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 => defined($ENV{XAPIAN_FLUSH_THRESHOLD}) ?
26 0x7fffffff : 1_000_000,
27 DEBUG => !!$ENV{DEBUG},
30 my $xapianlevels = qr/\A(?:full|medium)\z/;
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 $levels = qr/\A(?:full|medium|basic)\z/;
54 my $mainrepo = $ibx; # for "public-inbox-index" w/o entry in config
55 my $git_dir = $mainrepo;
58 my $indexlevel = 'full';
60 $mainrepo = $ibx->{mainrepo};
61 $altid = $ibx->{altid};
62 $version = $ibx->{version} || 1;
64 require PublicInbox::AltId;
66 PublicInbox::AltId->new($ibx, $_);
69 if ($ibx->{indexlevel}) {
70 if ($ibx->{indexlevel} =~ $levels) {
71 $indexlevel = $ibx->{indexlevel};
73 die("Invalid indexlevel $ibx->{indexlevel}\n");
77 $ibx = { mainrepo => $git_dir, version => 1 };
79 $ibx = PublicInbox::InboxWritable->new($ibx);
80 require Search::Xapian::WritableDatabase;
82 mainrepo => $mainrepo,
87 indexlevel => $indexlevel,
91 $self->{lock_path} = "$mainrepo/ssoma.lock";
92 my $dir = $self->xdir;
93 $self->{over} = PublicInbox::OverIdx->new("$dir/over.sqlite3");
94 } elsif ($version == 2) {
95 defined $part or die "partition is required for v2\n";
96 # partition is a number
97 $self->{partition} = $part;
98 $self->{lock_path} = undef;
100 die "unsupported inbox version=$version\n";
102 $self->{creat} = ($creat || 0) == 1;
108 my $xdb = delete $self->{xdb} or croak 'not acquired';
110 $self->lock_release if $self->{creat};
116 croak 'already acquired' if $self->{xdb};
117 my $dir = $self->xdir;
118 my $flag = Search::Xapian::DB_OPEN;
119 if ($self->{creat}) {
122 File::Path::mkpath($dir);
123 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
125 $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag);
129 my ($doc, $col, $num) = @_;
130 $num = Search::Xapian::sortable_serialise($num);
131 $doc->add_value($col, $num);
134 sub index_text ($$$$)
136 my ($self, $field, $n, $text) = @_;
137 my $tg = $self->term_generator;
139 if ($self->{indexlevel} eq 'full') {
140 $tg->index_text($field, $n, $text);
141 $tg->increase_termpos;
143 $tg->index_text_without_positions($field, $n, $text);
147 sub index_users ($$) {
148 my ($self, $smsg) = @_;
150 my $from = $smsg->from;
154 $self->index_text($from, 1, 'A'); # A - author
155 $self->index_text($to, 1, 'XTO') if $to ne '';
156 $self->index_text($cc, 1, 'XCC') if $cc ne '';
159 sub index_diff_inc ($$$$) {
160 my ($self, $text, $pfx, $xnq) = @_;
162 $self->index_text(join("\n", @$xnq), 1, 'XNQ');
165 $self->index_text($text, 1, $pfx);
168 sub index_old_diff_fn {
169 my ($self, $seen, $fa, $fb, $xnq) = @_;
171 # no renames or space support for traditional diffs,
172 # find the number of leading common paths to strip:
173 my @fa = split('/', $fa);
174 my @fb = split('/', $fb);
175 while (scalar(@fa) && scalar(@fb)) {
176 $fa = join('/', @fa);
177 $fb = join('/', @fb);
179 unless ($seen->{$fa}++) {
180 $self->index_diff_inc($fa, 'XDFN', $xnq);
190 sub index_diff ($$$) {
191 my ($self, $lines, $doc) = @_;
197 if ($in_diff && s/^ //) { # diff context
198 $self->index_diff_inc($_, 'XDFCTX', $xnq);
199 } elsif (/^-- $/) { # email signature begins
201 } elsif (m!^diff --git ("?a/.+) ("?b/.+)\z!) {
202 my ($fa, $fb) = ($1, $2);
203 my $fn = (split('/', git_unquote($fa), 2))[1];
204 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
205 $fn = (split('/', git_unquote($fb), 2))[1];
206 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
209 } elsif (m/^diff -(.+) (\S+) (\S+)$/) {
210 my ($opt, $fa, $fb) = ($1, $2, $3);
212 # only support unified:
213 next unless $opt =~ /[uU]/;
214 $in_diff = $self->index_old_diff_fn(\%seen, $fa, $fb,
216 } elsif (m!^--- ("?a/.+)!) {
217 my $fn = (split('/', git_unquote($1), 2))[1];
218 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
220 } elsif (m!^\+\+\+ ("?b/.+)!) {
221 my $fn = (split('/', git_unquote($1), 2))[1];
222 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
224 } elsif (/^--- (\S+)/) {
227 } elsif (defined $in_diff && /^\+\+\+ (\S+)/) {
228 $in_diff = $self->index_old_diff_fn(\%seen, $in_diff, $1,
230 } elsif ($in_diff && s/^\+//) { # diff added
231 $self->index_diff_inc($_, 'XDFB', $xnq);
232 } elsif ($in_diff && s/^-//) { # diff removed
233 $self->index_diff_inc($_, 'XDFA', $xnq);
234 } elsif (m!^index ([a-f0-9]+)\.\.([a-f0-9]+)!) {
235 my ($ba, $bb) = ($1, $2);
236 index_git_blob_id($doc, 'XDFPRE', $ba);
237 index_git_blob_id($doc, 'XDFPOST', $bb);
239 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*$/) {
240 # traditional diff w/o -p
241 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*(\S+.*)$/) {
242 # hunk header context
243 $self->index_diff_inc($1, 'XDFHH', $xnq);
244 # ignore the following lines:
245 } elsif (/^(?:dis)similarity index/ ||
246 /^(?:old|new) mode/ ||
247 /^(?:deleted|new) file mode/ ||
248 /^(?:copy|rename) (?:from|to) / ||
249 /^(?:dis)?similarity index / ||
250 /^\\ No newline at end of file/ ||
251 /^Binary files .* differ/) {
257 warn "non-diff line: $_\n" if DEBUG && $_ ne '';
262 $self->index_text(join("\n", @xnq), 1, 'XNQ');
265 sub index_body ($$$) {
266 my ($self, $lines, $doc) = @_;
267 my $txt = join("\n", @$lines);
269 # does it look like a diff?
270 if ($txt =~ /^(?:diff|---|\+\+\+) /ms) {
272 $self->index_diff($lines, $doc);
274 $self->index_text($txt, 1, 'XNQ');
277 $self->index_text($txt, 0, 'XQUOT');
282 sub add_xapian ($$$$$) {
283 my ($self, $mime, $num, $oid, $mids, $mid0) = @_;
284 my $smsg = PublicInbox::SearchMsg->new($mime);
285 my $doc = $smsg->{doc};
286 my $subj = $smsg->subject;
287 add_val($doc, PublicInbox::Search::TS(), $smsg->ts);
288 my @ds = gmtime($smsg->ds);
289 my $yyyymmdd = strftime('%Y%m%d', @ds);
290 add_val($doc, PublicInbox::Search::YYYYMMDD(), $yyyymmdd);
291 my $dt = strftime('%Y%m%d%H%M%S', @ds);
292 add_val($doc, PublicInbox::Search::DT(), $dt);
294 my $tg = $self->term_generator;
296 $tg->set_document($doc);
297 $self->index_text($subj, 1, 'S') if $subj;
298 $self->index_users($smsg);
300 msg_iter($mime, sub {
301 my ($part, $depth, @idx) = @{$_[0]};
302 my $ct = $part->content_type || 'text/plain';
303 my $fn = $part->filename;
304 if (defined $fn && $fn ne '') {
305 $self->index_text($fn, 1, 'XFN');
308 return if $ct =~ m!\btext/x?html\b!i;
310 my $s = eval { $part->body_str };
312 if ($ct =~ m!\btext/plain\b!i) {
313 # Try to assume UTF-8 because Alpine
314 # seems to do wacky things and set
316 $part->charset_set('UTF-8');
317 $s = eval { $part->body_str };
318 $s = $part->body if $@;
321 defined $s or return;
324 my $body = $part->body;
325 my @lines = split(/\n/, $body);
326 while (defined(my $l = shift @lines)) {
328 $self->index_body(\@orig, $doc) if @orig;
331 $self->index_body(\@quot, 0) if @quot;
335 $self->index_body(\@quot, 0) if @quot;
336 $self->index_body(\@orig, $doc) if @orig;
339 foreach my $mid (@$mids) {
340 $self->index_text($mid, 1, 'XM');
342 # because too many Message-IDs are prefixed with
344 if ($mid =~ /\w{12,}/) {
345 my @long = ($mid =~ /(\w{3,}+)/g);
346 $self->index_text(join(' ', @long), 1, 'XM');
349 $smsg->{to} = $smsg->{cc} = '';
350 PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids);
351 my $data = $smsg->to_doc_data($oid, $mid0);
352 $doc->set_data($data);
353 if (my $altid = $self->{-altid}) {
354 foreach my $alt (@$altid) {
355 my $pfx = $alt->{xprefix};
356 foreach my $mid (@$mids) {
357 my $id = $alt->mid2alt($mid);
358 next unless defined $id;
359 $doc->add_boolean_term($pfx . $id);
363 $doc->add_boolean_term('Q' . $_) foreach @$mids;
364 $self->{xdb}->replace_document($num, $doc);
368 # mime = Email::MIME object
369 my ($self, $mime, $bytes, $num, $oid, $mid0) = @_;
370 my $mids = mids($mime->header_obj);
371 $mid0 = $mids->[0] unless defined $mid0; # v1 compatibility
372 unless (defined $num) { # v1
374 $num = index_mm($self, $mime);
377 if ($self->{indexlevel} =~ $xapianlevels) {
378 $self->add_xapian($mime, $num, $oid, $mids, $mid0)
380 if (my $over = $self->{over}) {
381 $over->add_overview($mime, $bytes, $num, $oid, $mid0);
386 warn "failed to index message <".join('> <',@$mids).">: $@\n";
392 # returns begin and end PostingIterator
394 my ($self, $termval) = @_;
395 my $db = $self->{xdb};
397 ($db->postlist_begin($termval), $db->postlist_end($termval));
401 my ($self, $termval, $cb) = @_;
402 my $batch_size = 1000; # don't let @ids grow too large to avoid OOM
404 my ($head, $tail) = $self->find_doc_ids($termval);
405 return if $head == $tail;
407 for (; $head != $tail && @ids < $batch_size; $head->inc) {
408 push @ids, $head->get_docid;
415 my ($self, $mid) = @_;
416 my $db = $self->{xdb};
418 $mid = mid_clean($mid);
419 my $over = $self->{over};
422 batch_do($self, 'Q' . $mid, sub {
424 $db->delete_document($_) for @$ids;
425 $over->delete_articles($ids) if $over;
430 warn "failed to remove message <$mid>: $@\n";
432 warn "cannot remove non-existent <$mid>\n";
436 # MID is a hint in V2
438 my ($self, $oid, $mid) = @_;
439 my $db = $self->{xdb};
441 $self->{over}->remove_oid($oid, $mid) if $self->{over};
443 # XXX careful, we cannot use batch_do here since we conditionally
444 # delete documents based on other factors, so we cannot call
445 # find_doc_ids twice.
446 my ($head, $tail) = $self->find_doc_ids('Q' . $mid);
447 return if $head == $tail;
449 # there is only ONE element in @delete unless we
450 # have bugs in our v2writable deduplication check
452 for (; $head != $tail; $head->inc) {
453 my $docid = $head->get_docid;
454 my $doc = $db->get_document($docid);
455 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
457 if ($smsg->{blob} eq $oid) {
458 push(@delete, $docid);
461 $db->delete_document($_) foreach @delete;
465 sub term_generator { # write-only
468 my $tg = $self->{term_generator};
471 $tg = Search::Xapian::TermGenerator->new;
472 $tg->set_stemmer($self->stemmer);
474 $self->{term_generator} = $tg;
477 sub index_git_blob_id {
478 my ($doc, $pfx, $objid) = @_;
480 my $len = length($objid);
481 for (my $len = length($objid); $len >= 7; ) {
482 $doc->add_term($pfx.$objid);
483 $objid = substr($objid, 0, --$len);
488 my ($self, $mime) = @_;
489 my $mid = eval { mid_clean(mid_mime($mime)) };
490 $self->remove_message($mid) if defined $mid;
494 my ($self, $mime) = @_;
495 my $mid = mid_clean(mid_mime($mime));
496 my $mm = $self->{mm};
499 if (defined $self->{regen_down}) {
500 $num = $mm->num_for($mid) and return $num;
502 while (($num = $self->{regen_down}--) > 0) {
503 if ($mm->mid_set($num, $mid) != 0) {
507 } elsif (defined $self->{regen_up}) {
508 $num = $mm->num_for($mid) and return $num;
510 # this is to fixup old bugs due to add-remove-add
511 while (($num = ++$self->{regen_up})) {
512 if ($mm->mid_set($num, $mid) != 0) {
518 $num = $mm->mid_insert($mid) and return $num;
520 # fallback to num_for since filters like RubyLang set the number
525 my ($self, $mime) = @_;
526 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
530 my ($self, $mime, $bytes, $blob) = @_;
531 my $num = index_mm($self, $mime);
532 add_message($self, $mime, $bytes, $num, $blob);
536 my ($self, $mime) = @_;
537 unindex_blob($self, $mime);
538 unindex_mm($self, $mime);
542 my ($git, $blob, $sizeref) = @_;
544 my $str = $git->cat_file($blob, $sizeref);
545 # fixup bugs from import:
546 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
547 PublicInbox::MIME->new($str);
553 my ($self, $opts) = @_;
554 $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) })
557 sub batch_adjust ($$$$) {
558 my ($max, $bytes, $batch_cb, $latest) = @_;
562 $batch_cb->($latest);
568 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
569 my $hex = '[a-f0-9]';
570 my $h40 = $hex .'{40}';
571 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
572 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
573 my $git = $self->{git};
576 my $max = BATCH_BYTES;
581 while (defined($line = <$log>)) {
582 if ($line =~ /$addmsg/o) {
584 if (delete $D{$blob}) {
585 if (defined $self->{regen_down}) {
586 $self->{regen_down}--;
590 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
591 batch_adjust(\$max, $bytes, $batch_cb, $latest);
592 $add_cb->($self, $mime, $bytes, $blob);
593 } elsif ($line =~ /$delmsg/o) {
596 } elsif ($line =~ /^commit ($h40)/o) {
602 foreach my $blob (keys %D) {
603 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
604 $del_cb->($self, $mime);
606 $batch_cb->($latest, $newest);
611 die "BUG: _msgmap_init is only for v1\n" if $self->{version} != 1;
612 $self->{mm} ||= eval {
613 require PublicInbox::Msgmap;
614 PublicInbox::Msgmap->new($self->{mainrepo}, 1);
619 my ($self, $range) = @_;
620 my $git = $self->{git};
622 if (index($range, '..') < 0) {
624 # can't use 'rev-list --count' if we use --diff-filter
625 my $fh = $git->popen(qw(log --pretty=tformat:%h
626 --no-notes --no-color --no-renames
627 --diff-filter=AM), $range);
628 ++$regen_max while <$fh>;
629 my (undef, $max) = $self->{mm}->minmax;
631 if ($max && $max == $regen_max) {
632 # fix up old bugs in full indexes which caused messages to
633 # not appear in Msgmap
634 $self->{regen_up} = $max;
636 # normal regen is for for fresh data
637 $self->{regen_down} = $regen_max;
641 $git->popen(qw/log --no-notes --no-color --no-renames
642 --raw -r --no-abbrev/, $range);
645 sub is_ancestor ($$$) {
646 my ($git, $cur, $tip) = @_;
647 return 0 unless $git->check($cur);
648 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
649 qw(merge-base --is-ancestor), $cur, $tip ];
650 my $pid = spawn($cmd);
651 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
652 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
656 sub need_update ($$$) {
657 my ($self, $cur, $new) = @_;
658 my $git = $self->{git};
659 return 1 if $cur && !is_ancestor($git, $cur, $new);
660 my $range = $cur eq '' ? $new : "$cur..$new";
661 chomp(my $n = $git->qx(qw(rev-list --count), $range));
662 ($n eq '' || $n > 0);
665 # indexes all unindexed messages (v1 only)
667 my ($self, $opts) = @_;
668 my $tip = $opts->{ref} || 'HEAD';
669 my $reindex = $opts->{reindex};
670 my ($mkey, $last_commit, $lx, $xlog);
671 my $git = $self->{git};
674 my $xdb = $self->begin_txn_lazy;
675 my $mm = _msgmap_init($self);
678 $mkey = 'last_commit';
679 $last_commit = $xdb->get_metadata('last_commit');
683 $mkey = undef if $last_commit ne '';
686 # use last_commit from msgmap if it is older or unset
687 my $lm = $mm->last_commit || '';
688 if (!$lm || ($lm && $lx && is_ancestor($git, $lm, $lx))) {
692 $self->{over}->rollback_lazy;
693 $self->{over}->disconnect;
696 $xdb->cancel_transaction;
697 $xdb = _xdb_release($self);
699 # ensure we leak no FDs to "git log" with Xapian <= 1.2
700 my $range = $lx eq '' ? $tip : "$lx..$tip";
701 $xlog = _git_log($self, $range);
703 $xdb = $self->begin_txn_lazy;
704 } while ($xdb->get_metadata('last_commit') ne $last_commit);
706 my $dbh = $mm->{dbh} if $mm;
708 my ($commit, $newest) = @_;
711 my $cur = $mm->last_commit || '';
712 if (need_update($self, $cur, $newest)) {
713 $mm->last_commit($newest);
718 if ($mkey && $newest && $self->{indexlevel} =~ $xapianlevels) {
719 my $cur = $xdb->get_metadata($mkey);
720 if (need_update($self, $cur, $newest)) {
721 $xdb->set_metadata($mkey, $newest);
724 $self->commit_txn_lazy;
726 $xdb = _xdb_release($self);
727 # let another process do some work... <
729 $xdb = $self->begin_txn_lazy;
730 $dbh->begin_work if $dbh;
735 read_log($self, $xlog, *index_both, *unindex_both, $cb);
739 # order matters for unlocking
740 $_[0]->{xdb} = undef;
741 $_[0]->{lockfh} = undef;
744 # remote_* subs are only used by SearchIdxPart
747 if (my $w = $self->{w}) {
748 print $w "commit\n" or die "failed to write commit: $!";
750 $self->commit_txn_lazy;
756 if (my $w = delete $self->{w}) {
757 my $pid = delete $self->{pid} or die "no process to wait on\n";
758 print $w "close\n" or die "failed to write to pid:$pid: $!\n";
759 close $w or die "failed to close pipe for pid:$pid: $!\n";
760 waitpid($pid, 0) == $pid or die "remote process did not finish";
761 $? == 0 or die ref($self)." pid:$pid exited with: $?";
763 die "transaction in progress $self\n" if $self->{txn};
764 $self->_xdb_release if $self->{xdb};
769 my ($self, $oid, $mid) = @_;
770 if (my $w = $self->{w}) {
771 # triggers remove_by_oid in a partition
772 print $w "D $oid $mid\n" or die "failed to write remove $!";
774 $self->begin_txn_lazy;
775 $self->remove_by_oid($oid, $mid);
781 return if $self->{txn};
783 $self->{-inbox}->with_umask(sub {
784 my $xdb = $self->{xdb} || $self->_xdb_acquire;
785 $self->{over}->begin_lazy if $self->{over};
786 $xdb->begin_transaction;
792 sub commit_txn_lazy {
794 delete $self->{txn} or return;
795 $self->{-inbox}->with_umask(sub {
796 $self->{xdb}->commit_transaction;
797 $self->{over}->commit_lazy if $self->{over};
803 die "$$ $0 xdb not released\n" if $self->{xdb};
804 die "$$ $0 still in transaction\n" if $self->{txn};