1 # Copyright (C) 2015-2018 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <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 base qw(PublicInbox::Search);
15 use PublicInbox::MID qw/mid_clean id_compress mid_mime/;
16 use PublicInbox::MsgIter;
18 use POSIX qw(strftime);
19 require PublicInbox::Git;
22 MAX_MID_SIZE => 244, # max term size - 1 in Xapian
25 OLD_PERM_EVERYBODY => 2,
27 PERM_EVERYBODY => 0664,
28 BATCH_BYTES => 1_000_000,
29 DEBUG => !!$ENV{DEBUG},
44 return $s unless ($s =~ /\A"(.*)"\z/);
46 $s =~ s/\\([abfnrtv])/$GIT_ESC{$1}/g;
47 $s =~ s/\\([0-7]{1,3})/chr(oct($1))/ge;
52 my ($class, $ibx, $creat, $part) = @_;
53 my $mainrepo = $ibx; # for "public-inbox-index" w/o entry in config
54 my $git_dir = $mainrepo;
58 $mainrepo = $ibx->{mainrepo};
59 $altid = $ibx->{altid};
60 $version = $ibx->{version} || 1;
62 require PublicInbox::AltId;
64 PublicInbox::AltId->new($ibx, $_);
69 $git = PublicInbox::Git->new($git_dir); # v1 only
71 require Search::Xapian::WritableDatabase;
73 mainrepo => $mainrepo,
78 my $perm = $self->_git_config_perm;
79 my $umask = _umask_for($perm);
80 $self->{umask} = $umask;
82 $self->{lock_path} = "$mainrepo/ssoma.lock";
83 } elsif ($version == 2) {
84 defined $part or die "partition is required for v2\n";
85 # partition is a number or "all"
86 $self->{partition} = $part;
87 $self->{lock_path} = undef;
88 $self->{msgmap_path} = "$mainrepo/msgmap.sqlite3";
90 die "unsupported inbox version=$version\n";
92 $self->{creat} = ($creat || 0) == 1;
98 my $xdb = delete $self->{xdb} or croak 'not acquired';
100 _lock_release($self) if $self->{creat};
106 croak 'already acquired' if $self->{xdb};
107 my $dir = $self->xdir;
108 my $flag = Search::Xapian::DB_OPEN;
109 if ($self->{creat}) {
111 _lock_acquire($self);
112 File::Path::mkpath($dir);
113 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
115 $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag);
118 # we only acquire the flock if creating or reindexing;
119 # PublicInbox::Import already has the lock on its own.
122 croak 'already locked' if $self->{lockfh};
123 my $lock_path = $self->{lock_path} or return;
124 sysopen(my $lockfh, $lock_path, O_WRONLY|O_CREAT) or
125 die "failed to open lock $lock_path: $!\n";
126 flock($lockfh, LOCK_EX) or die "lock failed: $!\n";
127 $self->{lockfh} = $lockfh;
132 return unless $self->{lock_path};
133 my $lockfh = delete $self->{lockfh} or croak 'not locked';
134 flock($lockfh, LOCK_UN) or die "unlock failed: $!\n";
135 close $lockfh or die "close failed: $!\n";
139 my ($doc, $col, $num) = @_;
140 $num = Search::Xapian::sortable_serialise($num);
141 $doc->add_value($col, $num);
144 sub add_values ($$$$) {
145 my ($smsg, $bytes, $num, $lines) = @_;
148 my $doc = $smsg->{doc};
149 add_val($doc, &PublicInbox::Search::TS, $ts);
151 defined($num) and add_val($doc, &PublicInbox::Search::NUM, $num);
153 defined($bytes) and add_val($doc, &PublicInbox::Search::BYTES, $bytes);
155 add_val($doc, &PublicInbox::Search::LINES, $lines);
157 my $yyyymmdd = strftime('%Y%m%d', gmtime($ts));
158 add_val($doc, PublicInbox::Search::YYYYMMDD, $yyyymmdd);
161 sub index_users ($$) {
162 my ($tg, $smsg) = @_;
164 my $from = $smsg->from;
168 $tg->index_text($from, 1, 'A'); # A - author
169 $tg->increase_termpos;
170 $tg->index_text($to, 1, 'XTO') if $to ne '';
171 $tg->increase_termpos;
172 $tg->index_text($cc, 1, 'XCC') if $cc ne '';
173 $tg->increase_termpos;
176 sub index_text_inc ($$$) {
177 my ($tg, $text, $pfx) = @_;
178 $tg->index_text($text, 1, $pfx);
179 $tg->increase_termpos;
182 sub index_old_diff_fn {
183 my ($tg, $seen, $fa, $fb) = @_;
185 # no renames or space support for traditional diffs,
186 # find the number of leading common paths to strip:
187 my @fa = split('/', $fa);
188 my @fb = split('/', $fb);
189 while (scalar(@fa) && scalar(@fb)) {
190 $fa = join('/', @fa);
191 $fb = join('/', @fb);
193 index_text_inc($tg, $fa,'XDFN') unless $seen->{$fa}++;
202 sub index_diff ($$$) {
203 my ($tg, $lines, $doc) = @_;
207 if ($in_diff && s/^ //) { # diff context
208 index_text_inc($tg, $_, 'XDFCTX');
209 } elsif (/^-- $/) { # email signature begins
211 } elsif (m!^diff --git ("?a/.+) ("?b/.+)\z!) {
212 my ($fa, $fb) = ($1, $2);
213 my $fn = (split('/', git_unquote($fa), 2))[1];
214 index_text_inc($tg, $fn, 'XDFN') unless $seen{$fn}++;
215 $fn = (split('/', git_unquote($fb), 2))[1];
216 index_text_inc($tg, $fn, 'XDFN') unless $seen{$fn}++;
219 } elsif (m/^diff -(.+) (\S+) (\S+)$/) {
220 my ($opt, $fa, $fb) = ($1, $2, $3);
221 # only support unified:
222 next unless $opt =~ /[uU]/;
223 $in_diff = index_old_diff_fn($tg, \%seen, $fa, $fb);
224 } elsif (m!^--- ("?a/.+)!) {
225 my $fn = (split('/', git_unquote($1), 2))[1];
226 index_text_inc($tg, $fn, 'XDFN') unless $seen{$fn}++;
228 } elsif (m!^\+\+\+ ("?b/.+)!) {
229 my $fn = (split('/', git_unquote($1), 2))[1];
230 index_text_inc($tg, $fn, 'XDFN') unless $seen{$fn}++;
232 } elsif (/^--- (\S+)/) {
234 } elsif (defined $in_diff && /^\+\+\+ (\S+)/) {
235 $in_diff = index_old_diff_fn($tg, \%seen, $in_diff, $1);
236 } elsif ($in_diff && s/^\+//) { # diff added
237 index_text_inc($tg, $_, 'XDFB');
238 } elsif ($in_diff && s/^-//) { # diff removed
239 index_text_inc($tg, $_, 'XDFA');
240 } elsif (m!^index ([a-f0-9]+)\.\.([a-f0-9]+)!) {
241 my ($ba, $bb) = ($1, $2);
242 index_git_blob_id($doc, 'XDFPRE', $ba);
243 index_git_blob_id($doc, 'XDFPOST', $bb);
245 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*$/) {
246 # traditional diff w/o -p
247 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*(\S+.*)$/) {
248 # hunk header context
249 index_text_inc($tg, $1, 'XDFHH');
250 # ignore the following lines:
251 } elsif (/^(?:dis)similarity index/) {
252 } elsif (/^(?:old|new) mode/) {
253 } elsif (/^(?:deleted|new) file mode/) {
254 } elsif (/^(?:copy|rename) (?:from|to) /) {
255 } elsif (/^(?:dis)?similarity index /) {
256 } elsif (/^\\ No newline at end of file/) {
257 } elsif (/^Binary files .* differ/) {
261 warn "non-diff line: $_\n" if DEBUG && $_ ne '';
267 sub index_body ($$$) {
268 my ($tg, $lines, $doc) = @_;
269 my $txt = join("\n", @$lines);
270 $tg->index_text($txt, !!$doc, $doc ? 'XNQ' : 'XQUOT');
271 $tg->increase_termpos;
272 # does it look like a diff?
273 if ($doc && $txt =~ /^(?:diff|---|\+\+\+) /ms) {
275 index_diff($tg, $lines, $doc);
281 my ($self, $mime, $bytes, $num, $blob) = @_; # mime = Email::MIME object
282 my $db = $self->{xdb};
284 my ($doc_id, $old_tid);
285 my $mid = mid_clean(mid_mime($mime));
286 my $threader = $self->{threader};
289 die 'Message-ID too long' if length($mid) > MAX_MID_SIZE;
290 my $smsg = $self->lookup_message($mid);
292 # convert a ghost to a regular message
293 # it will also clobber any existing regular message
294 $doc_id = $smsg->{doc_id};
295 $old_tid = $smsg->thread_id unless $threader;
297 $smsg = PublicInbox::SearchMsg->new($mime);
298 my $doc = $smsg->{doc};
299 $doc->add_term('XMID' . $mid);
301 my $subj = $smsg->subject;
304 $xpath = $self->subject_path($subj);
305 $xpath = id_compress($xpath);
306 $doc->add_term('XPATH' . $xpath);
309 my $lines = $mime->body_raw =~ tr!\n!\n!;
310 add_values($smsg, $bytes, $num, $lines);
312 my $tg = $self->term_generator;
314 $tg->set_document($doc);
315 $tg->index_text($subj, 1, 'S') if $subj;
316 $tg->increase_termpos;
318 index_users($tg, $smsg);
320 msg_iter($mime, sub {
321 my ($part, $depth, @idx) = @{$_[0]};
322 my $ct = $part->content_type || 'text/plain';
323 my $fn = $part->filename;
324 if (defined $fn && $fn ne '') {
325 $tg->index_text($fn, 1, 'XFN');
328 return if $ct =~ m!\btext/x?html\b!i;
330 my $s = eval { $part->body_str };
332 if ($ct =~ m!\btext/plain\b!i) {
333 # Try to assume UTF-8 because Alpine
334 # seems to do wacky things and set
336 $part->charset_set('UTF-8');
337 $s = eval { $part->body_str };
338 $s = $part->body if $@;
341 defined $s or return;
344 my $body = $part->body;
345 my @lines = split(/\n/, $body);
346 while (defined(my $l = shift @lines)) {
348 index_body($tg, \@orig, $doc) if @orig;
351 index_body($tg, \@quot, 0) if @quot;
355 index_body($tg, \@quot, 0) if @quot;
356 index_body($tg, \@orig, $doc) if @orig;
359 # populates smsg->references for smsg->to_doc_data
360 my $refs = parse_references($smsg);
361 my $data = $smsg->to_doc_data($blob);
363 $threader->thread_msg($mid, $smsg->ts, $xpath, $data);
365 link_message($self, $smsg, $refs, $old_tid);
367 $tg->index_text($mid, 1, 'XM');
368 $doc->set_data($data);
370 if (my $altid = $self->{-altid}) {
371 foreach my $alt (@$altid) {
372 my $id = $alt->mid2alt($mid);
373 next unless defined $id;
374 $doc->add_term($alt->{xprefix} . $id);
377 if (defined $doc_id) {
378 $db->replace_document($doc_id, $doc);
380 $doc_id = $db->add_document($doc);
385 warn "failed to index message <$mid>: $@\n";
391 # returns deleted doc_id on success, undef on missing
393 my ($self, $mid) = @_;
394 my $db = $self->{xdb};
396 $mid = mid_clean($mid);
399 my ($head, $tail) = $self->find_doc_ids('XMID' . $mid);
400 if ($head->equal($tail)) {
401 warn "cannot remove non-existent <$mid>\n";
403 for (; $head != $tail; $head->inc) {
404 my $docid = $head->get_docid;
405 $db->delete_document($docid);
410 warn "failed to remove message <$mid>: $@\n";
416 sub term_generator { # write-only
419 my $tg = $self->{term_generator};
422 $tg = Search::Xapian::TermGenerator->new;
423 $tg->set_stemmer($self->stemmer);
425 $self->{term_generator} = $tg;
428 # increments last_thread_id counter
429 # returns a 64-bit integer represented as a decimal string
432 my $db = $self->{xdb};
433 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
435 $db->set_metadata('last_thread_id', ++$last_thread_id);
440 sub parse_references ($) {
442 my $doc = $smsg->{doc};
443 my $mid = $smsg->mid;
444 my $mime = $smsg->{mime};
445 my $hdr = $mime->header_obj;
447 # last References should be IRT, but some mail clients do things
448 # out of order, so trust IRT over References iff IRT exists
449 my @refs = (($hdr->header_raw('References') || '') =~ /<([^>]+)>/g);
450 push(@refs, (($hdr->header_raw('In-Reply-To') || '') =~ /<([^>]+)>/g));
453 my %uniq = ($mid => 1);
454 my @orig_refs = @refs;
457 # prevent circular references via References: here:
458 foreach my $ref (@orig_refs) {
459 if (length($ref) > MAX_MID_SIZE) {
460 warn "References: <$ref> too long, ignoring\n";
467 $smsg->{references} = '<'.join('> <', @refs).'>' if @refs;
472 my ($self, $smsg, $refs, $old_tid) = @_;
477 # first ref *should* be the thread root,
478 # but we can never trust clients to do the right thing
479 my $ref = shift @$refs;
480 $tid = $self->_resolve_mid_to_tid($ref);
481 $self->merge_threads($tid, $old_tid) if defined $old_tid;
483 # the rest of the refs should point to this tid:
484 foreach $ref (@$refs) {
485 my $ptid = $self->_resolve_mid_to_tid($ref);
486 merge_threads($self, $tid, $ptid);
489 $tid = defined $old_tid ? $old_tid : $self->next_thread_id;
491 $smsg->{doc}->add_term('G' . $tid);
495 my ($self, $mime, $bytes, $num, $blob) = @_;
496 $self->add_message($mime, $bytes, $num, $blob);
499 sub index_git_blob_id {
500 my ($doc, $pfx, $objid) = @_;
502 my $len = length($objid);
503 for (my $len = length($objid); $len >= 7; ) {
504 $doc->add_term($pfx.$objid);
505 $objid = substr($objid, 0, --$len);
510 my ($self, $mime) = @_;
511 my $mid = eval { mid_clean(mid_mime($mime)) };
512 $self->remove_message($mid) if defined $mid;
516 my ($self, $mime, $warn_existing) = @_;
517 my $mid = mid_clean(mid_mime($mime));
518 my $mm = $self->{mm};
519 my $num = $mm->mid_insert($mid);
520 return $num if defined $num;
522 warn "<$mid> reused\n" if $warn_existing;
523 # fallback to num_for since filters like RubyLang set the number
528 my ($self, $mime) = @_;
529 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
533 my ($self, $mime, $bytes, $blob) = @_;
534 my $num = $self->{mm}->num_for(mid_clean(mid_mime($mime)));
535 index_blob($self, $mime, $bytes, $num, $blob);
539 my ($self, $mime) = @_;
540 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
541 unindex_blob($self, $mime);
545 my ($self, $mime, $bytes, $blob) = @_;
546 my $num = index_mm($self, $mime);
547 index_blob($self, $mime, $bytes, $num, $blob);
551 my ($self, $mime) = @_;
552 unindex_blob($self, $mime);
553 unindex_mm($self, $mime);
557 my ($git, $blob, $sizeref) = @_;
559 my $str = $git->cat_file($blob, $sizeref);
560 # fixup bugs from import:
561 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
562 PublicInbox::MIME->new($str);
568 my ($self, $opts) = @_;
569 with_umask($self, sub { $self->_index_sync($opts) });
572 sub batch_adjust ($$$$) {
573 my ($max, $bytes, $batch_cb, $latest) = @_;
577 $batch_cb->($latest, 1);
583 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
584 my $hex = '[a-f0-9]';
585 my $h40 = $hex .'{40}';
586 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
587 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
588 my $git = $self->{git};
591 my $max = BATCH_BYTES;
594 while (defined($line = <$log>)) {
595 if ($line =~ /$addmsg/o) {
597 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
598 batch_adjust(\$max, $bytes, $batch_cb, $latest);
599 $add_cb->($self, $mime, $bytes, $blob);
600 } elsif ($line =~ /$delmsg/o) {
602 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
603 batch_adjust(\$max, $bytes, $batch_cb, $latest);
604 $del_cb->($self, $mime);
605 } elsif ($line =~ /^commit ($h40)/o) {
609 $batch_cb->($latest, 0);
614 $self->{mm} ||= eval {
615 require PublicInbox::Msgmap;
616 my $msgmap_path = $self->{msgmap_path};
617 if (defined $msgmap_path) { # v2
618 PublicInbox::Msgmap->new_file($msgmap_path, 1);
620 PublicInbox::Msgmap->new($self->{mainrepo}, 1);
626 my ($self, $range) = @_;
627 $self->{git}->popen(qw/log --reverse --no-notes --no-color
628 --raw -r --no-abbrev/, $range);
631 # indexes all unindexed messages
633 my ($self, $opts) = @_;
634 my $tip = $opts->{ref} || 'HEAD';
635 my $reindex = $opts->{reindex};
636 my ($mkey, $last_commit, $lx, $xlog);
637 $self->{git}->batch_prepare;
638 my $xdb = _xdb_acquire($self);
639 $xdb->begin_transaction;
642 $mkey = 'last_commit';
643 $last_commit = $xdb->get_metadata('last_commit');
647 $mkey = undef if $last_commit ne '';
649 $xdb->cancel_transaction;
650 $xdb = _xdb_release($self);
652 # ensure we leak no FDs to "git log"
653 my $range = $lx eq '' ? $tip : "$lx..$tip";
654 $xlog = _git_log($self, $range);
656 $xdb = _xdb_acquire($self);
657 $xdb->begin_transaction;
658 } while ($xdb->get_metadata('last_commit') ne $last_commit);
660 my $mm = _msgmap_init($self);
661 my $dbh = $mm->{dbh} if $mm;
664 my ($commit, $more) = @_;
666 $mm->last_commit($commit) if $commit;
670 $xdb->set_metadata($mkey, $commit) if $mkey && $commit;
671 $xdb->commit_transaction;
672 $xdb = _xdb_release($self);
674 # let another process do some work... <
677 $xdb = _xdb_acquire($self);
678 $xdb->begin_transaction;
680 $dbh->begin_work if $dbh;
686 my $lm = $mm->last_commit || '';
688 # Common case is the indexes are synced,
689 # we only need to run git-log once:
690 rlog($self, $xlog, *index_both, *unindex_both, $cb);
692 # Uncommon case, msgmap and xapian are out-of-sync
693 # do not care for performance (but git is fast :>)
694 # This happens if we have to reindex Xapian since
695 # msgmap is a frozen format and our Xapian format
697 my $r = $lm eq '' ? $tip : "$lm..$tip";
699 # first, ensure msgmap is up-to-date:
700 my $mkey_prev = $mkey;
701 $mkey = undef; # ignore xapian, for now
702 my $mlog = _git_log($self, $r);
704 rlog($self, $mlog, *index_mm, *unindex_mm, $cb);
705 $mm_only = $mlog = undef;
707 # now deal with Xapian
710 rlog($self, $xlog, *index_mm2, *unindex_mm2, $cb);
713 # user didn't install DBD::SQLite and DBI
714 rlog($self, $xlog, *index_blob, *unindex_blob, $cb);
718 # this will create a ghost as necessary
719 sub _resolve_mid_to_tid {
720 my ($self, $mid) = @_;
722 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
727 my ($self, $mid) = @_;
729 my $tid = $self->next_thread_id;
730 my $doc = Search::Xapian::Document->new;
731 $doc->add_term('XMID' . $mid);
732 $doc->add_term('G' . $tid);
733 $doc->add_term('T' . 'ghost');
735 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
736 $self->{xdb}->add_document($doc);
742 my ($self, $winner_tid, $loser_tid) = @_;
743 return if $winner_tid == $loser_tid;
744 my ($head, $tail) = $self->find_doc_ids('G' . $loser_tid);
745 my $db = $self->{xdb};
747 for (; $head != $tail; $head->inc) {
748 my $docid = $head->get_docid;
749 my $doc = $db->get_document($docid);
750 $doc->remove_term('G' . $loser_tid);
751 $doc->add_term('G' . $winner_tid);
752 $db->replace_document($docid, $doc);
756 sub _read_git_config_perm {
758 my @cmd = qw(config);
759 if ($self->{version} == 2) {
760 push @cmd, "--file=$self->{mainrepo}/inbox-config";
762 my $fh = $self->{git}->popen(@cmd, 'core.sharedRepository');
765 chomp $perm if defined $perm;
769 sub _git_config_perm {
771 my $perm = scalar @_ ? $_[0] : _read_git_config_perm($self);
772 return PERM_GROUP if (!defined($perm) || $perm eq '');
773 return PERM_UMASK if ($perm eq 'umask');
774 return PERM_GROUP if ($perm eq 'group');
775 if ($perm =~ /\A(?:all|world|everybody)\z/) {
776 return PERM_EVERYBODY;
778 return PERM_GROUP if ($perm =~ /\A(?:true|yes|on|1)\z/);
779 return PERM_UMASK if ($perm =~ /\A(?:false|no|off|0)\z/);
782 return PERM_UMASK if ($i == PERM_UMASK);
783 return PERM_GROUP if ($i == OLD_PERM_GROUP);
784 return PERM_EVERYBODY if ($i == OLD_PERM_EVERYBODY);
786 if (($i & 0600) != 0600) {
787 die "core.sharedRepository mode invalid: ".
788 sprintf('%.3o', $i) . "\nOwner must have permissions\n";
794 my ($perm) = @_; # _git_config_perm return value
796 return umask if $rv == 0;
798 # set +x bit if +r or +w were set
799 $rv |= 0100 if ($rv & 0600);
800 $rv |= 0010 if ($rv & 0060);
801 $rv |= 0001 if ($rv & 0006);
806 my ($self, $cb) = @_;
807 my $old = umask $self->{umask};
808 my $rv = eval { $cb->() };
816 # order matters for unlocking
817 $_[0]->{xdb} = undef;
818 $_[0]->{lockfh} = undef;
821 # remote_* subs are only used by SearchIdxPart and SearchIdxThread:
824 print { $self->{w} } "commit\n" or die "failed to write commit: $!";
829 my $pid = delete $self->{pid} or die "no process to wait on\n";
830 my $w = delete $self->{w} or die "no pipe to write to\n";
831 print $w "close\n" or die "failed to write to pid:$pid: $!\n";
832 close $w or die "failed to close pipe for pid:$pid: $!\n";
833 waitpid($pid, 0) == $pid or die "remote process did not finish";
834 $? == 0 or die ref($self)." exited with: $?";