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 $mainrepo = $ibx; # for "public-inbox-index" w/o entry in config
51 my $git_dir = $mainrepo;
54 my $indexlevel = 'full';
56 $mainrepo = $ibx->{mainrepo};
57 $altid = $ibx->{altid};
58 $version = $ibx->{version} || 1;
60 require PublicInbox::AltId;
62 PublicInbox::AltId->new($ibx, $_);
66 $ibx = { mainrepo => $git_dir, version => 1 };
68 $ibx = PublicInbox::InboxWritable->new($ibx);
69 require Search::Xapian::WritableDatabase;
71 mainrepo => $mainrepo,
76 indexlevel => $indexlevel,
80 $self->{lock_path} = "$mainrepo/ssoma.lock";
81 my $dir = $self->xdir;
82 $self->{over} = PublicInbox::OverIdx->new("$dir/over.sqlite3");
83 } elsif ($version == 2) {
84 defined $part or die "partition is required for v2\n";
85 # partition is a number
86 $self->{partition} = $part;
87 $self->{lock_path} = undef;
89 die "unsupported inbox version=$version\n";
91 $self->{creat} = ($creat || 0) == 1;
97 my $xdb = delete $self->{xdb} or croak 'not acquired';
99 $self->lock_release if $self->{creat};
105 croak 'already acquired' if $self->{xdb};
106 my $dir = $self->xdir;
107 my $flag = Search::Xapian::DB_OPEN;
108 if ($self->{creat}) {
111 File::Path::mkpath($dir);
112 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
114 $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag);
118 my ($doc, $col, $num) = @_;
119 $num = Search::Xapian::sortable_serialise($num);
120 $doc->add_value($col, $num);
123 sub index_text ($$$$)
125 my ($self, $field, $n, $text) = @_;
126 my $tg = $self->term_generator;
128 if ($self->{indexlevel} eq 'full') {
129 $tg->index_text($field, $n, $text);
130 $tg->increase_termpos;
132 $tg->index_text_without_positions($field, $n, $text);
136 sub index_users ($$) {
137 my ($self, $smsg) = @_;
139 my $from = $smsg->from;
143 $self->index_text($from, 1, 'A'); # A - author
144 $self->index_text($to, 1, 'XTO') if $to ne '';
145 $self->index_text($cc, 1, 'XCC') if $cc ne '';
148 sub index_diff_inc ($$$$) {
149 my ($self, $text, $pfx, $xnq) = @_;
151 $self->index_text(join("\n", @$xnq), 1, 'XNQ');
154 $self->index_text($text, 1, $pfx);
157 sub index_old_diff_fn {
158 my ($self, $seen, $fa, $fb, $xnq) = @_;
160 # no renames or space support for traditional diffs,
161 # find the number of leading common paths to strip:
162 my @fa = split('/', $fa);
163 my @fb = split('/', $fb);
164 while (scalar(@fa) && scalar(@fb)) {
165 $fa = join('/', @fa);
166 $fb = join('/', @fb);
168 unless ($seen->{$fa}++) {
169 $self->index_diff_inc($fa, 'XDFN', $xnq);
179 sub index_diff ($$$) {
180 my ($self, $lines, $doc) = @_;
186 if ($in_diff && s/^ //) { # diff context
187 $self->index_diff_inc($_, 'XDFCTX', $xnq);
188 } elsif (/^-- $/) { # email signature begins
190 } elsif (m!^diff --git ("?a/.+) ("?b/.+)\z!) {
191 my ($fa, $fb) = ($1, $2);
192 my $fn = (split('/', git_unquote($fa), 2))[1];
193 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
194 $fn = (split('/', git_unquote($fb), 2))[1];
195 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
198 } elsif (m/^diff -(.+) (\S+) (\S+)$/) {
199 my ($opt, $fa, $fb) = ($1, $2, $3);
201 # only support unified:
202 next unless $opt =~ /[uU]/;
203 $in_diff = $self->index_old_diff_fn(\%seen, $fa, $fb,
205 } elsif (m!^--- ("?a/.+)!) {
206 my $fn = (split('/', git_unquote($1), 2))[1];
207 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
209 } elsif (m!^\+\+\+ ("?b/.+)!) {
210 my $fn = (split('/', git_unquote($1), 2))[1];
211 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
213 } elsif (/^--- (\S+)/) {
216 } elsif (defined $in_diff && /^\+\+\+ (\S+)/) {
217 $in_diff = $self->index_old_diff_fn(\%seen, $in_diff, $1,
219 } elsif ($in_diff && s/^\+//) { # diff added
220 $self->index_diff_inc($_, 'XDFB', $xnq);
221 } elsif ($in_diff && s/^-//) { # diff removed
222 $self->index_diff_inc($_, 'XDFA', $xnq);
223 } elsif (m!^index ([a-f0-9]+)\.\.([a-f0-9]+)!) {
224 my ($ba, $bb) = ($1, $2);
225 index_git_blob_id($doc, 'XDFPRE', $ba);
226 index_git_blob_id($doc, 'XDFPOST', $bb);
228 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*$/) {
229 # traditional diff w/o -p
230 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*(\S+.*)$/) {
231 # hunk header context
232 $self->index_diff_inc($1, 'XDFHH', $xnq);
233 # ignore the following lines:
234 } elsif (/^(?:dis)similarity index/ ||
235 /^(?:old|new) mode/ ||
236 /^(?:deleted|new) file mode/ ||
237 /^(?:copy|rename) (?:from|to) / ||
238 /^(?:dis)?similarity index / ||
239 /^\\ No newline at end of file/ ||
240 /^Binary files .* differ/) {
246 warn "non-diff line: $_\n" if DEBUG && $_ ne '';
251 $self->index_text(join("\n", @xnq), 1, 'XNQ');
254 sub index_body ($$$) {
255 my ($self, $lines, $doc) = @_;
256 my $txt = join("\n", @$lines);
258 # does it look like a diff?
259 if ($txt =~ /^(?:diff|---|\+\+\+) /ms) {
261 $self->index_diff($lines, $doc);
263 $self->index_text($txt, 1, 'XNQ');
266 $self->index_text($txt, 0, 'XQUOT');
271 sub add_xapian ($$$$$) {
272 my ($self, $mime, $num, $oid, $mids, $mid0) = @_;
273 my $smsg = PublicInbox::SearchMsg->new($mime);
274 my $doc = $smsg->{doc};
275 my $subj = $smsg->subject;
276 add_val($doc, PublicInbox::Search::TS(), $smsg->ts);
277 my @ds = gmtime($smsg->ds);
278 my $yyyymmdd = strftime('%Y%m%d', @ds);
279 add_val($doc, PublicInbox::Search::YYYYMMDD(), $yyyymmdd);
280 my $dt = strftime('%Y%m%d%H%M%S', @ds);
281 add_val($doc, PublicInbox::Search::DT(), $dt);
283 my $tg = $self->term_generator;
285 $tg->set_document($doc);
286 $self->index_text($subj, 1, 'S') if $subj;
287 $self->index_users($smsg);
289 msg_iter($mime, sub {
290 my ($part, $depth, @idx) = @{$_[0]};
291 my $ct = $part->content_type || 'text/plain';
292 my $fn = $part->filename;
293 if (defined $fn && $fn ne '') {
294 $self->index_text($fn, 1, 'XFN');
297 return if $ct =~ m!\btext/x?html\b!i;
299 my $s = eval { $part->body_str };
301 if ($ct =~ m!\btext/plain\b!i) {
302 # Try to assume UTF-8 because Alpine
303 # seems to do wacky things and set
305 $part->charset_set('UTF-8');
306 $s = eval { $part->body_str };
307 $s = $part->body if $@;
310 defined $s or return;
313 my $body = $part->body;
314 my @lines = split(/\n/, $body);
315 while (defined(my $l = shift @lines)) {
317 $self->index_body(\@orig, $doc) if @orig;
320 $self->index_body(\@quot, 0) if @quot;
324 $self->index_body(\@quot, 0) if @quot;
325 $self->index_body(\@orig, $doc) if @orig;
328 foreach my $mid (@$mids) {
329 $self->index_text($mid, 1, 'XM');
331 # because too many Message-IDs are prefixed with
333 if ($mid =~ /\w{12,}/) {
334 my @long = ($mid =~ /(\w{3,}+)/g);
335 $self->index_text(join(' ', @long), 1, 'XM');
338 $smsg->{to} = $smsg->{cc} = '';
339 PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids);
340 my $data = $smsg->to_doc_data($oid, $mid0);
341 $doc->set_data($data);
342 if (my $altid = $self->{-altid}) {
343 foreach my $alt (@$altid) {
344 my $pfx = $alt->{xprefix};
345 foreach my $mid (@$mids) {
346 my $id = $alt->mid2alt($mid);
347 next unless defined $id;
348 $doc->add_boolean_term($pfx . $id);
352 $doc->add_boolean_term('Q' . $_) foreach @$mids;
353 $self->{xdb}->replace_document($num, $doc);
357 # mime = Email::MIME object
358 my ($self, $mime, $bytes, $num, $oid, $mid0) = @_;
359 my $xapianlevels = qr/\A(?:full|medium)\z/;
360 my $mids = mids($mime->header_obj);
361 $mid0 = $mids->[0] unless defined $mid0; # v1 compatibility
362 unless (defined $num) { # v1
364 $num = index_mm($self, $mime);
367 if ($self->{indexlevel} =~ $xapianlevels) {
368 $self->add_xapian($mime, $num, $oid, $mids, $mid0)
370 if (my $over = $self->{over}) {
371 $over->add_overview($mime, $bytes, $num, $oid, $mid0);
376 warn "failed to index message <".join('> <',@$mids).">: $@\n";
382 # returns begin and end PostingIterator
384 my ($self, $termval) = @_;
385 my $db = $self->{xdb};
387 ($db->postlist_begin($termval), $db->postlist_end($termval));
391 my ($self, $termval, $cb) = @_;
392 my $batch_size = 1000; # don't let @ids grow too large to avoid OOM
394 my ($head, $tail) = $self->find_doc_ids($termval);
395 return if $head == $tail;
397 for (; $head != $tail && @ids < $batch_size; $head->inc) {
398 push @ids, $head->get_docid;
405 my ($self, $mid) = @_;
406 my $db = $self->{xdb};
408 $mid = mid_clean($mid);
409 my $over = $self->{over};
412 batch_do($self, 'Q' . $mid, sub {
414 $db->delete_document($_) for @$ids;
415 $over->delete_articles($ids) if $over;
420 warn "failed to remove message <$mid>: $@\n";
422 warn "cannot remove non-existent <$mid>\n";
426 # MID is a hint in V2
428 my ($self, $oid, $mid) = @_;
429 my $db = $self->{xdb};
431 $self->{over}->remove_oid($oid, $mid) if $self->{over};
433 # XXX careful, we cannot use batch_do here since we conditionally
434 # delete documents based on other factors, so we cannot call
435 # find_doc_ids twice.
436 my ($head, $tail) = $self->find_doc_ids('Q' . $mid);
437 return if $head == $tail;
439 # there is only ONE element in @delete unless we
440 # have bugs in our v2writable deduplication check
442 for (; $head != $tail; $head->inc) {
443 my $docid = $head->get_docid;
444 my $doc = $db->get_document($docid);
445 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
447 if ($smsg->{blob} eq $oid) {
448 push(@delete, $docid);
451 $db->delete_document($_) foreach @delete;
455 sub term_generator { # write-only
458 my $tg = $self->{term_generator};
461 $tg = Search::Xapian::TermGenerator->new;
462 $tg->set_stemmer($self->stemmer);
464 $self->{term_generator} = $tg;
467 sub index_git_blob_id {
468 my ($doc, $pfx, $objid) = @_;
470 my $len = length($objid);
471 for (my $len = length($objid); $len >= 7; ) {
472 $doc->add_term($pfx.$objid);
473 $objid = substr($objid, 0, --$len);
478 my ($self, $mime) = @_;
479 my $mid = eval { mid_clean(mid_mime($mime)) };
480 $self->remove_message($mid) if defined $mid;
484 my ($self, $mime) = @_;
485 my $mid = mid_clean(mid_mime($mime));
486 my $mm = $self->{mm};
489 if (defined $self->{regen_down}) {
490 $num = $mm->num_for($mid) and return $num;
492 while (($num = $self->{regen_down}--) > 0) {
493 if ($mm->mid_set($num, $mid) != 0) {
497 } elsif (defined $self->{regen_up}) {
498 $num = $mm->num_for($mid) and return $num;
500 # this is to fixup old bugs due to add-remove-add
501 while (($num = ++$self->{regen_up})) {
502 if ($mm->mid_set($num, $mid) != 0) {
508 $num = $mm->mid_insert($mid) and return $num;
510 # fallback to num_for since filters like RubyLang set the number
515 my ($self, $mime) = @_;
516 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
520 my ($self, $mime, $bytes, $blob) = @_;
521 my $num = index_mm($self, $mime);
522 add_message($self, $mime, $bytes, $num, $blob);
526 my ($self, $mime) = @_;
527 unindex_blob($self, $mime);
528 unindex_mm($self, $mime);
532 my ($git, $blob, $sizeref) = @_;
534 my $str = $git->cat_file($blob, $sizeref);
535 # fixup bugs from import:
536 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
537 PublicInbox::MIME->new($str);
543 my ($self, $opts) = @_;
544 $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) })
547 sub batch_adjust ($$$$) {
548 my ($max, $bytes, $batch_cb, $latest) = @_;
552 $batch_cb->($latest);
558 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
559 my $hex = '[a-f0-9]';
560 my $h40 = $hex .'{40}';
561 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
562 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
563 my $git = $self->{git};
566 my $max = BATCH_BYTES;
571 while (defined($line = <$log>)) {
572 if ($line =~ /$addmsg/o) {
574 if (delete $D{$blob}) {
575 if (defined $self->{regen_down}) {
576 $self->{regen_down}--;
580 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
581 batch_adjust(\$max, $bytes, $batch_cb, $latest);
582 $add_cb->($self, $mime, $bytes, $blob);
583 } elsif ($line =~ /$delmsg/o) {
586 } elsif ($line =~ /^commit ($h40)/o) {
592 foreach my $blob (keys %D) {
593 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
594 $del_cb->($self, $mime);
596 $batch_cb->($latest, $newest);
601 die "BUG: _msgmap_init is only for v1\n" if $self->{version} != 1;
602 $self->{mm} ||= eval {
603 require PublicInbox::Msgmap;
604 PublicInbox::Msgmap->new($self->{mainrepo}, 1);
609 my ($self, $range) = @_;
610 my $git = $self->{git};
612 if (index($range, '..') < 0) {
614 # can't use 'rev-list --count' if we use --diff-filter
615 my $fh = $git->popen(qw(log --pretty=tformat:%h
616 --no-notes --no-color --no-renames
617 --diff-filter=AM), $range);
618 ++$regen_max while <$fh>;
619 my (undef, $max) = $self->{mm}->minmax;
621 if ($max && $max == $regen_max) {
622 # fix up old bugs in full indexes which caused messages to
623 # not appear in Msgmap
624 $self->{regen_up} = $max;
626 # normal regen is for for fresh data
627 $self->{regen_down} = $regen_max;
631 $git->popen(qw/log --no-notes --no-color --no-renames
632 --raw -r --no-abbrev/, $range);
635 sub is_ancestor ($$$) {
636 my ($git, $cur, $tip) = @_;
637 return 0 unless $git->check($cur);
638 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
639 qw(merge-base --is-ancestor), $cur, $tip ];
640 my $pid = spawn($cmd);
641 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
642 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
646 sub need_update ($$$) {
647 my ($self, $cur, $new) = @_;
648 my $git = $self->{git};
649 return 1 if $cur && !is_ancestor($git, $cur, $new);
650 my $range = $cur eq '' ? $new : "$cur..$new";
651 chomp(my $n = $git->qx(qw(rev-list --count), $range));
652 ($n eq '' || $n > 0);
655 # indexes all unindexed messages (v1 only)
657 my ($self, $opts) = @_;
658 my $tip = $opts->{ref} || 'HEAD';
659 my $reindex = $opts->{reindex};
660 my ($mkey, $last_commit, $lx, $xlog);
661 my $git = $self->{git};
664 my $xdb = $self->begin_txn_lazy;
665 my $mm = _msgmap_init($self);
668 $mkey = 'last_commit';
669 $last_commit = $xdb->get_metadata('last_commit');
673 $mkey = undef if $last_commit ne '';
676 # use last_commit from msgmap if it is older or unset
677 my $lm = $mm->last_commit || '';
678 if (!$lm || ($lm && $lx && is_ancestor($git, $lm, $lx))) {
682 $self->{over}->rollback_lazy;
683 $self->{over}->disconnect;
686 $xdb->cancel_transaction;
687 $xdb = _xdb_release($self);
689 # ensure we leak no FDs to "git log" with Xapian <= 1.2
690 my $range = $lx eq '' ? $tip : "$lx..$tip";
691 $xlog = _git_log($self, $range);
693 $xdb = $self->begin_txn_lazy;
694 } while ($xdb->get_metadata('last_commit') ne $last_commit);
696 my $dbh = $mm->{dbh} if $mm;
698 my ($commit, $newest) = @_;
701 my $cur = $mm->last_commit || '';
702 if (need_update($self, $cur, $newest)) {
703 $mm->last_commit($newest);
708 if ($mkey && $newest) {
709 my $cur = $xdb->get_metadata($mkey);
710 if (need_update($self, $cur, $newest)) {
711 $xdb->set_metadata($mkey, $newest);
714 $self->commit_txn_lazy;
716 $xdb = _xdb_release($self);
717 # let another process do some work... <
719 $xdb = $self->begin_txn_lazy;
720 $dbh->begin_work if $dbh;
725 read_log($self, $xlog, *index_both, *unindex_both, $cb);
729 # order matters for unlocking
730 $_[0]->{xdb} = undef;
731 $_[0]->{lockfh} = undef;
734 # remote_* subs are only used by SearchIdxPart
737 if (my $w = $self->{w}) {
738 print $w "commit\n" or die "failed to write commit: $!";
740 $self->commit_txn_lazy;
746 if (my $w = delete $self->{w}) {
747 my $pid = delete $self->{pid} or die "no process to wait on\n";
748 print $w "close\n" or die "failed to write to pid:$pid: $!\n";
749 close $w or die "failed to close pipe for pid:$pid: $!\n";
750 waitpid($pid, 0) == $pid or die "remote process did not finish";
751 $? == 0 or die ref($self)." pid:$pid exited with: $?";
753 die "transaction in progress $self\n" if $self->{txn};
754 $self->_xdb_release if $self->{xdb};
759 my ($self, $oid, $mid) = @_;
760 if (my $w = $self->{w}) {
761 # triggers remove_by_oid in a partition
762 print $w "D $oid $mid\n" or die "failed to write remove $!";
764 $self->begin_txn_lazy;
765 $self->remove_by_oid($oid, $mid);
771 return if $self->{txn};
773 $self->{-inbox}->with_umask(sub {
774 my $xdb = $self->{xdb} || $self->_xdb_acquire;
775 $self->{over}->begin_lazy if $self->{over};
776 $xdb->begin_transaction;
782 sub commit_txn_lazy {
784 delete $self->{txn} or return;
785 $self->{-inbox}->with_umask(sub {
786 $self->{xdb}->commit_transaction;
787 $self->{over}->commit_lazy if $self->{over};
793 die "$$ $0 xdb not released\n" if $self->{xdb};
794 die "$$ $0 still in transaction\n" if $self->{txn};