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
5 # Indexes mail with Xapian and our (SQLite-based) ::Msgmap for use
6 # with the web and NNTP interfaces. This index maintains thread
7 # relationships for use by PublicInbox::SearchThread.
8 # This writes to the search index.
9 package PublicInbox::SearchIdx;
12 use Fcntl qw(:flock :DEFAULT);
13 use PublicInbox::MIME;
14 use Email::MIME::ContentType;
15 $Email::MIME::ContentType::STRICT_PARAMS = 0;
16 use base qw(PublicInbox::Search);
17 use PublicInbox::MID qw/mid_clean id_compress mid_mime/;
18 use PublicInbox::MsgIter;
20 use POSIX qw(strftime);
21 require PublicInbox::Git;
23 use constant MAX_MID_SIZE => 244; # max term size - 1 in Xapian
27 OLD_PERM_EVERYBODY => 2,
29 PERM_EVERYBODY => 0664,
33 my ($class, $inbox, $creat) = @_;
37 $git_dir = $inbox->{mainrepo};
38 $altid = $inbox->{altid};
40 require PublicInbox::AltId;
42 PublicInbox::AltId->new($inbox, $_);
46 require Search::Xapian::WritableDatabase;
47 my $self = bless { git_dir => $git_dir, -altid => $altid }, $class;
48 my $perm = $self->_git_config_perm;
49 my $umask = _umask_for($perm);
50 $self->{umask} = $umask;
51 $self->{lock_path} = "$git_dir/ssoma.lock";
52 $self->{git} = PublicInbox::Git->new($git_dir);
53 $self->{creat} = ($creat || 0) == 1;
59 my $xdb = delete $self->{xdb} or croak 'not acquired';
61 _lock_release($self) if $self->{creat};
67 croak 'already acquired' if $self->{xdb};
68 my $dir = PublicInbox::Search->xdir($self->{git_dir});
69 my $flag = Search::Xapian::DB_OPEN;
73 File::Path::mkpath($dir);
74 $self->{batch_size} = 100;
75 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
77 $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag);
80 # we only acquire the flock if creating or reindexing;
81 # PublicInbox::Import already has the lock on its own.
84 croak 'already locked' if $self->{lockfh};
85 sysopen(my $lockfh, $self->{lock_path}, O_WRONLY|O_CREAT) or
86 die "failed to open lock $self->{lock_path}: $!\n";
87 flock($lockfh, LOCK_EX) or die "lock failed: $!\n";
88 $self->{lockfh} = $lockfh;
93 my $lockfh = delete $self->{lockfh} or croak 'not locked';
94 flock($lockfh, LOCK_UN) or die "unlock failed: $!\n";
95 close $lockfh or die "close failed: $!\n";
99 my ($doc, $col, $num) = @_;
100 $num = Search::Xapian::sortable_serialise($num);
101 $doc->add_value($col, $num);
104 sub add_values ($$$) {
105 my ($smsg, $bytes, $num) = @_;
108 my $doc = $smsg->{doc};
109 add_val($doc, &PublicInbox::Search::TS, $ts);
111 defined($num) and add_val($doc, &PublicInbox::Search::NUM, $num);
113 defined($bytes) and add_val($doc, &PublicInbox::Search::BYTES, $bytes);
115 add_val($doc, &PublicInbox::Search::LINES,
116 $smsg->{mime}->body_raw =~ tr!\n!\n!);
118 my $yyyymmdd = strftime('%Y%m%d', gmtime($ts));
119 add_val($doc, PublicInbox::Search::YYYYMMDD, $yyyymmdd);
122 sub index_users ($$) {
123 my ($tg, $smsg) = @_;
125 my $from = $smsg->from;
129 $tg->index_text($from, 1, 'A'); # A - author
130 $tg->increase_termpos;
131 $tg->index_text($to, 1, 'XTO') if $to ne '';
132 $tg->increase_termpos;
133 $tg->index_text($cc, 1, 'XCC') if $cc ne '';
134 $tg->increase_termpos;
137 sub index_body ($$$) {
138 my ($tg, $lines, $inc) = @_;
139 $tg->index_text(join("\n", @$lines), $inc, $inc ? 'XNQ' : 'XQUOT');
141 $tg->increase_termpos;
145 my ($self, $mime, $bytes, $num, $blob) = @_; # mime = Email::MIME object
146 my $db = $self->{xdb};
148 my ($doc_id, $old_tid);
149 my $mid = mid_clean(mid_mime($mime));
152 die 'Message-ID too long' if length($mid) > MAX_MID_SIZE;
153 my $smsg = $self->lookup_message($mid);
155 # convert a ghost to a regular message
156 # it will also clobber any existing regular message
157 $doc_id = $smsg->{doc_id};
158 $old_tid = $smsg->thread_id;
160 $smsg = PublicInbox::SearchMsg->new($mime);
161 my $doc = $smsg->{doc};
162 $doc->add_term('Q' . $mid);
164 my $subj = $smsg->subject;
166 my $path = $self->subject_path($subj);
167 $doc->add_term('XPATH' . id_compress($path));
170 add_values($smsg, $bytes, $num);
172 my $tg = $self->term_generator;
174 $tg->set_document($doc);
175 $tg->index_text($subj, 1, 'S') if $subj;
176 $tg->increase_termpos;
178 index_users($tg, $smsg);
180 msg_iter($mime, sub {
181 my ($part, $depth, @idx) = @{$_[0]};
182 my $ct = $part->content_type || 'text/plain';
183 my $fn = $part->filename;
184 if (defined $fn && $fn ne '') {
185 $tg->index_text($fn, 1, 'XFN');
188 return if $ct =~ m!\btext/x?html\b!i;
190 my $s = eval { $part->body_str };
192 if ($ct =~ m!\btext/plain\b!i) {
193 # Try to assume UTF-8 because Alpine
194 # seems to do wacky things and set
196 $part->charset_set('UTF-8');
197 $s = eval { $part->body_str };
198 $s = $part->body if $@;
201 defined $s or return;
204 my $body = $part->body;
205 my @lines = split(/\n/, $body);
206 while (defined(my $l = shift @lines)) {
208 index_body($tg, \@orig, 1) if @orig;
211 index_body($tg, \@quot, 0) if @quot;
215 index_body($tg, \@quot, 0) if @quot;
216 index_body($tg, \@orig, 1) if @orig;
219 link_message($self, $smsg, $old_tid);
220 $tg->index_text($mid, 1, 'XMID');
221 $doc->set_data($smsg->to_doc_data($blob));
223 if (my $altid = $self->{-altid}) {
224 foreach my $alt (@$altid) {
225 my $id = $alt->mid2alt($mid);
226 next unless defined $id;
227 $doc->add_term($alt->{xprefix} . $id);
230 if (defined $doc_id) {
231 $db->replace_document($doc_id, $doc);
233 $doc_id = $db->add_document($doc);
238 warn "failed to index message <$mid>: $@\n";
244 # returns deleted doc_id on success, undef on missing
246 my ($self, $mid) = @_;
247 my $db = $self->{xdb};
249 $mid = mid_clean($mid);
252 $doc_id = $self->find_unique_doc_id('mid', $mid);
253 $db->delete_document($doc_id) if defined $doc_id;
257 warn "failed to remove message <$mid>: $@\n";
263 sub term_generator { # write-only
266 my $tg = $self->{term_generator};
269 $tg = Search::Xapian::TermGenerator->new;
270 $tg->set_stemmer($self->stemmer);
272 $self->{term_generator} = $tg;
275 # increments last_thread_id counter
276 # returns a 64-bit integer represented as a hex string
279 my $db = $self->{xdb};
280 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
282 $db->set_metadata('last_thread_id', ++$last_thread_id);
288 my ($self, $smsg, $old_tid) = @_;
289 my $doc = $smsg->{doc};
290 my $mid = $smsg->mid;
291 my $mime = $smsg->{mime};
292 my $hdr = $mime->header_obj;
294 # last References should be IRT, but some mail clients do things
295 # out of order, so trust IRT over References iff IRT exists
296 my @refs = ($hdr->header_raw('References'),
297 $hdr->header_raw('In-Reply-To'));
298 @refs = ((join(' ', @refs)) =~ /<([^>]+)>/g);
302 my %uniq = ($mid => 1);
303 my @orig_refs = @refs;
306 # prevent circular references via References: here:
307 foreach my $ref (@orig_refs) {
308 if (length($ref) > MAX_MID_SIZE) {
309 warn "References: <$ref> too long, ignoring\n";
318 $smsg->{references} = '<'.join('> <', @refs).'>';
320 # first ref *should* be the thread root,
321 # but we can never trust clients to do the right thing
322 my $ref = shift @refs;
323 $tid = $self->_resolve_mid_to_tid($ref);
324 $self->merge_threads($tid, $old_tid) if defined $old_tid;
326 # the rest of the refs should point to this tid:
327 foreach $ref (@refs) {
328 my $ptid = $self->_resolve_mid_to_tid($ref);
329 merge_threads($self, $tid, $ptid);
332 $tid = defined $old_tid ? $old_tid : $self->next_thread_id;
334 $doc->add_term('G' . $tid);
338 my ($self, $mime, $bytes, $num, $blob) = @_;
339 $self->add_message($mime, $bytes, $num, $blob);
343 my ($self, $mime) = @_;
344 my $mid = eval { mid_clean(mid_mime($mime)) };
345 $self->remove_message($mid) if defined $mid;
349 my ($self, $mime) = @_;
350 $self->{mm}->mid_insert(mid_clean(mid_mime($mime)));
354 my ($self, $mime) = @_;
355 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
359 my ($self, $mime, $bytes, $blob) = @_;
360 my $num = $self->{mm}->num_for(mid_clean(mid_mime($mime)));
361 index_blob($self, $mime, $bytes, $num, $blob);
365 my ($self, $mime) = @_;
366 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
367 unindex_blob($self, $mime);
371 my ($self, $mime, $bytes, $blob) = @_;
372 my $num = index_mm($self, $mime);
373 index_blob($self, $mime, $bytes, $num, $blob);
377 my ($self, $mime) = @_;
378 unindex_blob($self, $mime);
379 unindex_mm($self, $mime);
383 my ($git, $blob, $sizeref) = @_;
385 my $str = $git->cat_file($blob, $sizeref);
386 # fixup bugs from import:
387 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
388 PublicInbox::MIME->new($str);
394 my ($self, $opts) = @_;
395 with_umask($self, sub { $self->_index_sync($opts) });
399 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
400 my $hex = '[a-f0-9]';
401 my $h40 = $hex .'{40}';
402 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
403 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
404 my $git = $self->{git};
407 my $max = $self->{batch_size}; # may be undef
410 while (defined($line = <$log>)) {
411 if ($line =~ /$addmsg/o) {
413 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
414 $add_cb->($self, $mime, $bytes, $blob);
415 } elsif ($line =~ /$delmsg/o) {
417 my $mime = do_cat_mail($git, $blob) or next;
418 $del_cb->($self, $mime);
419 } elsif ($line =~ /^commit ($h40)/o) {
420 if (defined $max && --$max <= 0) {
421 $max = $self->{batch_size};
422 $batch_cb->($latest, 1);
427 $batch_cb->($latest, 0);
433 require PublicInbox::Msgmap;
434 PublicInbox::Msgmap->new($self->{git_dir}, 1);
439 my ($self, $range) = @_;
440 $self->{git}->popen(qw/log --reverse --no-notes --no-color
441 --raw -r --no-abbrev/, $range);
444 # indexes all unindexed messages
446 my ($self, $opts) = @_;
447 my $tip = $opts->{ref} || 'HEAD';
448 my $reindex = $opts->{reindex};
449 my ($mkey, $last_commit, $lx, $xlog);
450 $self->{git}->batch_prepare;
451 my $xdb = _xdb_acquire($self);
452 $xdb->begin_transaction;
455 $mkey = 'last_commit';
456 $last_commit = $xdb->get_metadata('last_commit');
460 $mkey = undef if $last_commit ne '';
462 $xdb->cancel_transaction;
463 $xdb = _xdb_release($self);
465 # ensure we leak no FDs to "git log"
466 my $range = $lx eq '' ? $tip : "$lx..$tip";
467 $xlog = _git_log($self, $range);
469 $xdb = _xdb_acquire($self);
470 $xdb->begin_transaction;
471 } while ($xdb->get_metadata('last_commit') ne $last_commit);
473 my $mm = _msgmap_init($self);
474 my $dbh = $mm->{dbh} if $mm;
477 my ($commit, $more) = @_;
479 $mm->last_commit($commit) if $commit;
483 $xdb->set_metadata($mkey, $commit) if $mkey && $commit;
484 $xdb->commit_transaction;
485 $xdb = _xdb_release($self);
487 # let another process do some work... <
490 $xdb = _xdb_acquire($self);
491 $xdb->begin_transaction;
493 $dbh->begin_work if $dbh;
499 my $lm = $mm->last_commit || '';
501 # Common case is the indexes are synced,
502 # we only need to run git-log once:
503 rlog($self, $xlog, *index_both, *unindex_both, $cb);
505 # Uncommon case, msgmap and xapian are out-of-sync
506 # do not care for performance (but git is fast :>)
507 # This happens if we have to reindex Xapian since
508 # msgmap is a frozen format and our Xapian format
510 my $r = $lm eq '' ? $tip : "$lm..$tip";
512 # first, ensure msgmap is up-to-date:
513 my $mkey_prev = $mkey;
514 $mkey = undef; # ignore xapian, for now
515 my $mlog = _git_log($self, $r);
517 rlog($self, $mlog, *index_mm, *unindex_mm, $cb);
518 $mm_only = $mlog = undef;
520 # now deal with Xapian
523 rlog($self, $xlog, *index_mm2, *unindex_mm2, $cb);
526 # user didn't install DBD::SQLite and DBI
527 rlog($self, $xlog, *index_blob, *unindex_blob, $cb);
531 # this will create a ghost as necessary
532 sub _resolve_mid_to_tid {
533 my ($self, $mid) = @_;
535 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
540 my ($self, $mid) = @_;
542 my $tid = $self->next_thread_id;
543 my $doc = Search::Xapian::Document->new;
544 $doc->add_term('Q' . $mid);
545 $doc->add_term('G' . $tid);
546 $doc->add_term('T' . 'ghost');
548 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
549 $self->{xdb}->add_document($doc);
555 my ($self, $winner_tid, $loser_tid) = @_;
556 return if $winner_tid == $loser_tid;
557 my ($head, $tail) = $self->find_doc_ids('G' . $loser_tid);
558 my $db = $self->{xdb};
560 for (; $head != $tail; $head->inc) {
561 my $docid = $head->get_docid;
562 my $doc = $db->get_document($docid);
563 $doc->remove_term('G' . $loser_tid);
564 $doc->add_term('G' . $winner_tid);
565 $db->replace_document($docid, $doc);
569 sub _read_git_config_perm {
571 my @cmd = qw(config core.sharedRepository);
572 my $fh = PublicInbox::Git->new($self->{git_dir})->popen(@cmd);
575 chomp $perm if defined $perm;
579 sub _git_config_perm {
581 my $perm = scalar @_ ? $_[0] : _read_git_config_perm($self);
582 return PERM_GROUP if (!defined($perm) || $perm eq '');
583 return PERM_UMASK if ($perm eq 'umask');
584 return PERM_GROUP if ($perm eq 'group');
585 if ($perm =~ /\A(?:all|world|everybody)\z/) {
586 return PERM_EVERYBODY;
588 return PERM_GROUP if ($perm =~ /\A(?:true|yes|on|1)\z/);
589 return PERM_UMASK if ($perm =~ /\A(?:false|no|off|0)\z/);
592 return PERM_UMASK if ($i == PERM_UMASK);
593 return PERM_GROUP if ($i == OLD_PERM_GROUP);
594 return PERM_EVERYBODY if ($i == OLD_PERM_EVERYBODY);
596 if (($i & 0600) != 0600) {
597 die "core.sharedRepository mode invalid: ".
598 sprintf('%.3o', $i) . "\nOwner must have permissions\n";
604 my ($perm) = @_; # _git_config_perm return value
606 return umask if $rv == 0;
608 # set +x bit if +r or +w were set
609 $rv |= 0100 if ($rv & 0600);
610 $rv |= 0010 if ($rv & 0060);
611 $rv |= 0001 if ($rv & 0006);
616 my ($self, $cb) = @_;
617 my $old = umask $self->{umask};
618 my $rv = eval { $cb->() };
626 # order matters for unlocking
627 $_[0]->{xdb} = undef;
628 $_[0]->{lockfh} = undef;