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/;
10 use PublicInbox::MID qw/mid_clean mid_compressed/;
12 # This is English-only, everything else is non-standard and may be confused as
13 # a prefix common in patch emails
14 our $REPLY_RE = qr/^re:\s+/i;
15 our $LANG = 'english';
19 # SCHEMA_VERSION history
21 # 1 - subject_path is lower-cased
22 # 2 - subject_path is mid_compressed in the index, only
23 # 3 - message-ID is compressed if it includes '%' (hack!)
24 # 4 - change "Re: " normalization, avoid circular Reference ghosts
25 # 5 - subject_path drops trailing '.'
26 # 6 - preserve References: order in document data
28 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
32 my %bool_pfx_internal = (
33 type => 'T', # "mail" or "ghost"
34 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
37 my %bool_pfx_external = (
39 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
40 references => 'XREFS',
48 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
50 sub xpfx { $all_pfx{$_[0]} }
53 my %meta_pfx = (mid => 1, thread => 1, path => 1, type => 1);
54 while (my ($k, $v) = each %all_pfx) {
55 $PFX2TERM_RMAP{$v} = $k if $meta_pfx{$k};
58 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
61 my ($class, $git_dir, $writable) = @_;
62 # allow concurrent versions for easier rollback:
63 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
66 if ($writable) { # not used by the WWW interface
67 require Search::Xapian::WritableDatabase;
68 my $flag = Search::Xapian::DB_OPEN;
71 File::Path::mkpath($dir);
72 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
74 $db = Search::Xapian::WritableDatabase->new($dir, $flag);
76 $db = Search::Xapian::Database->new($dir);
78 bless { xdb => $db, git_dir => $git_dir }, $class;
81 sub reopen { $_[0]->{xdb}->reopen }
84 my ($self, $mime) = @_; # mime = Email::MIME object
85 my $db = $self->{xdb};
88 my $mid_orig = mid_clean($mime->header_obj->header('Message-ID'));
89 my $mid = mid_compressed($mid_orig);
91 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
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 my $tg = $self->term_generator;
135 $tg->set_document($doc);
136 $tg->index_text($subj, 1, 'S') if $subj;
137 $tg->increase_termpos;
138 $tg->index_text($subj) if $subj;
139 $tg->increase_termpos;
141 $tg->index_text($smsg->from->format);
142 $tg->increase_termpos;
144 $mime->walk_parts(sub {
146 return if $part->subparts; # walk_parts already recurses
147 my $ct = $part->content_type || $ct_msg;
149 # account for filter bugs...
150 $ct =~ m!\btext/plain\b!i or return;
153 my $body = $part->body;
155 my @lines = split(/\n/, $body);
156 while (defined(my $l = shift @lines)) {
164 $tg->index_text(join("\n", @quot), 0);
166 $tg->increase_termpos;
169 $tg->index_text(join("\n", @orig));
171 $tg->increase_termpos;
176 $doc_id = $smsg->doc_id;
177 $self->link_message($smsg, 0);
178 $doc->set_data($smsg->to_doc_data);
179 $db->replace_document($doc_id, $doc);
181 $self->link_message($smsg, 0);
182 $doc->set_data($smsg->to_doc_data);
183 $doc_id = $db->add_document($doc);
188 warn "failed to index message <$mid_orig>: $@\n";
194 # returns deleted doc_id on success, undef on missing
196 my ($self, $mid_orig) = @_;
197 my $db = $self->{xdb};
199 $mid_orig = mid_clean($mid_orig);
200 my $mid = mid_compressed($mid_orig);
203 $doc_id = $self->find_unique_doc_id('mid', $mid);
204 $db->delete_document($doc_id) if defined $doc_id;
208 warn "failed to remove message <$mid_orig>: $@\n";
216 my ($self, $query_string, $opts) = @_;
217 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
219 $self->do_enquire($query, $opts);
222 sub get_subject_path {
223 my ($self, $path, $opts) = @_;
224 my $query = $self->qp->parse_query("path:".mid_compressed($path), 0);
225 $self->do_enquire($query);
228 # given a message ID, get followups to a message
230 my ($self, $mid, $opts) = @_;
231 $mid = mid_clean($mid);
232 $mid = mid_compressed($mid);
234 my $irt = $qp->parse_query("inreplyto:$mid", 0);
235 my $ref = $qp->parse_query("references:$mid", 0);
236 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
238 $self->do_enquire($query);
242 my ($self, $mid, $opts) = @_;
243 my $smsg = eval { $self->lookup_message($mid) };
245 return { total => 0, msgs => [] } unless $smsg;
247 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
248 my $qsub = $qp->parse_query('path:'.mid_compressed($smsg->path));
249 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
250 $self->do_enquire($query, $opts);
256 my ($self, $query, $opts) = @_;
257 my $enquire = $self->enquire;
259 $query = Search::Xapian::Query->new(OP_AND, $query, $mail_query);
260 $enquire->set_query($query);
261 $enquire->set_sort_by_relevance_then_value(TS, 0);
263 my $offset = $opts->{offset} || 0;
264 my $limit = $opts->{limit} || 50;
265 my $mset = $enquire->get_mset($offset, $limit);
267 PublicInbox::SearchMsg->load_doc($_->get_document);
270 { total => $mset->get_matches_estimated, msgs => \@msgs }
274 sub stemmer { Search::Xapian::Stem->new($LANG) }
280 my $qp = $self->{query_parser};
284 $qp = Search::Xapian::QueryParser->new;
285 $qp->set_default_op(OP_AND);
286 $qp->set_database($self->{xdb});
287 $qp->set_stemmer($self->stemmer);
288 $qp->set_stemming_strategy(STEM_SOME);
289 $qp->add_valuerangeprocessor($self->ts_range_processor);
290 $qp->add_valuerangeprocessor($self->date_range_processor);
292 while (my ($name, $prefix) = each %bool_pfx_external) {
293 $qp->add_boolean_prefix($name, $prefix);
296 while (my ($name, $prefix) = each %prob_prefix) {
297 $qp->add_prefix($name, $prefix);
300 $self->{query_parser} = $qp;
303 sub term_generator { # write-only
306 my $tg = $self->{term_generator};
309 $tg = Search::Xapian::TermGenerator->new;
310 $tg->set_stemmer($self->stemmer);
312 $self->{term_generator} = $tg;
315 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
317 # increments last_thread_id counter
318 # returns a 64-bit integer represented as a hex string
321 my $db = $self->{xdb};
322 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
324 $db->set_metadata('last_thread_id', ++$last_thread_id);
329 sub ts_range_processor {
330 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
333 sub date_range_processor {
334 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
338 my ($self, $smsg, $is_ghost) = @_;
341 $smsg->ensure_metadata;
343 $self->link_message_to_parents($smsg);
347 sub link_message_to_parents {
348 my ($self, $smsg) = @_;
349 my $doc = $smsg->{doc};
350 my $mid = mid_compressed($smsg->mid);
351 my $mime = $smsg->mime;
352 my $refs = $mime->header_obj->header('References');
353 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
354 my $irt = $mime->header_obj->header('In-Reply-To');
356 $irt = mid_compressed(mid_clean($irt));
358 # maybe some crazies will try to make a circular reference:
362 # last References should be $irt
363 # we will de-dupe later
370 my @crefs = map { mid_compressed($_) } @refs;
371 my %uniq = ($mid => 1);
373 # prevent circular references via References: here:
375 foreach my $ref (@crefs) {
382 $doc->add_term(xpfx('inreplyto') . $irt) if defined $irt;
383 $smsg->{references_sorted} = '<'.join('><', @refs).'>';
385 my $ref_pfx = xpfx('references');
387 # first ref *should* be the thread root,
388 # but we can never trust clients to do the right thing
389 my $ref = shift @refs;
390 $doc->add_term($ref_pfx . $ref);
391 $tid = $self->_resolve_mid_to_tid($ref);
393 # the rest of the refs should point to this tid:
394 foreach $ref (@refs) {
395 $doc->add_term($ref_pfx . $ref);
396 my $ptid = $self->_resolve_mid_to_tid($ref);
398 $self->merge_threads($tid, $ptid);
402 $tid = $self->next_thread_id;
404 $doc->add_term(xpfx('thread') . $tid);
408 my ($self, $mid) = @_;
409 $mid = mid_clean($mid);
410 $mid = mid_compressed($mid);
412 my $doc_id = $self->find_unique_doc_id('mid', $mid);
414 if (defined $doc_id) {
416 my $doc = $self->{xdb}->get_document($doc_id);
417 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
418 $smsg->doc_id($doc_id);
423 sub find_unique_doc_id {
424 my ($self, $term, $value) = @_;
426 my ($begin, $end) = $self->find_doc_ids($term, $value);
428 return undef if $begin->equal($end); # not found
430 my $rv = $begin->get_docid;
434 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
438 # returns begin and end PostingIterator
440 my ($self, $term, $value) = @_;
442 $self->find_doc_ids_for_term(xpfx($term) . $value);
445 # returns begin and end PostingIterator
446 sub find_doc_ids_for_term {
447 my ($self, $term) = @_;
448 my $db = $self->{xdb};
450 ($db->postlist_begin($term), $db->postlist_end($term));
453 # this will create a ghost as necessary
454 sub _resolve_mid_to_tid {
455 my ($self, $mid) = @_;
457 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
462 my ($self, $mid, $tid) = @_;
464 $mid = mid_compressed($mid);
465 $tid = $self->next_thread_id unless defined $tid;
467 my $doc = Search::Xapian::Document->new;
468 $doc->add_term(xpfx('mid') . $mid);
469 $doc->add_term(xpfx('thread') . $tid);
470 $doc->add_term(xpfx('type') . 'ghost');
472 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
473 $self->link_message($smsg, 1);
474 $self->{xdb}->add_document($doc);
480 my ($self, $winner_tid, $loser_tid) = @_;
481 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
482 my $thread_pfx = xpfx('thread');
483 my $db = $self->{xdb};
485 for (; $head != $tail; $head->inc) {
486 my $docid = $head->get_docid;
487 my $doc = $db->get_document($docid);
488 $doc->remove_term($thread_pfx . $loser_tid);
489 $doc->add_term($thread_pfx . $winner_tid);
490 $db->replace_document($docid, $doc);
494 # normalize subjects so they are suitable as pathnames for URLs
497 $subj = subject_normalized($subj);
498 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
502 sub subject_normalized {
504 $subj =~ s/\A\s+//s; # no leading space
505 $subj =~ s/\s+\z//s; # no trailing space
506 $subj =~ s/\s+/ /gs; # no redundant spaces
507 $subj =~ s/\.+\z//; # no trailing '.'
508 $subj =~ s/$REPLY_RE//igo; # remove reply prefix
513 my ($git, $blob) = @_;
515 my $str = $git->cat_file($blob);
516 Email::MIME->new($str);
522 my ($self, $git, $blob) = @_;
523 my $mime = do_cat_mail($git, $blob) or return;
524 eval { $self->add_message($mime) };
525 warn "W: index_blob $blob: $@\n" if $@;
529 my ($self, $git, $blob) = @_;
530 my $mime = do_cat_mail($git, $blob) or return;
531 my $mid = $mime->header_obj->header('Message-ID');
532 eval { $self->remove_message($mid) } if defined $mid;
533 warn "W: unindex_blob $blob: $@\n" if $@;
538 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
541 # indexes all unindexed messages
543 my ($self, $head) = @_;
544 require PublicInbox::GitCatFile;
545 my $db = $self->{xdb};
546 my $hex = '[a-f0-9]';
547 my $h40 = $hex .'{40}';
548 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
549 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
552 $db->begin_transaction;
554 my $git = PublicInbox::GitCatFile->new($self->{git_dir});
556 my $latest = $db->get_metadata('last_commit');
557 my $range = length $latest ? "$latest..$head" : $head;
560 # get indexed messages
561 my @cmd = ('git', "--git-dir=$self->{git_dir}", "log",
562 qw/--reverse --no-notes --no-color --raw -r
563 --no-abbrev/, $range);
564 my $pid = open(my $log, '-|', @cmd) or
565 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
567 while (my $line = <$log>) {
568 if ($line =~ /$addmsg/o) {
569 $self->index_blob($git, $1);
570 } elsif ($line =~ /$delmsg/o) {
571 $self->unindex_blob($git, $1);
572 } elsif ($line =~ /^commit ($h40)/o) {
577 $db->set_metadata('last_commit', $latest) if defined $latest;
580 warn "indexing failed: $@\n";
581 $db->cancel_transaction;
583 $db->commit_transaction;