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/;
15 # SCHEMA_VERSION history
17 # 1 - subject_path is lower-cased
18 # 2 - subject_path is mid_compressed in the index, only
19 # 3 - message-ID is compressed if it includes '%' (hack!)
22 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
26 my %bool_pfx_internal = (
27 type => 'T', # "mail" or "ghost"
28 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
31 my %bool_pfx_external = (
33 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
34 references => 'XREFS',
42 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
44 sub xpfx { $all_pfx{$_[0]} }
47 while (my ($k, $v) = each %all_pfx) {
48 next if $prob_prefix{$k};
49 $PFX2TERM_RMAP{$v} = $k;
52 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
55 my ($class, $git_dir, $writable) = @_;
56 # allow concurrent versions for easier rollback:
57 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
60 if ($writable) { # not used by the WWW interface
61 require Search::Xapian::WritableDatabase;
62 my $flag = Search::Xapian::DB_OPEN;
65 File::Path::mkpath($dir);
66 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
68 $db = Search::Xapian::WritableDatabase->new($dir, $flag);
70 $db = Search::Xapian::Database->new($dir);
72 bless { xdb => $db, git_dir => $git_dir }, $class;
75 sub reopen { $_[0]->{xdb}->reopen }
78 my ($self, $mime) = @_; # mime = Email::MIME object
79 my $db = $self->{xdb};
82 my $mid_orig = mid_clean($mime->header_obj->header_raw('Message-ID'));
83 my $mid = mid_compressed($mid_orig);
85 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
86 my $enc_msg = PublicInbox::View::enc_for($ct_msg);
89 my $smsg = $self->lookup_message($mid);
93 $smsg->ensure_metadata;
94 # convert a ghost to a regular message
95 # it will also clobber any existing regular message
99 my $type = xpfx('type');
101 $doc->remove_term($type . 'ghost');
105 # probably does not exist:
106 eval { $doc->remove_term($type . 'mail') };
107 $doc->add_term($type . 'mail');
109 $smsg = PublicInbox::SearchMsg->new($mime);
111 $doc->add_term(xpfx('mid') . $mid);
114 my $subj = $smsg->subject;
117 $doc->add_term(xpfx('subject') . $subj);
119 my $path = subject_path($subj);
120 $doc->add_term(xpfx('path') . mid_compressed($path));
123 my $from = $smsg->from_name;
124 my $date = $smsg->date;
125 my $ts = Search::Xapian::sortable_serialise($smsg->ts);
126 $doc->add_value(PublicInbox::Search::TS, $ts);
128 $doc->set_data($smsg->to_doc_data);
130 my $tg = $self->term_generator;
132 $tg->set_document($doc);
133 $tg->index_text($subj, 1, 'S') if $subj;
134 $tg->increase_termpos;
135 $tg->index_text($subj) if $subj;
136 $tg->increase_termpos;
138 $tg->index_text($smsg->from->format);
139 $tg->increase_termpos;
141 $mime->walk_parts(sub {
143 return if $part->subparts; # walk_parts already recurses
144 my $ct = $part->content_type || $ct_msg;
146 # account for filter bugs...
147 $ct =~ m!\btext/plain\b!i or return;
149 my $enc = PublicInbox::View::enc_for($ct, $enc_msg);
151 foreach my $l (split(/\n/, $enc->decode($part->body))) {
159 $tg->index_text(join("\n", @quot), 0);
160 $tg->increase_termpos;
163 $tg->index_text(join("\n", @orig));
164 $tg->increase_termpos;
169 $doc_id = $smsg->doc_id;
170 $self->link_message($smsg, 0);
171 $db->replace_document($doc_id, $doc);
173 $self->link_message($smsg, 0);
174 $doc_id = $db->add_document($doc);
179 warn "failed to index message <$mid_orig>: $@\n";
185 # returns deleted doc_id on success, undef on missing
187 my ($self, $mid_orig) = @_;
188 my $db = $self->{xdb};
190 $mid_orig = mid_clean($mid_orig);
191 my $mid = mid_compressed($mid_orig);
194 $doc_id = $self->find_unique_doc_id('mid', $mid);
195 $db->delete_document($doc_id) if defined $doc_id;
199 warn "failed to remove message <$mid_orig>: $@\n";
207 my ($self, $query_string, $opts) = @_;
208 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
210 $query = Search::Xapian::Query->new(OP_AND, $mail_query, $query);
211 $self->do_enquire($query, $opts);
214 sub get_subject_path {
215 my ($self, $path, $opts) = @_;
216 my $query = $self->qp->parse_query("path:".mid_compressed($path), 0);
217 $self->do_enquire($query);
220 # given a message ID, get followups to a message
222 my ($self, $mid, $opts) = @_;
223 $mid = mid_clean($mid);
224 $mid = mid_compressed($mid);
226 my $irt = $qp->parse_query("inreplyto:$mid", 0);
227 my $ref = $qp->parse_query("references:$mid", 0);
228 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
230 $self->do_enquire($query);
234 my ($self, $mid, $opts) = @_;
235 my $smsg = eval { $self->lookup_message($mid) };
237 return { count => 0, msgs => [] } unless $smsg;
239 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
240 my $qsub = $qp->parse_query('path:'.mid_compressed($smsg->path));
241 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
242 $self->do_enquire($query);
248 my ($self, $query, $opts) = @_;
249 my $enquire = $self->enquire;
251 $enquire->set_query($query);
252 $enquire->set_sort_by_relevance_then_value(TS, 0);
254 my $offset = $opts->{offset} || 0;
255 my $limit = $opts->{limit} || 50;
256 my $mset = $enquire->get_mset($offset, $limit);
258 PublicInbox::SearchMsg->load_doc($_->get_document);
261 { count => $mset->get_matches_estimated, msgs => \@msgs }
265 sub stemmer { Search::Xapian::Stem->new(LANG) }
271 my $qp = $self->{query_parser};
275 $qp = Search::Xapian::QueryParser->new;
276 $qp->set_default_op(OP_AND);
277 $qp->set_database($self->{xdb});
278 $qp->set_stemmer($self->stemmer);
279 $qp->set_stemming_strategy(STEM_SOME);
280 $qp->add_valuerangeprocessor($self->ts_range_processor);
281 $qp->add_valuerangeprocessor($self->date_range_processor);
283 while (my ($name, $prefix) = each %bool_pfx_external) {
284 $qp->add_boolean_prefix($name, $prefix);
287 while (my ($name, $prefix) = each %prob_prefix) {
288 $qp->add_prefix($name, $prefix);
291 $self->{query_parser} = $qp;
294 sub term_generator { # write-only
297 my $tg = $self->{term_generator};
300 $tg = Search::Xapian::TermGenerator->new;
301 $tg->set_stemmer($self->stemmer);
303 $self->{term_generator} = $tg;
306 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
308 # increments last_thread_id counter
309 # returns a 64-bit integer represented as a hex string
312 my $db = $self->{xdb};
313 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
315 $db->set_metadata('last_thread_id', ++$last_thread_id);
320 sub ts_range_processor {
321 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
324 sub date_range_processor {
325 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
329 my ($self, $smsg, $is_ghost) = @_;
332 $smsg->ensure_metadata;
334 $self->link_message_to_parents($smsg);
338 sub link_message_to_parents {
339 my ($self, $smsg) = @_;
340 my $doc = $smsg->{doc};
341 my $mid = mid_compressed($smsg->mid);
342 my $mime = $smsg->mime;
343 my $refs = $mime->header_obj->header_raw('References');
344 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
345 my $irt = $mime->header_obj->header_raw('In-Reply-To');
347 if ($irt =~ /<([^>]+)>/) {
351 # maybe some crazies will try to make a circular reference:
361 my @crefs = map { mid_compressed($_) } @refs;
362 my %uniq = ($mid => 1);
364 # prevent circular references via References: here:
366 foreach my $ref (@crefs) {
371 $irt = undef if (defined $irt && !$uniq{$irt});
375 $doc->add_term(xpfx('inreplyto') . $irt);
378 my $ref_pfx = xpfx('references');
380 # first ref *should* be the thread root,
381 # but we can never trust clients to do the right thing
382 my $ref = shift @refs;
383 $doc->add_term($ref_pfx . $ref);
384 $tid = $self->_resolve_mid_to_tid($ref);
386 # the rest of the refs should point to this tid:
387 foreach $ref (@refs) {
388 $doc->add_term($ref_pfx . $ref);
389 my $ptid = $self->_resolve_mid_to_tid($ref);
391 $self->merge_threads($tid, $ptid);
395 $tid = $self->next_thread_id;
397 $doc->add_term(xpfx('thread') . $tid);
401 my ($self, $mid) = @_;
402 $mid = mid_clean($mid);
403 $mid = mid_compressed($mid);
405 my $doc_id = $self->find_unique_doc_id('mid', $mid);
407 if (defined $doc_id) {
409 my $doc = $self->{xdb}->get_document($doc_id);
410 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
411 $smsg->doc_id($doc_id);
416 sub find_unique_doc_id {
417 my ($self, $term, $value) = @_;
419 my ($begin, $end) = $self->find_doc_ids($term, $value);
421 return undef if $begin->equal($end); # not found
423 my $rv = $begin->get_docid;
427 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
431 # returns begin and end PostingIterator
433 my ($self, $term, $value) = @_;
435 $self->find_doc_ids_for_term(xpfx($term) . $value);
438 # returns begin and end PostingIterator
439 sub find_doc_ids_for_term {
440 my ($self, $term) = @_;
441 my $db = $self->{xdb};
443 ($db->postlist_begin($term), $db->postlist_end($term));
446 # this will create a ghost as necessary
447 sub _resolve_mid_to_tid {
448 my ($self, $mid) = @_;
450 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
455 my ($self, $mid, $tid) = @_;
457 $mid = mid_compressed($mid);
458 $tid = $self->next_thread_id unless defined $tid;
460 my $doc = Search::Xapian::Document->new;
461 $doc->add_term(xpfx('mid') . $mid);
462 $doc->add_term(xpfx('thread') . $tid);
463 $doc->add_term(xpfx('type') . 'ghost');
465 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
466 $self->link_message($smsg, 1);
467 $self->{xdb}->add_document($doc);
473 my ($self, $winner_tid, $loser_tid) = @_;
474 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
475 my $thread_pfx = xpfx('thread');
476 my $db = $self->{xdb};
478 for (; $head != $tail; $head->inc) {
479 my $docid = $head->get_docid;
480 my $doc = $db->get_document($docid);
481 $doc->remove_term($thread_pfx . $loser_tid);
482 $doc->add_term($thread_pfx . $winner_tid);
483 $db->replace_document($docid, $doc);
487 # normalize subjects so they are suitable as pathnames for URLs
493 $subj =~ s/^(?:re|aw):\s*//i; # remove reply prefix (aw: German)
494 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
499 my ($git, $blob) = @_;
501 my $str = $git->cat_file($blob);
502 Email::MIME->new($str);
508 my ($self, $git, $blob) = @_;
509 my $mime = do_cat_mail($git, $blob) or return;
510 eval { $self->add_message($mime) };
511 warn "W: index_blob $blob: $@\n" if $@;
515 my ($self, $git, $blob) = @_;
516 my $mime = do_cat_mail($git, $blob) or return;
517 my $mid = $mime->header_obj->header_raw('Message-ID');
518 eval { $self->remove_message($mid) } if defined $mid;
519 warn "W: unindex_blob $blob: $@\n" if $@;
524 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
527 # indexes all unindexed messages
530 require PublicInbox::GitCatFile;
531 my $db = $self->{xdb};
532 my $hex = '[a-f0-9]';
533 my $h40 = $hex .'{40}';
534 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
535 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
537 $db->begin_transaction;
539 my $git = PublicInbox::GitCatFile->new($self->{git_dir});
541 my $latest = $db->get_metadata('last_commit');
542 my $range = length $latest ? "$latest..HEAD" : 'HEAD';
545 # get indexed messages
546 my @cmd = ('git', "--git-dir=$self->{git_dir}", "log",
547 qw/--reverse --no-notes --no-color --raw -r
548 --no-abbrev/, $range);
549 my $pid = open(my $log, '-|', @cmd) or
550 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
552 while (my $line = <$log>) {
553 if ($line =~ /$addmsg/o) {
554 $self->index_blob($git, $1);
555 } elsif ($line =~ /$delmsg/o) {
556 $self->unindex_blob($git, $1);
557 } elsif ($line =~ /^commit ($h40)/o) {
562 $db->set_metadata('last_commit', $latest) if defined $latest;
565 warn "indexing failed: $@\n";
566 $db->cancel_transaction;
568 $db->commit_transaction;