc->add_term(xpfx('thread') . $tid);
379 my ($self, $mid) = @_;
380 $mid = mid_clean($mid);
381 $mid = mid_compressed($mid);
383 my $doc_id = $self->find_unique_doc_id('mid', $mid);
385 if (defined $doc_id) {
387 my $doc = $self->{xdb}->get_document($doc_id);
388 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
389 $smsg->doc_id($doc_id);
394 sub find_unique_doc_id {
395 my ($self, $term, $value) = @_;
397 my ($begin, $end) = $self->find_doc_ids($term, $value);
399 return undef if $begin->equal($end); # not found
401 my $rv = $begin->get_docid;
405 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
409 # returns begin and end PostingIterator
411 my ($self, $term, $value) = @_;
413 $self->find_doc_ids_for_term(xpfx($term) . $value);
416 # returns begin and end PostingIterator
417 sub find_doc_ids_for_term {
418 my ($self, $term) = @_;
419 my $db = $self->{xdb};
421 ($db->postlist_begin($term), $db->postlist_end($term));
424 # this will create a ghost as necessary
425 sub _resolve_mid_to_tid {
426 my ($self, $mid) = @_;
428 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
433 my ($self, $mid, $tid) = @_;
435 $mid = mid_compressed($mid);
436 $tid = $self->next_thread_id unless defined $tid;
438 my $doc = Search::Xapian::Document->new;
439 $doc->add_term(xpfx('mid') . $mid);
440 $doc->add_term(xpfx('thread') . $tid);
441 $doc->add_term(xpfx('type') . 'ghost');
443 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
444 $self->link_message($smsg, 1);
445 $self->{xdb}->add_document($doc);
451 my ($self, $winner_tid, $loser_tid) = @_;
452 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
453 my $thread_pfx = xpfx('thread');
454 my $db = $self->{xdb};
456 for (; $head != $tail; $head->inc) {
457 my $docid = $head->get_docid;
458 my $doc = $db->get_document($docid);
459 $doc->remove_term($thread_pfx . $loser_tid);
460 $doc->add_term($thread_pfx . $winner_tid);
461 $db->replace_document($docid, $doc);
465 # normalize subjects so they are suitable as pathnames for URLs
471 $subj =~ s/^(?:re|aw):\s*//i; # remove reply prefix (aw: German)
472 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
477 my ($git, $blob) = @_;
479 my $str = $git->cat_file($blob);
480 Email::MIME->new($str);
486 my ($self, $git, $blob) = @_;
487 my $mime = do_cat_mail($git, $blob) or return;
488 eval { $self->add_message($mime) };
489 warn "W: index_blob $blob: $@\n" if $@;
493 my ($self, $git, $blob) = @_;
494 my $mime = do_cat_mail($git, $blob) or return;
495 my $mid = $mime->header('Message-ID');
496 eval { $self->remove_message($mid) } if defined $mid;
497 warn "W: unindex_blob $blob: $@\n" if $@;
502 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
505 # indexes all unindexed messages
507 my ($self, $git) = @_;
508 my $db = $self->{xdb};
509 my $latest = $db->get_metadata('last_commit');
510 my $range = length $latest ? "$latest..HEAD" : 'HEAD';
513 my $hex = '[a-f0-9]';
514 my $h40 = $hex .'{40}';
515 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
516 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
518 # get indexed messages
519 my @cmd = ('git', "--git-dir=$git->{git_dir}", "log",
520 qw/--reverse --no-notes --no-color --raw -r --no-abbrev/,
523 my $pid = open(my $log, '-|', @cmd) or
524 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
526 while (my $line = <$log>) {
527 if ($line =~ /$addmsg/o) {
528 $self->index_blob($git, $1);
529 } elsif ($line =~ /$delmsg/o) {
530 $self->unindex_blob($git, $1);
531 } elsif ($line =~ /^commit ($h40)/o) {
533 if (defined $latest) {
534 $db->set_metadata('last_commit', $latest)
540 $db->set_metadata('last_commit', $latest) if defined $latest;