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;
10 use Search::Xapian qw/:standard/;
11 require PublicInbox::View;
12 use Date::Parse qw/str2time/;
16 our @EXPORT = qw/xpfx mid_compressed/;
23 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
26 use constant MID_MAX => SHA1HEX_LEN;
29 my %bool_pfx_internal = (
30 type => 'T', # "mail" or "ghost"
31 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
34 my %bool_pfx_external = (
36 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
37 references => 'XREFS',
45 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
47 sub xpfx { $all_pfx{$_[0]} }
50 while (my ($k, $v) = each %all_pfx) {
51 next if $prob_prefix{$k};
52 $PFX2TERM_RMAP{$v} = $k;
55 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
60 return $mid if (length($mid) <= MID_MAX);
61 Digest::SHA::sha1_hex($mid);
65 my ($class, $git_dir, $writable) = @_;
66 # allow concurrent versions for easier rollback:
67 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
70 if ($writable) { # not used by the WWW interface
71 require Search::Xapian::WritableDatabase;
73 File::Path::mkpath($dir);
74 $db = Search::Xapian::WritableDatabase->new($dir,
75 Search::Xapian::DB_CREATE_OR_OPEN);
77 $db = Search::Xapian::Database->new($dir);
79 bless { xdb => $db, git_dir => $git_dir }, $class;
82 sub reopen { $_[0]->{xdb}->reopen }
85 my ($self, $mime) = @_; # mime = Email::MIME object
86 my $db = $self->{xdb};
89 my $mid = clean_mid($mime->header('Message-ID'));
90 $mid = mid_compressed($mid);
92 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
93 my $enc_msg = PublicInbox::View::enc_for($ct_msg);
95 $db->begin_transaction;
97 my $smsg = $self->lookup_message($mid);
101 $smsg->ensure_metadata;
102 # convert a ghost to a regular message
103 # it will also clobber any existing regular message
107 my $type = xpfx('type');
109 $doc->remove_term($type . 'ghost');
113 # probably does not exist:
114 eval { $doc->remove_term($type . 'mail') };
115 $doc->add_term($type . 'mail');
117 $smsg = PublicInbox::SearchMsg->new($mime);
119 $doc->add_term(xpfx('mid') . $mid);
122 my $subj = $mime->header('Subject');
123 $subj = '' unless defined $subj;
126 $doc->add_term(xpfx('subject') . $subj);
128 my $path = subject_path($subj);
129 $doc->add_term(xpfx('path') . $path);
132 my $from = $mime->header('From') || '';
136 @from = Email::Address->parse($from);
137 $from = $from[0]->name;
140 my $ts = eval { str2time($mime->header('Date')) } || 0;
141 my $date = POSIX::strftime('%Y-%m-%d %H:%M', gmtime($ts));
142 $ts = Search::Xapian::sortable_serialise($ts);
143 $doc->add_value(PublicInbox::Search::TS, $ts);
145 # this is what we show in index results:
148 $doc->set_data("$mid\n$subj\n$from\n$date");
150 my $tg = $self->term_generator;
152 $tg->set_document($doc);
153 $tg->index_text($subj, 1, 'S') if $subj;
154 $tg->increase_termpos;
155 $tg->index_text($subj) if $subj;
156 $tg->increase_termpos;
159 $tg->index_text($from[0]->format);
160 $tg->increase_termpos;
163 $mime->walk_parts(sub {
165 return if $part->subparts; # walk_parts already recurses
166 my $ct = $part->content_type || $ct_msg;
168 # account for filter bugs...
169 $ct =~ m!\btext/plain\b!i or return;
171 my $enc = PublicInbox::View::enc_for($ct, $enc_msg);
173 foreach my $l (split(/\n/, $enc->decode($part->body))) {
181 $tg->index_text(join("\n", @quot), 0);
182 $tg->increase_termpos;
185 $tg->index_text(join("\n", @orig));
186 $tg->increase_termpos;
191 $doc_id = $smsg->doc_id;
192 $self->link_message($smsg, 0);
193 $db->replace_document($doc_id, $doc);
195 $self->link_message($smsg, 0);
196 $doc_id = $db->add_document($doc);
201 warn "failed to index message <$mid>: $@\n";
202 $db->cancel_transaction;
204 $db->commit_transaction;
209 # returns deleted doc_id on success, undef on missing
211 my ($self, $mid) = @_;
212 my $db = $self->{xdb};
214 $mid = clean_mid($mid);
215 $mid = mid_compressed($mid);
217 $db->begin_transaction;
219 $doc_id = $self->find_unique_doc_id('mid', $mid);
220 $db->delete_document($doc_id) if defined $doc_id;
224 warn "failed to remove message <$mid>: $@\n";
225 $db->cancel_transaction;
227 $db->commit_transaction;
234 my ($self, $query_string, $opts) = @_;
235 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
237 $query = Search::Xapian::Query->new(OP_AND, $mail_query, $query);
238 $self->do_enquire($query, $opts);
241 # given a message ID, get replies to a message
243 my ($self, $mid, $opts) = @_;
244 $mid = clean_mid($mid);
245 $mid = mid_compressed($mid);
247 my $irt = $qp->parse_query("inreplyto:$mid", 0);
248 my $ref = $qp->parse_query("references:$mid", 0);
249 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
251 $self->do_enquire($query);
255 my ($self, $mid, $opts) = @_;
256 my $smsg = eval { $self->lookup_message($mid) };
258 return { count => 0, msgs => [] } unless $smsg;
260 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
261 my $qsub = $qp->parse_query('path:'.$smsg->path);
262 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
263 $self->do_enquire($query);
269 my ($self, $query, $opts) = @_;
270 my $enquire = $self->enquire;
272 $enquire->set_query($query);
273 $enquire->set_sort_by_relevance_then_value(TS, 0);
275 my $offset = $opts->{offset} || 0;
276 my $limit = $opts->{limit} || 50;
277 my $mset = $enquire->get_mset($offset, $limit);
278 my @msgs = map { $_->get_document->get_data } $mset->items;
280 { count => $mset->get_matches_estimated, msgs => \@msgs }
284 sub stemmer { Search::Xapian::Stem->new(LANG) }
290 my $qp = $self->{query_parser};
294 $qp = Search::Xapian::QueryParser->new;
295 $qp->set_default_op(OP_AND);
296 $qp->set_database($self->{xdb});
297 $qp->set_stemmer($self->stemmer);
298 $qp->set_stemming_strategy(STEM_SOME);
299 $qp->add_valuerangeprocessor($self->ts_range_processor);
300 $qp->add_valuerangeprocessor($self->date_range_processor);
302 while (my ($name, $prefix) = each %bool_pfx_external) {
303 $qp->add_boolean_prefix($name, $prefix);
306 while (my ($name, $prefix) = each %prob_prefix) {
307 $qp->add_prefix($name, $prefix);
310 $self->{query_parser} = $qp;
313 sub term_generator { # write-only
316 my $tg = $self->{term_generator};
319 $tg = Search::Xapian::TermGenerator->new;
320 $tg->set_stemmer($self->stemmer);
322 $self->{term_generator} = $tg;
325 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
327 # increments last_thread_id counter
328 # returns a 64-bit integer represented as a hex string
331 my $db = $self->{xdb};
332 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
334 $db->set_metadata('last_thread_id', ++$last_thread_id);
339 sub ts_range_processor {
340 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
343 sub date_range_processor {
344 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
349 defined($mid) or die "no Message-ID";
350 # MDA->precheck did more checking for us
357 my ($self, $smsg, $is_ghost) = @_;
360 $smsg->ensure_metadata;
362 $self->link_message_to_parents($smsg);
366 sub link_message_to_parents {
367 my ($self, $smsg) = @_;
368 my $doc = $smsg->{doc};
369 my $mid = mid_compressed($smsg->mid);
370 my $mime = $smsg->mime;
371 my $refs = $mime->header('References');
372 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
373 my $irt = $mime->header('In-Reply-To');
375 if ($irt =~ /<([^>]+)>/) {
383 @refs = map { mid_compressed($_) } @refs;
385 @refs = grep { !$uniq{$_}++ } @refs; # uniq
387 $doc->add_term(xpfx('inreplyto') . $refs[-1]);
389 my $ref_pfx = xpfx('references');
391 # first ref *should* be the thread root,
392 # but we can never trust clients to do the right thing
393 my $ref = shift @refs;
394 $doc->add_term($ref_pfx . $ref);
395 $tid = $self->_resolve_mid_to_tid($ref);
397 # the rest of the refs should point to this tid:
398 foreach $ref (@refs) {
399 $doc->add_term($ref_pfx . $ref);
400 my $ptid = $self->_resolve_mid_to_tid($ref);
402 $self->merge_threads($tid, $ptid);
406 $tid = $self->next_thread_id;
408 $doc->add_term(xpfx('thread') . $tid);
412 my ($self, $mid) = @_;
413 $mid = clean_mid($mid);
414 $mid = mid_compressed($mid);
416 my $doc_id = $self->find_unique_doc_id('mid', $mid);
418 if (defined $doc_id) {
420 my $doc = $self->{xdb}->get_document($doc_id);
421 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
422 $smsg->doc_id($doc_id);
427 sub find_unique_doc_id {
428 my ($self, $term, $value) = @_;
430 my ($begin, $end) = $self->find_doc_ids($term, $value);
432 return undef if $begin->equal($end); # not found
434 my $rv = $begin->get_docid;
438 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
442 # returns begin and end PostingIterator
444 my ($self, $term, $value) = @_;
446 $self->find_doc_ids_for_term(xpfx($term) . $value);
449 # returns begin and end PostingIterator
450 sub find_doc_ids_for_term {
451 my ($self, $term) = @_;
452 my $db = $self->{xdb};
454 ($db->postlist_begin($term), $db->postlist_end($term));
457 # this will create a ghost as necessary
458 sub _resolve_mid_to_tid {
459 my ($self, $mid) = @_;
461 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
466 my ($self, $mid, $tid) = @_;
468 $mid = mid_compressed($mid);
469 $tid = $self->next_thread_id unless defined $tid;
471 my $doc = Search::Xapian::Document->new;
472 $doc->add_term(xpfx('mid') . $mid);
473 $doc->add_term(xpfx('thread') . $tid);
474 $doc->add_term(xpfx('type') . 'ghost');
476 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
477 $self->link_message($smsg, 1);
478 $self->{xdb}->add_document($doc);
484 my ($self, $winner_tid, $loser_tid) = @_;
485 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
486 my $thread_pfx = xpfx('thread');
487 my $db = $self->{xdb};
489 for (; $head != $tail; $head->inc) {
490 my $docid = $head->get_docid;
491 my $doc = $db->get_document($docid);
492 $doc->remove_term($thread_pfx . $loser_tid);
493 $doc->add_term($thread_pfx . $winner_tid);
494 $db->replace_document($docid, $doc);
498 # normalize subjects so they are suitable as pathnames for URLs
504 $subj =~ s/^(?:re|aw):\s*//i; # remove reply prefix (aw: German)
505 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
510 my ($git, $blob) = @_;
512 my $str = $git->cat_file($blob);
513 Email::MIME->new($str);
519 my ($self, $git, $blob) = @_;
520 my $mime = do_cat_mail($git, $blob) or return;
521 eval { $self->add_message($mime) };
522 warn "W: index_blob $blob: $@\n" if $@;
526 my ($self, $git, $blob) = @_;
527 my $mime = do_cat_mail($git, $blob) or return;
528 my $mid = $mime->header('Message-ID');
529 eval { $self->remove_message($mid) } if defined $mid;
530 warn "W: unindex_blob $blob: $@\n" if $@;
535 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
538 # indexes all unindexed messages
540 my ($self, $git) = @_;
541 my $db = $self->{xdb};
542 my $latest = $db->get_metadata('last_commit');
543 my $range = length $latest ? "$latest..HEAD" : 'HEAD';
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}$!;
551 # get indexed messages
552 my @cmd = ('git', "--git-dir=$git->{git_dir}", "log",
553 qw/--reverse --no-notes --no-color --raw -r --no-abbrev/,
556 my $pid = open(my $log, '-|', @cmd) or
557 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
559 while (my $line = <$log>) {
560 if ($line =~ /$addmsg/o) {
561 $self->index_blob($git, $1);
562 } elsif ($line =~ /$delmsg/o) {
563 $self->unindex_blob($git, $1);
564 } elsif ($line =~ /^commit ($h40)/o) {
566 if (defined $latest) {
567 $db->set_metadata('last_commit', $latest)
573 $db->set_metadata('last_commit', $latest) if defined $latest;