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/;
13 # This is English-only, everything else is non-standard and may be confused as
14 # a prefix common in patch emails
15 our $REPLY_RE = qr/^re:\s+/i;
16 our $LANG = 'english';
20 # SCHEMA_VERSION history
22 # 1 - subject_path is lower-cased
23 # 2 - subject_path is mid_compressed in the index, only
24 # 3 - message-ID is compressed if it includes '%' (hack!)
25 # 4 - change "Re: " normalization, avoid circular Reference ghosts
27 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
31 my %bool_pfx_internal = (
32 type => 'T', # "mail" or "ghost"
33 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
36 my %bool_pfx_external = (
38 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
39 references => 'XREFS',
47 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
49 sub xpfx { $all_pfx{$_[0]} }
52 while (my ($k, $v) = each %all_pfx) {
53 next if $prob_prefix{$k};
54 $PFX2TERM_RMAP{$v} = $k;
57 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
60 my ($class, $git_dir, $writable) = @_;
61 # allow concurrent versions for easier rollback:
62 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
65 if ($writable) { # not used by the WWW interface
66 require Search::Xapian::WritableDatabase;
67 my $flag = Search::Xapian::DB_OPEN;
70 File::Path::mkpath($dir);
71 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
73 $db = Search::Xapian::WritableDatabase->new($dir, $flag);
75 $db = Search::Xapian::Database->new($dir);
77 bless { xdb => $db, git_dir => $git_dir }, $class;
80 sub reopen { $_[0]->{xdb}->reopen }
83 my ($self, $mime) = @_; # mime = Email::MIME object
84 my $db = $self->{xdb};
87 my $mid_orig = mid_clean($mime->header_obj->header_raw('Message-ID'));
88 my $mid = mid_compressed($mid_orig);
90 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
91 my $enc_msg = PublicInbox::View::enc_for($ct_msg);
94 my $smsg = $self->lookup_message($mid);
98 $smsg->ensure_metadata;
99 # convert a ghost to a regular message
100 # it will also clobber any existing regular message
104 my $type = xpfx('type');
106 $doc->remove_term($type . 'ghost');
110 # probably does not exist:
111 eval { $doc->remove_term($type . 'mail') };
112 $doc->add_term($type . 'mail');
114 $smsg = PublicInbox::SearchMsg->new($mime);
116 $doc->add_term(xpfx('mid') . $mid);
119 my $subj = $smsg->subject;
122 $doc->add_term(xpfx('subject') . $subj);
124 my $path = subject_path($subj);
125 $doc->add_term(xpfx('path') . mid_compressed($path));
128 my $from = $smsg->from_name;
129 my $date = $smsg->date;
130 my $ts = Search::Xapian::sortable_serialise($smsg->ts);
131 $doc->add_value(PublicInbox::Search::TS, $ts);
133 $doc->set_data($smsg->to_doc_data);
135 my $tg = $self->term_generator;
137 $tg->set_document($doc);
138 $tg->index_text($subj, 1, 'S') if $subj;
139 $tg->increase_termpos;
140 $tg->index_text($subj) if $subj;
141 $tg->increase_termpos;
143 $tg->index_text($smsg->from->format);
144 $tg->increase_termpos;
146 $mime->walk_parts(sub {
148 return if $part->subparts; # walk_parts already recurses
149 my $ct = $part->content_type || $ct_msg;
151 # account for filter bugs...
152 $ct =~ m!\btext/plain\b!i or return;
154 my $enc = PublicInbox::View::enc_for($ct, $enc_msg);
156 foreach my $l (split(/\n/, $enc->decode($part->body))) {
164 $tg->index_text(join("\n", @quot), 0);
165 $tg->increase_termpos;
168 $tg->index_text(join("\n", @orig));
169 $tg->increase_termpos;
174 $doc_id = $smsg->doc_id;
175 $self->link_message($smsg, 0);
176 $db->replace_document($doc_id, $doc);
178 $self->link_message($smsg, 0);
179 $doc_id = $db->add_document($doc);
184 warn "failed to index message <$mid_orig>: $@\n";
190 # returns deleted doc_id on success, undef on missing
192 my ($self, $mid_orig) = @_;
193 my $db = $self->{xdb};
195 $mid_orig = mid_clean($mid_orig);
196 my $mid = mid_compressed($mid_orig);
199 $doc_id = $self->find_unique_doc_id('mid', $mid);
200 $db->delete_document($doc_id) if defined $doc_id;
204 warn "failed to remove message <$mid_orig>: $@\n";
212 my ($self, $query_string, $opts) = @_;
213 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
215 $query = Search::Xapian::Query->new(OP_AND, $mail_query, $query);
216 $self->do_enquire($query, $opts);
219 sub get_subject_path {
220 my ($self, $path, $opts) = @_;
221 my $query = $self->qp->parse_query("path:".mid_compressed($path), 0);
222 $self->do_enquire($query);
225 # given a message ID, get followups to a message
227 my ($self, $mid, $opts) = @_;
228 $mid = mid_clean($mid);
229 $mid = mid_compressed($mid);
231 my $irt = $qp->parse_query("inreplyto:$mid", 0);
232 my $ref = $qp->parse_query("references:$mid", 0);
233 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
235 $self->do_enquire($query);
239 my ($self, $mid, $opts) = @_;
240 my $smsg = eval { $self->lookup_message($mid) };
242 return { count => 0, msgs => [] } unless $smsg;
244 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
245 my $qsub = $qp->parse_query('path:'.mid_compressed($smsg->path));
246 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
247 $self->do_enquire($query);
253 my ($self, $query, $opts) = @_;
254 my $enquire = $self->enquire;
256 $enquire->set_query($query);
257 $enquire->set_sort_by_relevance_then_value(TS, 0);
259 my $offset = $opts->{offset} || 0;
260 my $limit = $opts->{limit} || 50;
261 my $mset = $enquire->get_mset($offset, $limit);
263 PublicInbox::SearchMsg->load_doc($_->get_document);
266 { count => $mset->get_matches_estimated, msgs => \@msgs }
270 sub stemmer { Search::Xapian::Stem->new($LANG) }
276 my $qp = $self->{query_parser};
280 $qp = Search::Xapian::QueryParser->new;
281 $qp->set_default_op(OP_AND);
282 $qp->set_database($self->{xdb});
283 $qp->set_stemmer($self->stemmer);
284 $qp->set_stemming_strategy(STEM_SOME);
285 $qp->add_valuerangeprocessor($self->ts_range_processor);
286 $qp->add_valuerangeprocessor($self->date_range_processor);
288 while (my ($name, $prefix) = each %bool_pfx_external) {
289 $qp->add_boolean_prefix($name, $prefix);
292 while (my ($name, $prefix) = each %prob_prefix) {
293 $qp->add_prefix($name, $prefix);
296 $self->{query_parser} = $qp;
299 sub term_generator { # write-only
302 my $tg = $self->{term_generator};
305 $tg = Search::Xapian::TermGenerator->new;
306 $tg->set_stemmer($self->stemmer);
308 $self->{term_generator} = $tg;
311 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
313 # increments last_thread_id counter
314 # returns a 64-bit integer represented as a hex string
317 my $db = $self->{xdb};
318 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
320 $db->set_metadata('last_thread_id', ++$last_thread_id);
325 sub ts_range_processor {
326 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
329 sub date_range_processor {
330 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
334 my ($self, $smsg, $is_ghost) = @_;
337 $smsg->ensure_metadata;
339 $self->link_message_to_parents($smsg);
343 sub link_message_to_parents {
344 my ($self, $smsg) = @_;
345 my $doc = $smsg->{doc};
346 my $mid = mid_compressed($smsg->mid);
347 my $mime = $smsg->mime;
348 my $refs = $mime->header_obj->header_raw('References');
349 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
350 my $irt = $mime->header_obj->header_raw('In-Reply-To');
352 if ($irt =~ /<([^>]+)>/) {
356 # maybe some crazies will try to make a circular reference:
366 my @crefs = map { mid_compressed($_) } @refs;
367 my %uniq = ($mid => 1);
369 # prevent circular references via References: here:
371 foreach my $ref (@crefs) {
376 $irt = undef if (defined $irt && !$uniq{$irt});
380 $doc->add_term(xpfx('inreplyto') . $irt);
383 my $ref_pfx = xpfx('references');
385 # first ref *should* be the thread root,
386 # but we can never trust clients to do the right thing
387 my $ref = shift @refs;
388 $doc->add_term($ref_pfx . $ref);
389 $tid = $self->_resolve_mid_to_tid($ref);
391 # the rest of the refs should point to this tid:
392 foreach $ref (@refs) {
393 $doc->add_term($ref_pfx . $ref);
394 my $ptid = $self->_resolve_mid_to_tid($ref);
396 $self->merge_threads($tid, $ptid);
400 $tid = $self->next_thread_id;
402 $doc->add_term(xpfx('thread') . $tid);
406 my ($self, $mid) = @_;
407 $mid = mid_clean($mid);
408 $mid = mid_compressed($mid);
410 my $doc_id = $self->find_unique_doc_id('mid', $mid);
412 if (defined $doc_id) {
414 my $doc = $self->{xdb}->get_document($doc_id);
415 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
416 $smsg->doc_id($doc_id);
421 sub find_unique_doc_id {
422 my ($self, $term, $value) = @_;
424 my ($begin, $end) = $self->find_doc_ids($term, $value);
426 return undef if $begin->equal($end); # not found
428 my $rv = $begin->get_docid;
432 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
436 # returns begin and end PostingIterator
438 my ($self, $term, $value) = @_;
440 $self->find_doc_ids_for_term(xpfx($term) . $value);
443 # returns begin and end PostingIterator
444 sub find_doc_ids_for_term {
445 my ($self, $term) = @_;
446 my $db = $self->{xdb};
448 ($db->postlist_begin($term), $db->postlist_end($term));
451 # this will create a ghost as necessary
452 sub _resolve_mid_to_tid {
453 my ($self, $mid) = @_;
455 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
460 my ($self, $mid, $tid) = @_;
462 $mid = mid_compressed($mid);
463 $tid = $self->next_thread_id unless defined $tid;
465 my $doc = Search::Xapian::Document->new;
466 $doc->add_term(xpfx('mid') . $mid);
467 $doc->add_term(xpfx('thread') . $tid);
468 $doc->add_term(xpfx('type') . 'ghost');
470 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
471 $self->link_message($smsg, 1);
472 $self->{xdb}->add_document($doc);
478 my ($self, $winner_tid, $loser_tid) = @_;
479 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
480 my $thread_pfx = xpfx('thread');
481 my $db = $self->{xdb};
483 for (; $head != $tail; $head->inc) {
484 my $docid = $head->get_docid;
485 my $doc = $db->get_document($docid);
486 $doc->remove_term($thread_pfx . $loser_tid);
487 $doc->add_term($thread_pfx . $winner_tid);
488 $db->replace_document($docid, $doc);
492 # normalize subjects so they are suitable as pathnames for URLs
498 $subj =~ s/$REPLY_RE//igo; # remove reply prefix
499 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
504 my ($git, $blob) = @_;
506 my $str = $git->cat_file($blob);
507 Email::MIME->new($str);
513 my ($self, $git, $blob) = @_;
514 my $mime = do_cat_mail($git, $blob) or return;
515 eval { $self->add_message($mime) };
516 warn "W: index_blob $blob: $@\n" if $@;
520 my ($self, $git, $blob) = @_;
521 my $mime = do_cat_mail($git, $blob) or return;
522 my $mid = $mime->header_obj->header_raw('Message-ID');
523 eval { $self->remove_message($mid) } if defined $mid;
524 warn "W: unindex_blob $blob: $@\n" if $@;
529 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
532 # indexes all unindexed messages
535 require PublicInbox::GitCatFile;
536 my $db = $self->{xdb};
537 my $hex = '[a-f0-9]';
538 my $h40 = $hex .'{40}';
539 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
540 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
542 $db->begin_transaction;
544 my $git = PublicInbox::GitCatFile->new($self->{git_dir});
546 my $latest = $db->get_metadata('last_commit');
547 my $range = length $latest ? "$latest..HEAD" : 'HEAD';
550 # get indexed messages
551 my @cmd = ('git', "--git-dir=$self->{git_dir}", "log",
552 qw/--reverse --no-notes --no-color --raw -r
553 --no-abbrev/, $range);
554 my $pid = open(my $log, '-|', @cmd) or
555 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
557 while (my $line = <$log>) {
558 if ($line =~ /$addmsg/o) {
559 $self->index_blob($git, $1);
560 } elsif ($line =~ /$delmsg/o) {
561 $self->unindex_blob($git, $1);
562 } elsif ($line =~ /^commit ($h40)/o) {
567 $db->set_metadata('last_commit', $latest) if defined $latest;
570 warn "indexing failed: $@\n";
571 $db->cancel_transaction;
573 $db->commit_transaction;