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 '.'
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('Message-ID'));
88 my $mid = mid_compressed($mid_orig);
90 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
93 my $smsg = $self->lookup_message($mid);
97 $smsg->ensure_metadata;
98 # convert a ghost to a regular message
99 # it will also clobber any existing regular message
103 my $type = xpfx('type');
105 $doc->remove_term($type . 'ghost');
109 # probably does not exist:
110 eval { $doc->remove_term($type . 'mail') };
111 $doc->add_term($type . 'mail');
113 $smsg = PublicInbox::SearchMsg->new($mime);
115 $doc->add_term(xpfx('mid') . $mid);
118 my $subj = $smsg->subject;
121 $doc->add_term(xpfx('subject') . $subj);
123 my $path = subject_path($subj);
124 $doc->add_term(xpfx('path') . mid_compressed($path));
127 my $from = $smsg->from_name;
128 my $date = $smsg->date;
129 my $ts = Search::Xapian::sortable_serialise($smsg->ts);
130 $doc->add_value(PublicInbox::Search::TS, $ts);
132 $doc->set_data($smsg->to_doc_data);
134 my $tg = $self->term_generator;
136 $tg->set_document($doc);
137 $tg->index_text($subj, 1, 'S') if $subj;
138 $tg->increase_termpos;
139 $tg->index_text($subj) if $subj;
140 $tg->increase_termpos;
142 $tg->index_text($smsg->from->format);
143 $tg->increase_termpos;
145 $mime->walk_parts(sub {
147 return if $part->subparts; # walk_parts already recurses
148 my $ct = $part->content_type || $ct_msg;
150 # account for filter bugs...
151 $ct =~ m!\btext/plain\b!i or return;
154 my $body = $part->body;
156 my @lines = split(/\n/, $body);
157 while (defined(my $l = shift @lines)) {
165 $tg->index_text(join("\n", @quot), 0);
167 $tg->increase_termpos;
170 $tg->index_text(join("\n", @orig));
172 $tg->increase_termpos;
177 $doc_id = $smsg->doc_id;
178 $self->link_message($smsg, 0);
179 $db->replace_document($doc_id, $doc);
181 $self->link_message($smsg, 0);
182 $doc_id = $db->add_document($doc);
187 warn "failed to index message <$mid_orig>: $@\n";
193 # returns deleted doc_id on success, undef on missing
195 my ($self, $mid_orig) = @_;
196 my $db = $self->{xdb};
198 $mid_orig = mid_clean($mid_orig);
199 my $mid = mid_compressed($mid_orig);
202 $doc_id = $self->find_unique_doc_id('mid', $mid);
203 $db->delete_document($doc_id) if defined $doc_id;
207 warn "failed to remove message <$mid_orig>: $@\n";
215 my ($self, $query_string, $opts) = @_;
216 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
218 $self->do_enquire($query, $opts);
221 sub get_subject_path {
222 my ($self, $path, $opts) = @_;
223 my $query = $self->qp->parse_query("path:".mid_compressed($path), 0);
224 $self->do_enquire($query);
227 # given a message ID, get followups to a message
229 my ($self, $mid, $opts) = @_;
230 $mid = mid_clean($mid);
231 $mid = mid_compressed($mid);
233 my $irt = $qp->parse_query("inreplyto:$mid", 0);
234 my $ref = $qp->parse_query("references:$mid", 0);
235 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
237 $self->do_enquire($query);
241 my ($self, $mid, $opts) = @_;
242 my $smsg = eval { $self->lookup_message($mid) };
244 return { count => 0, msgs => [] } unless $smsg;
246 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
247 my $qsub = $qp->parse_query('path:'.mid_compressed($smsg->path));
248 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
249 $self->do_enquire($query);
255 my ($self, $query, $opts) = @_;
256 my $enquire = $self->enquire;
258 $query = Search::Xapian::Query->new(OP_AND, $query, $mail_query);
259 $enquire->set_query($query);
260 $enquire->set_sort_by_relevance_then_value(TS, 0);
262 my $offset = $opts->{offset} || 0;
263 my $limit = $opts->{limit} || 50;
264 my $mset = $enquire->get_mset($offset, $limit);
266 PublicInbox::SearchMsg->load_doc($_->get_document);
269 { count => $mset->get_matches_estimated, msgs => \@msgs }
273 sub stemmer { Search::Xapian::Stem->new($LANG) }
279 my $qp = $self->{query_parser};
283 $qp = Search::Xapian::QueryParser->new;
284 $qp->set_default_op(OP_AND);
285 $qp->set_database($self->{xdb});
286 $qp->set_stemmer($self->stemmer);
287 $qp->set_stemming_strategy(STEM_SOME);
288 $qp->add_valuerangeprocessor($self->ts_range_processor);
289 $qp->add_valuerangeprocessor($self->date_range_processor);
291 while (my ($name, $prefix) = each %bool_pfx_external) {
292 $qp->add_boolean_prefix($name, $prefix);
295 while (my ($name, $prefix) = each %prob_prefix) {
296 $qp->add_prefix($name, $prefix);
299 $self->{query_parser} = $qp;
302 sub term_generator { # write-only
305 my $tg = $self->{term_generator};
308 $tg = Search::Xapian::TermGenerator->new;
309 $tg->set_stemmer($self->stemmer);
311 $self->{term_generator} = $tg;
314 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
316 # increments last_thread_id counter
317 # returns a 64-bit integer represented as a hex string
320 my $db = $self->{xdb};
321 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
323 $db->set_metadata('last_thread_id', ++$last_thread_id);
328 sub ts_range_processor {
329 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
332 sub date_range_processor {
333 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
337 my ($self, $smsg, $is_ghost) = @_;
340 $smsg->ensure_metadata;
342 $self->link_message_to_parents($smsg);
346 sub link_message_to_parents {
347 my ($self, $smsg) = @_;
348 my $doc = $smsg->{doc};
349 my $mid = mid_compressed($smsg->mid);
350 my $mime = $smsg->mime;
351 my $refs = $mime->header_obj->header('References');
352 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
353 my $irt = $mime->header_obj->header('In-Reply-To');
355 if ($irt =~ /<([^>]+)>/) {
359 # maybe some crazies will try to make a circular reference:
369 my @crefs = map { mid_compressed($_) } @refs;
370 my %uniq = ($mid => 1);
372 # prevent circular references via References: here:
374 foreach my $ref (@crefs) {
379 $irt = undef if (defined $irt && !$uniq{$irt});
383 $doc->add_term(xpfx('inreplyto') . $irt);
386 my $ref_pfx = xpfx('references');
388 # first ref *should* be the thread root,
389 # but we can never trust clients to do the right thing
390 my $ref = shift @refs;
391 $doc->add_term($ref_pfx . $ref);
392 $tid = $self->_resolve_mid_to_tid($ref);
394 # the rest of the refs should point to this tid:
395 foreach $ref (@refs) {
396 $doc->add_term($ref_pfx . $ref);
397 my $ptid = $self->_resolve_mid_to_tid($ref);
399 $self->merge_threads($tid, $ptid);
403 $tid = $self->next_thread_id;
405 $doc->add_term(xpfx('thread') . $tid);
409 my ($self, $mid) = @_;
410 $mid = mid_clean($mid);
411 $mid = mid_compressed($mid);
413 my $doc_id = $self->find_unique_doc_id('mid', $mid);
415 if (defined $doc_id) {
417 my $doc = $self->{xdb}->get_document($doc_id);
418 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
419 $smsg->doc_id($doc_id);
424 sub find_unique_doc_id {
425 my ($self, $term, $value) = @_;
427 my ($begin, $end) = $self->find_doc_ids($term, $value);
429 return undef if $begin->equal($end); # not found
431 my $rv = $begin->get_docid;
435 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
439 # returns begin and end PostingIterator
441 my ($self, $term, $value) = @_;
443 $self->find_doc_ids_for_term(xpfx($term) . $value);
446 # returns begin and end PostingIterator
447 sub find_doc_ids_for_term {
448 my ($self, $term) = @_;
449 my $db = $self->{xdb};
451 ($db->postlist_begin($term), $db->postlist_end($term));
454 # this will create a ghost as necessary
455 sub _resolve_mid_to_tid {
456 my ($self, $mid) = @_;
458 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
463 my ($self, $mid, $tid) = @_;
465 $mid = mid_compressed($mid);
466 $tid = $self->next_thread_id unless defined $tid;
468 my $doc = Search::Xapian::Document->new;
469 $doc->add_term(xpfx('mid') . $mid);
470 $doc->add_term(xpfx('thread') . $tid);
471 $doc->add_term(xpfx('type') . 'ghost');
473 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
474 $self->link_message($smsg, 1);
475 $self->{xdb}->add_document($doc);
481 my ($self, $winner_tid, $loser_tid) = @_;
482 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
483 my $thread_pfx = xpfx('thread');
484 my $db = $self->{xdb};
486 for (; $head != $tail; $head->inc) {
487 my $docid = $head->get_docid;
488 my $doc = $db->get_document($docid);
489 $doc->remove_term($thread_pfx . $loser_tid);
490 $doc->add_term($thread_pfx . $winner_tid);
491 $db->replace_document($docid, $doc);
495 # normalize subjects so they are suitable as pathnames for URLs
498 $subj = subject_normalized($subj);
499 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
503 sub subject_normalized {
505 $subj =~ s/\A\s+//s; # no leading space
506 $subj =~ s/\s+\z//s; # no trailing space
507 $subj =~ s/\s+/ /gs; # no redundant spaces
508 $subj =~ s/\.+\z//; # no trailing '.'
509 $subj =~ s/$REPLY_RE//igo; # remove reply prefix
514 my ($git, $blob) = @_;
516 my $str = $git->cat_file($blob);
517 Email::MIME->new($str);
523 my ($self, $git, $blob) = @_;
524 my $mime = do_cat_mail($git, $blob) or return;
525 eval { $self->add_message($mime) };
526 warn "W: index_blob $blob: $@\n" if $@;
530 my ($self, $git, $blob) = @_;
531 my $mime = do_cat_mail($git, $blob) or return;
532 my $mid = $mime->header_obj->header('Message-ID');
533 eval { $self->remove_message($mid) } if defined $mid;
534 warn "W: unindex_blob $blob: $@\n" if $@;
539 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
542 # indexes all unindexed messages
544 my ($self, $head) = @_;
545 require PublicInbox::GitCatFile;
546 my $db = $self->{xdb};
547 my $hex = '[a-f0-9]';
548 my $h40 = $hex .'{40}';
549 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
550 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
553 $db->begin_transaction;
555 my $git = PublicInbox::GitCatFile->new($self->{git_dir});
557 my $latest = $db->get_metadata('last_commit');
558 my $range = length $latest ? "$latest..$head" : $head;
561 # get indexed messages
562 my @cmd = ('git', "--git-dir=$self->{git_dir}", "log",
563 qw/--reverse --no-notes --no-color --raw -r
564 --no-abbrev/, $range);
565 my $pid = open(my $log, '-|', @cmd) or
566 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
568 while (my $line = <$log>) {
569 if ($line =~ /$addmsg/o) {
570 $self->index_blob($git, $1);
571 } elsif ($line =~ /$delmsg/o) {
572 $self->unindex_blob($git, $1);
573 } elsif ($line =~ /^commit ($h40)/o) {
578 $db->set_metadata('last_commit', $latest) if defined $latest;
581 warn "indexing failed: $@\n";
582 $db->cancel_transaction;
584 $db->commit_transaction;