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 base qw(PublicInbox::Search PublicInbox::Lock);
13 use PublicInbox::MIME;
14 use PublicInbox::InboxWritable;
15 use PublicInbox::MID qw/mid_clean id_compress mid_mime mids/;
16 use PublicInbox::MsgIter;
18 use POSIX qw(strftime);
19 use PublicInbox::OverIdx;
20 use PublicInbox::Spawn qw(spawn);
21 require PublicInbox::Git;
22 use Compress::Zlib qw(compress);
25 BATCH_BYTES => 1_000_000,
26 DEBUG => !!$ENV{DEBUG},
41 return $s unless ($s =~ /\A"(.*)"\z/);
43 $s =~ s/\\([abfnrtv])/$GIT_ESC{$1}/g;
44 $s =~ s/\\([0-7]{1,3})/chr(oct($1))/ge;
49 my ($class, $ibx, $creat, $part) = @_;
50 my $levels = qr/\A(?:full|medium|basic)\z/;
51 my $mainrepo = $ibx; # for "public-inbox-index" w/o entry in config
52 my $git_dir = $mainrepo;
55 my $indexlevel = 'full';
57 $mainrepo = $ibx->{mainrepo};
58 $altid = $ibx->{altid};
59 $version = $ibx->{version} || 1;
61 require PublicInbox::AltId;
63 PublicInbox::AltId->new($ibx, $_);
66 if ($ibx->{indexlevel}) {
67 if ($ibx->{indexlevel} =~ $levels) {
68 $indexlevel = $ibx->{indexlevel};
70 die("Invalid indexlevel $ibx->{indexlevel}\n");
74 $ibx = { mainrepo => $git_dir, version => 1 };
76 $ibx = PublicInbox::InboxWritable->new($ibx);
77 require Search::Xapian::WritableDatabase;
79 mainrepo => $mainrepo,
84 indexlevel => $indexlevel,
88 $self->{lock_path} = "$mainrepo/ssoma.lock";
89 my $dir = $self->xdir;
90 $self->{over} = PublicInbox::OverIdx->new("$dir/over.sqlite3");
91 } elsif ($version == 2) {
92 defined $part or die "partition is required for v2\n";
93 # partition is a number
94 $self->{partition} = $part;
95 $self->{lock_path} = undef;
97 die "unsupported inbox version=$version\n";
99 $self->{creat} = ($creat || 0) == 1;
105 my $xdb = delete $self->{xdb} or croak 'not acquired';
107 $self->lock_release if $self->{creat};
113 croak 'already acquired' if $self->{xdb};
114 my $dir = $self->xdir;
115 my $flag = Search::Xapian::DB_OPEN;
116 if ($self->{creat}) {
119 File::Path::mkpath($dir);
120 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
122 $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag);
126 my ($doc, $col, $num) = @_;
127 $num = Search::Xapian::sortable_serialise($num);
128 $doc->add_value($col, $num);
131 sub index_text ($$$$)
133 my ($self, $field, $n, $text) = @_;
134 my $tg = $self->term_generator;
136 if ($self->{indexlevel} eq 'full') {
137 $tg->index_text($field, $n, $text);
138 $tg->increase_termpos;
140 $tg->index_text_without_positions($field, $n, $text);
144 sub index_users ($$) {
145 my ($self, $smsg) = @_;
147 my $from = $smsg->from;
151 $self->index_text($from, 1, 'A'); # A - author
152 $self->index_text($to, 1, 'XTO') if $to ne '';
153 $self->index_text($cc, 1, 'XCC') if $cc ne '';
156 sub index_diff_inc ($$$$) {
157 my ($self, $text, $pfx, $xnq) = @_;
159 $self->index_text(join("\n", @$xnq), 1, 'XNQ');
162 $self->index_text($text, 1, $pfx);
165 sub index_old_diff_fn {
166 my ($self, $seen, $fa, $fb, $xnq) = @_;
168 # no renames or space support for traditional diffs,
169 # find the number of leading common paths to strip:
170 my @fa = split('/', $fa);
171 my @fb = split('/', $fb);
172 while (scalar(@fa) && scalar(@fb)) {
173 $fa = join('/', @fa);
174 $fb = join('/', @fb);
176 unless ($seen->{$fa}++) {
177 $self->index_diff_inc($fa, 'XDFN', $xnq);
187 sub index_diff ($$$) {
188 my ($self, $lines, $doc) = @_;
194 if ($in_diff && s/^ //) { # diff context
195 $self->index_diff_inc($_, 'XDFCTX', $xnq);
196 } elsif (/^-- $/) { # email signature begins
198 } elsif (m!^diff --git ("?a/.+) ("?b/.+)\z!) {
199 my ($fa, $fb) = ($1, $2);
200 my $fn = (split('/', git_unquote($fa), 2))[1];
201 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
202 $fn = (split('/', git_unquote($fb), 2))[1];
203 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
206 } elsif (m/^diff -(.+) (\S+) (\S+)$/) {
207 my ($opt, $fa, $fb) = ($1, $2, $3);
209 # only support unified:
210 next unless $opt =~ /[uU]/;
211 $in_diff = $self->index_old_diff_fn(\%seen, $fa, $fb,
213 } elsif (m!^--- ("?a/.+)!) {
214 my $fn = (split('/', git_unquote($1), 2))[1];
215 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
217 } elsif (m!^\+\+\+ ("?b/.+)!) {
218 my $fn = (split('/', git_unquote($1), 2))[1];
219 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
221 } elsif (/^--- (\S+)/) {
224 } elsif (defined $in_diff && /^\+\+\+ (\S+)/) {
225 $in_diff = $self->index_old_diff_fn(\%seen, $in_diff, $1,
227 } elsif ($in_diff && s/^\+//) { # diff added
228 $self->index_diff_inc($_, 'XDFB', $xnq);
229 } elsif ($in_diff && s/^-//) { # diff removed
230 $self->index_diff_inc($_, 'XDFA', $xnq);
231 } elsif (m!^index ([a-f0-9]+)\.\.([a-f0-9]+)!) {
232 my ($ba, $bb) = ($1, $2);
233 index_git_blob_id($doc, 'XDFPRE', $ba);
234 index_git_blob_id($doc, 'XDFPOST', $bb);
236 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*$/) {
237 # traditional diff w/o -p
238 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*(\S+.*)$/) {
239 # hunk header context
240 $self->index_diff_inc($1, 'XDFHH', $xnq);
241 # ignore the following lines:
242 } elsif (/^(?:dis)similarity index/ ||
243 /^(?:old|new) mode/ ||
244 /^(?:deleted|new) file mode/ ||
245 /^(?:copy|rename) (?:from|to) / ||
246 /^(?:dis)?similarity index / ||
247 /^\\ No newline at end of file/ ||
248 /^Binary files .* differ/) {
254 warn "non-diff line: $_\n" if DEBUG && $_ ne '';
259 $self->index_text(join("\n", @xnq), 1, 'XNQ');
262 sub index_body ($$$) {
263 my ($self, $lines, $doc) = @_;
264 my $txt = join("\n", @$lines);
266 # does it look like a diff?
267 if ($txt =~ /^(?:diff|---|\+\+\+) /ms) {
269 $self->index_diff($lines, $doc);
271 $self->index_text($txt, 1, 'XNQ');
274 $self->index_text($txt, 0, 'XQUOT');
279 sub add_xapian ($$$$$) {
280 my ($self, $mime, $num, $oid, $mids, $mid0) = @_;
281 my $smsg = PublicInbox::SearchMsg->new($mime);
282 my $doc = $smsg->{doc};
283 my $subj = $smsg->subject;
284 add_val($doc, PublicInbox::Search::TS(), $smsg->ts);
285 my @ds = gmtime($smsg->ds);
286 my $yyyymmdd = strftime('%Y%m%d', @ds);
287 add_val($doc, PublicInbox::Search::YYYYMMDD(), $yyyymmdd);
288 my $dt = strftime('%Y%m%d%H%M%S', @ds);
289 add_val($doc, PublicInbox::Search::DT(), $dt);
291 my $tg = $self->term_generator;
293 $tg->set_document($doc);
294 $self->index_text($subj, 1, 'S') if $subj;
295 $self->index_users($smsg);
297 msg_iter($mime, sub {
298 my ($part, $depth, @idx) = @{$_[0]};
299 my $ct = $part->content_type || 'text/plain';
300 my $fn = $part->filename;
301 if (defined $fn && $fn ne '') {
302 $self->index_text($fn, 1, 'XFN');
305 return if $ct =~ m!\btext/x?html\b!i;
307 my $s = eval { $part->body_str };
309 if ($ct =~ m!\btext/plain\b!i) {
310 # Try to assume UTF-8 because Alpine
311 # seems to do wacky things and set
313 $part->charset_set('UTF-8');
314 $s = eval { $part->body_str };
315 $s = $part->body if $@;
318 defined $s or return;
321 my $body = $part->body;
322 my @lines = split(/\n/, $body);
323 while (defined(my $l = shift @lines)) {
325 $self->index_body(\@orig, $doc) if @orig;
328 $self->index_body(\@quot, 0) if @quot;
332 $self->index_body(\@quot, 0) if @quot;
333 $self->index_body(\@orig, $doc) if @orig;
336 foreach my $mid (@$mids) {
337 $self->index_text($mid, 1, 'XM');
339 # because too many Message-IDs are prefixed with
341 if ($mid =~ /\w{12,}/) {
342 my @long = ($mid =~ /(\w{3,}+)/g);
343 $self->index_text(join(' ', @long), 1, 'XM');
346 $smsg->{to} = $smsg->{cc} = '';
347 PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids);
348 my $data = $smsg->to_doc_data($oid, $mid0);
349 $doc->set_data($data);
350 if (my $altid = $self->{-altid}) {
351 foreach my $alt (@$altid) {
352 my $pfx = $alt->{xprefix};
353 foreach my $mid (@$mids) {
354 my $id = $alt->mid2alt($mid);
355 next unless defined $id;
356 $doc->add_boolean_term($pfx . $id);
360 $doc->add_boolean_term('Q' . $_) foreach @$mids;
361 $self->{xdb}->replace_document($num, $doc);
365 # mime = Email::MIME object
366 my ($self, $mime, $bytes, $num, $oid, $mid0) = @_;
367 my $xapianlevels = qr/\A(?:full|medium)\z/;
368 my $mids = mids($mime->header_obj);
369 $mid0 = $mids->[0] unless defined $mid0; # v1 compatibility
370 unless (defined $num) { # v1
372 $num = index_mm($self, $mime);
375 if ($self->{indexlevel} =~ $xapianlevels) {
376 $self->add_xapian($mime, $num, $oid, $mids, $mid0)
378 if (my $over = $self->{over}) {
379 $over->add_overview($mime, $bytes, $num, $oid, $mid0);
384 warn "failed to index message <".join('> <',@$mids).">: $@\n";
390 # returns begin and end PostingIterator
392 my ($self, $termval) = @_;
393 my $db = $self->{xdb};
395 ($db->postlist_begin($termval), $db->postlist_end($termval));
399 my ($self, $termval, $cb) = @_;
400 my $batch_size = 1000; # don't let @ids grow too large to avoid OOM
402 my ($head, $tail) = $self->find_doc_ids($termval);
403 return if $head == $tail;
405 for (; $head != $tail && @ids < $batch_size; $head->inc) {
406 push @ids, $head->get_docid;
413 my ($self, $mid) = @_;
414 my $db = $self->{xdb};
416 $mid = mid_clean($mid);
417 my $over = $self->{over};
420 batch_do($self, 'Q' . $mid, sub {
422 $db->delete_document($_) for @$ids;
423 $over->delete_articles($ids) if $over;
428 warn "failed to remove message <$mid>: $@\n";
430 warn "cannot remove non-existent <$mid>\n";
434 # MID is a hint in V2
436 my ($self, $oid, $mid) = @_;
437 my $db = $self->{xdb};
439 $self->{over}->remove_oid($oid, $mid) if $self->{over};
441 # XXX careful, we cannot use batch_do here since we conditionally
442 # delete documents based on other factors, so we cannot call
443 # find_doc_ids twice.
444 my ($head, $tail) = $self->find_doc_ids('Q' . $mid);
445 return if $head == $tail;
447 # there is only ONE element in @delete unless we
448 # have bugs in our v2writable deduplication check
450 for (; $head != $tail; $head->inc) {
451 my $docid = $head->get_docid;
452 my $doc = $db->get_document($docid);
453 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
455 if ($smsg->{blob} eq $oid) {
456 push(@delete, $docid);
459 $db->delete_document($_) foreach @delete;
463 sub term_generator { # write-only
466 my $tg = $self->{term_generator};
469 $tg = Search::Xapian::TermGenerator->new;
470 $tg->set_stemmer($self->stemmer);
472 $self->{term_generator} = $tg;
475 sub index_git_blob_id {
476 my ($doc, $pfx, $objid) = @_;
478 my $len = length($objid);
479 for (my $len = length($objid); $len >= 7; ) {
480 $doc->add_term($pfx.$objid);
481 $objid = substr($objid, 0, --$len);
486 my ($self, $mime) = @_;
487 my $mid = eval { mid_clean(mid_mime($mime)) };
488 $self->remove_message($mid) if defined $mid;
492 my ($self, $mime) = @_;
493 my $mid = mid_clean(mid_mime($mime));
494 my $mm = $self->{mm};
497 if (defined $self->{regen_down}) {
498 $num = $mm->num_for($mid) and return $num;
500 while (($num = $self->{regen_down}--) > 0) {
501 if ($mm->mid_set($num, $mid) != 0) {
505 } elsif (defined $self->{regen_up}) {
506 $num = $mm->num_for($mid) and return $num;
508 # this is to fixup old bugs due to add-remove-add
509 while (($num = ++$self->{regen_up})) {
510 if ($mm->mid_set($num, $mid) != 0) {
516 $num = $mm->mid_insert($mid) and return $num;
518 # fallback to num_for since filters like RubyLang set the number
523 my ($self, $mime) = @_;
524 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
528 my ($self, $mime, $bytes, $blob) = @_;
529 my $num = index_mm($self, $mime);
530 add_message($self, $mime, $bytes, $num, $blob);
534 my ($self, $mime) = @_;
535 unindex_blob($self, $mime);
536 unindex_mm($self, $mime);
540 my ($git, $blob, $sizeref) = @_;
542 my $str = $git->cat_file($blob, $sizeref);
543 # fixup bugs from import:
544 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
545 PublicInbox::MIME->new($str);
551 my ($self, $opts) = @_;
552 $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) })
555 sub batch_adjust ($$$$) {
556 my ($max, $bytes, $batch_cb, $latest) = @_;
560 $batch_cb->($latest);
566 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
567 my $hex = '[a-f0-9]';
568 my $h40 = $hex .'{40}';
569 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
570 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
571 my $git = $self->{git};
574 my $max = BATCH_BYTES;
579 while (defined($line = <$log>)) {
580 if ($line =~ /$addmsg/o) {
582 if (delete $D{$blob}) {
583 if (defined $self->{regen_down}) {
584 $self->{regen_down}--;
588 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
589 batch_adjust(\$max, $bytes, $batch_cb, $latest);
590 $add_cb->($self, $mime, $bytes, $blob);
591 } elsif ($line =~ /$delmsg/o) {
594 } elsif ($line =~ /^commit ($h40)/o) {
600 foreach my $blob (keys %D) {
601 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
602 $del_cb->($self, $mime);
604 $batch_cb->($latest, $newest);
609 die "BUG: _msgmap_init is only for v1\n" if $self->{version} != 1;
610 $self->{mm} ||= eval {
611 require PublicInbox::Msgmap;
612 PublicInbox::Msgmap->new($self->{mainrepo}, 1);
617 my ($self, $range) = @_;
618 my $git = $self->{git};
620 if (index($range, '..') < 0) {
622 # can't use 'rev-list --count' if we use --diff-filter
623 my $fh = $git->popen(qw(log --pretty=tformat:%h
624 --no-notes --no-color --no-renames
625 --diff-filter=AM), $range);
626 ++$regen_max while <$fh>;
627 my (undef, $max) = $self->{mm}->minmax;
629 if ($max && $max == $regen_max) {
630 # fix up old bugs in full indexes which caused messages to
631 # not appear in Msgmap
632 $self->{regen_up} = $max;
634 # normal regen is for for fresh data
635 $self->{regen_down} = $regen_max;
639 $git->popen(qw/log --no-notes --no-color --no-renames
640 --raw -r --no-abbrev/, $range);
643 sub is_ancestor ($$$) {
644 my ($git, $cur, $tip) = @_;
645 return 0 unless $git->check($cur);
646 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
647 qw(merge-base --is-ancestor), $cur, $tip ];
648 my $pid = spawn($cmd);
649 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
650 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
654 sub need_update ($$$) {
655 my ($self, $cur, $new) = @_;
656 my $git = $self->{git};
657 return 1 if $cur && !is_ancestor($git, $cur, $new);
658 my $range = $cur eq '' ? $new : "$cur..$new";
659 chomp(my $n = $git->qx(qw(rev-list --count), $range));
660 ($n eq '' || $n > 0);
663 # indexes all unindexed messages (v1 only)
665 my ($self, $opts) = @_;
666 my $tip = $opts->{ref} || 'HEAD';
667 my $reindex = $opts->{reindex};
668 my ($mkey, $last_commit, $lx, $xlog);
669 my $git = $self->{git};
672 my $xdb = $self->begin_txn_lazy;
673 my $mm = _msgmap_init($self);
676 $mkey = 'last_commit';
677 $last_commit = $xdb->get_metadata('last_commit');
681 $mkey = undef if $last_commit ne '';
684 # use last_commit from msgmap if it is older or unset
685 my $lm = $mm->last_commit || '';
686 if (!$lm || ($lm && $lx && is_ancestor($git, $lm, $lx))) {
690 $self->{over}->rollback_lazy;
691 $self->{over}->disconnect;
694 $xdb->cancel_transaction;
695 $xdb = _xdb_release($self);
697 # ensure we leak no FDs to "git log" with Xapian <= 1.2
698 my $range = $lx eq '' ? $tip : "$lx..$tip";
699 $xlog = _git_log($self, $range);
701 $xdb = $self->begin_txn_lazy;
702 } while ($xdb->get_metadata('last_commit') ne $last_commit);
704 my $dbh = $mm->{dbh} if $mm;
706 my ($commit, $newest) = @_;
709 my $cur = $mm->last_commit || '';
710 if (need_update($self, $cur, $newest)) {
711 $mm->last_commit($newest);
716 if ($mkey && $newest) {
717 my $cur = $xdb->get_metadata($mkey);
718 if (need_update($self, $cur, $newest)) {
719 $xdb->set_metadata($mkey, $newest);
722 $self->commit_txn_lazy;
724 $xdb = _xdb_release($self);
725 # let another process do some work... <
727 $xdb = $self->begin_txn_lazy;
728 $dbh->begin_work if $dbh;
733 read_log($self, $xlog, *index_both, *unindex_both, $cb);
737 # order matters for unlocking
738 $_[0]->{xdb} = undef;
739 $_[0]->{lockfh} = undef;
742 # remote_* subs are only used by SearchIdxPart
745 if (my $w = $self->{w}) {
746 print $w "commit\n" or die "failed to write commit: $!";
748 $self->commit_txn_lazy;
754 if (my $w = delete $self->{w}) {
755 my $pid = delete $self->{pid} or die "no process to wait on\n";
756 print $w "close\n" or die "failed to write to pid:$pid: $!\n";
757 close $w or die "failed to close pipe for pid:$pid: $!\n";
758 waitpid($pid, 0) == $pid or die "remote process did not finish";
759 $? == 0 or die ref($self)." pid:$pid exited with: $?";
761 die "transaction in progress $self\n" if $self->{txn};
762 $self->_xdb_release if $self->{xdb};
767 my ($self, $oid, $mid) = @_;
768 if (my $w = $self->{w}) {
769 # triggers remove_by_oid in a partition
770 print $w "D $oid $mid\n" or die "failed to write remove $!";
772 $self->begin_txn_lazy;
773 $self->remove_by_oid($oid, $mid);
779 return if $self->{txn};
781 $self->{-inbox}->with_umask(sub {
782 my $xdb = $self->{xdb} || $self->_xdb_acquire;
783 $self->{over}->begin_lazy if $self->{over};
784 $xdb->begin_transaction;
790 sub commit_txn_lazy {
792 delete $self->{txn} or return;
793 $self->{-inbox}->with_umask(sub {
794 $self->{xdb}->commit_transaction;
795 $self->{over}->commit_lazy if $self->{over};
801 die "$$ $0 xdb not released\n" if $self->{xdb};
802 die "$$ $0 still in transaction\n" if $self->{txn};