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},
42 return $s unless ($s =~ /\A"(.*)"\z/);
44 $s =~ s/\\([abfnrtv])/$GIT_ESC{$1}/g;
45 $s =~ s/\\([0-7]{1,3})/chr(oct($1))/ge;
50 my ($class, $ibx, $creat, $part) = @_;
51 my $levels = qr/\A(?:full|medium|basic)\z/;
52 my $mainrepo = $ibx; # for "public-inbox-index" w/o entry in config
53 my $git_dir = $mainrepo;
56 my $indexlevel = 'full';
58 $mainrepo = $ibx->{mainrepo};
59 $altid = $ibx->{altid};
60 $version = $ibx->{version} || 1;
62 require PublicInbox::AltId;
64 PublicInbox::AltId->new($ibx, $_);
67 if ($ibx->{indexlevel}) {
68 if ($ibx->{indexlevel} =~ $levels) {
69 $indexlevel = $ibx->{indexlevel};
71 die("Invalid indexlevel $ibx->{indexlevel}\n");
75 $ibx = { mainrepo => $git_dir, version => 1 };
77 $ibx = PublicInbox::InboxWritable->new($ibx);
78 require Search::Xapian::WritableDatabase;
80 mainrepo => $mainrepo,
85 indexlevel => $indexlevel,
89 $self->{lock_path} = "$mainrepo/ssoma.lock";
90 my $dir = $self->xdir;
91 $self->{over} = PublicInbox::OverIdx->new("$dir/over.sqlite3");
92 } elsif ($version == 2) {
93 defined $part or die "partition is required for v2\n";
94 # partition is a number
95 $self->{partition} = $part;
96 $self->{lock_path} = undef;
98 die "unsupported inbox version=$version\n";
100 $self->{creat} = ($creat || 0) == 1;
106 my $xdb = delete $self->{xdb} or croak 'not acquired';
108 $self->lock_release if $self->{creat};
114 croak 'already acquired' if $self->{xdb};
115 my $dir = $self->xdir;
116 my $flag = Search::Xapian::DB_OPEN;
117 if ($self->{creat}) {
120 File::Path::mkpath($dir);
121 $flag = Search::Xapian::DB_CREATE_OR_OPEN;
123 $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag);
127 my ($doc, $col, $num) = @_;
128 $num = Search::Xapian::sortable_serialise($num);
129 $doc->add_value($col, $num);
132 sub index_text ($$$$)
134 my ($self, $field, $n, $text) = @_;
135 my $tg = $self->term_generator;
137 if ($self->{indexlevel} eq 'full') {
138 $tg->index_text($field, $n, $text);
139 $tg->increase_termpos;
141 $tg->index_text_without_positions($field, $n, $text);
145 sub index_users ($$) {
146 my ($self, $smsg) = @_;
148 my $from = $smsg->from;
152 $self->index_text($from, 1, 'A'); # A - author
153 $self->index_text($to, 1, 'XTO') if $to ne '';
154 $self->index_text($cc, 1, 'XCC') if $cc ne '';
157 sub index_diff_inc ($$$$) {
158 my ($self, $text, $pfx, $xnq) = @_;
160 $self->index_text(join("\n", @$xnq), 1, 'XNQ');
163 $self->index_text($text, 1, $pfx);
166 sub index_old_diff_fn {
167 my ($self, $seen, $fa, $fb, $xnq) = @_;
169 # no renames or space support for traditional diffs,
170 # find the number of leading common paths to strip:
171 my @fa = split('/', $fa);
172 my @fb = split('/', $fb);
173 while (scalar(@fa) && scalar(@fb)) {
174 $fa = join('/', @fa);
175 $fb = join('/', @fb);
177 unless ($seen->{$fa}++) {
178 $self->index_diff_inc($fa, 'XDFN', $xnq);
188 sub index_diff ($$$) {
189 my ($self, $lines, $doc) = @_;
195 if ($in_diff && s/^ //) { # diff context
196 $self->index_diff_inc($_, 'XDFCTX', $xnq);
197 } elsif (/^-- $/) { # email signature begins
199 } elsif (m!^diff --git ("?a/.+) ("?b/.+)\z!) {
200 my ($fa, $fb) = ($1, $2);
201 my $fn = (split('/', git_unquote($fa), 2))[1];
202 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
203 $fn = (split('/', git_unquote($fb), 2))[1];
204 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
207 } elsif (m/^diff -(.+) (\S+) (\S+)$/) {
208 my ($opt, $fa, $fb) = ($1, $2, $3);
210 # only support unified:
211 next unless $opt =~ /[uU]/;
212 $in_diff = $self->index_old_diff_fn(\%seen, $fa, $fb,
214 } elsif (m!^--- ("?a/.+)!) {
215 my $fn = (split('/', git_unquote($1), 2))[1];
216 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
218 } elsif (m!^\+\+\+ ("?b/.+)!) {
219 my $fn = (split('/', git_unquote($1), 2))[1];
220 $seen{$fn}++ or $self->index_diff_inc($fn, 'XDFN', $xnq);
222 } elsif (/^--- (\S+)/) {
225 } elsif (defined $in_diff && /^\+\+\+ (\S+)/) {
226 $in_diff = $self->index_old_diff_fn(\%seen, $in_diff, $1,
228 } elsif ($in_diff && s/^\+//) { # diff added
229 $self->index_diff_inc($_, 'XDFB', $xnq);
230 } elsif ($in_diff && s/^-//) { # diff removed
231 $self->index_diff_inc($_, 'XDFA', $xnq);
232 } elsif (m!^index ([a-f0-9]+)\.\.([a-f0-9]+)!) {
233 my ($ba, $bb) = ($1, $2);
234 index_git_blob_id($doc, 'XDFPRE', $ba);
235 index_git_blob_id($doc, 'XDFPOST', $bb);
237 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*$/) {
238 # traditional diff w/o -p
239 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*(\S+.*)$/) {
240 # hunk header context
241 $self->index_diff_inc($1, 'XDFHH', $xnq);
242 # ignore the following lines:
243 } elsif (/^(?:dis)similarity index/ ||
244 /^(?:old|new) mode/ ||
245 /^(?:deleted|new) file mode/ ||
246 /^(?:copy|rename) (?:from|to) / ||
247 /^(?:dis)?similarity index / ||
248 /^\\ No newline at end of file/ ||
249 /^Binary files .* differ/) {
255 warn "non-diff line: $_\n" if DEBUG && $_ ne '';
260 $self->index_text(join("\n", @xnq), 1, 'XNQ');
263 sub index_body ($$$) {
264 my ($self, $lines, $doc) = @_;
265 my $txt = join("\n", @$lines);
267 # does it look like a diff?
268 if ($txt =~ /^(?:diff|---|\+\+\+) /ms) {
270 $self->index_diff($lines, $doc);
272 $self->index_text($txt, 1, 'XNQ');
275 $self->index_text($txt, 0, 'XQUOT');
280 sub add_xapian ($$$$$) {
281 my ($self, $mime, $num, $oid, $mids, $mid0) = @_;
282 my $smsg = PublicInbox::SearchMsg->new($mime);
283 my $doc = $smsg->{doc};
284 my $subj = $smsg->subject;
285 add_val($doc, PublicInbox::Search::TS(), $smsg->ts);
286 my @ds = gmtime($smsg->ds);
287 my $yyyymmdd = strftime('%Y%m%d', @ds);
288 add_val($doc, PublicInbox::Search::YYYYMMDD(), $yyyymmdd);
289 my $dt = strftime('%Y%m%d%H%M%S', @ds);
290 add_val($doc, PublicInbox::Search::DT(), $dt);
292 my $tg = $self->term_generator;
294 $tg->set_document($doc);
295 $self->index_text($subj, 1, 'S') if $subj;
296 $self->index_users($smsg);
298 msg_iter($mime, sub {
299 my ($part, $depth, @idx) = @{$_[0]};
300 my $ct = $part->content_type || 'text/plain';
301 my $fn = $part->filename;
302 if (defined $fn && $fn ne '') {
303 $self->index_text($fn, 1, 'XFN');
306 return if $ct =~ m!\btext/x?html\b!i;
308 my $s = eval { $part->body_str };
310 if ($ct =~ m!\btext/plain\b!i) {
311 # Try to assume UTF-8 because Alpine
312 # seems to do wacky things and set
314 $part->charset_set('UTF-8');
315 $s = eval { $part->body_str };
316 $s = $part->body if $@;
319 defined $s or return;
322 my $body = $part->body;
323 my @lines = split(/\n/, $body);
324 while (defined(my $l = shift @lines)) {
326 $self->index_body(\@orig, $doc) if @orig;
329 $self->index_body(\@quot, 0) if @quot;
333 $self->index_body(\@quot, 0) if @quot;
334 $self->index_body(\@orig, $doc) if @orig;
337 foreach my $mid (@$mids) {
338 $self->index_text($mid, 1, 'XM');
340 # because too many Message-IDs are prefixed with
342 if ($mid =~ /\w{12,}/) {
343 my @long = ($mid =~ /(\w{3,}+)/g);
344 $self->index_text(join(' ', @long), 1, 'XM');
347 $smsg->{to} = $smsg->{cc} = '';
348 PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids);
349 my $data = $smsg->to_doc_data($oid, $mid0);
350 $doc->set_data($data);
351 if (my $altid = $self->{-altid}) {
352 foreach my $alt (@$altid) {
353 my $pfx = $alt->{xprefix};
354 foreach my $mid (@$mids) {
355 my $id = $alt->mid2alt($mid);
356 next unless defined $id;
357 $doc->add_boolean_term($pfx . $id);
361 $doc->add_boolean_term('Q' . $_) foreach @$mids;
362 $self->{xdb}->replace_document($num, $doc);
366 # mime = Email::MIME object
367 my ($self, $mime, $bytes, $num, $oid, $mid0) = @_;
368 my $xapianlevels = qr/\A(?:full|medium)\z/;
369 my $mids = mids($mime->header_obj);
370 $mid0 = $mids->[0] unless defined $mid0; # v1 compatibility
371 unless (defined $num) { # v1
373 $num = index_mm($self, $mime);
376 if ($self->{indexlevel} =~ $xapianlevels) {
377 $self->add_xapian($mime, $num, $oid, $mids, $mid0)
379 if (my $over = $self->{over}) {
380 $over->add_overview($mime, $bytes, $num, $oid, $mid0);
385 warn "failed to index message <".join('> <',@$mids).">: $@\n";
391 # returns begin and end PostingIterator
393 my ($self, $termval) = @_;
394 my $db = $self->{xdb};
396 ($db->postlist_begin($termval), $db->postlist_end($termval));
400 my ($self, $termval, $cb) = @_;
401 my $batch_size = 1000; # don't let @ids grow too large to avoid OOM
403 my ($head, $tail) = $self->find_doc_ids($termval);
404 return if $head == $tail;
406 for (; $head != $tail && @ids < $batch_size; $head->inc) {
407 push @ids, $head->get_docid;
414 my ($self, $mid) = @_;
415 my $db = $self->{xdb};
417 $mid = mid_clean($mid);
418 my $over = $self->{over};
421 batch_do($self, 'Q' . $mid, sub {
423 $db->delete_document($_) for @$ids;
424 $over->delete_articles($ids) if $over;
429 warn "failed to remove message <$mid>: $@\n";
431 warn "cannot remove non-existent <$mid>\n";
435 # MID is a hint in V2
437 my ($self, $oid, $mid) = @_;
438 my $db = $self->{xdb};
440 $self->{over}->remove_oid($oid, $mid) if $self->{over};
442 # XXX careful, we cannot use batch_do here since we conditionally
443 # delete documents based on other factors, so we cannot call
444 # find_doc_ids twice.
445 my ($head, $tail) = $self->find_doc_ids('Q' . $mid);
446 return if $head == $tail;
448 # there is only ONE element in @delete unless we
449 # have bugs in our v2writable deduplication check
451 for (; $head != $tail; $head->inc) {
452 my $docid = $head->get_docid;
453 my $doc = $db->get_document($docid);
454 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
456 if ($smsg->{blob} eq $oid) {
457 push(@delete, $docid);
460 $db->delete_document($_) foreach @delete;
464 sub term_generator { # write-only
467 my $tg = $self->{term_generator};
470 $tg = Search::Xapian::TermGenerator->new;
471 $tg->set_stemmer($self->stemmer);
473 $self->{term_generator} = $tg;
476 sub index_git_blob_id {
477 my ($doc, $pfx, $objid) = @_;
479 my $len = length($objid);
480 for (my $len = length($objid); $len >= 7; ) {
481 $doc->add_term($pfx.$objid);
482 $objid = substr($objid, 0, --$len);
487 my ($self, $mime) = @_;
488 my $mid = eval { mid_clean(mid_mime($mime)) };
489 $self->remove_message($mid) if defined $mid;
493 my ($self, $mime) = @_;
494 my $mid = mid_clean(mid_mime($mime));
495 my $mm = $self->{mm};
498 if (defined $self->{regen_down}) {
499 $num = $mm->num_for($mid) and return $num;
501 while (($num = $self->{regen_down}--) > 0) {
502 if ($mm->mid_set($num, $mid) != 0) {
506 } elsif (defined $self->{regen_up}) {
507 $num = $mm->num_for($mid) and return $num;
509 # this is to fixup old bugs due to add-remove-add
510 while (($num = ++$self->{regen_up})) {
511 if ($mm->mid_set($num, $mid) != 0) {
517 $num = $mm->mid_insert($mid) and return $num;
519 # fallback to num_for since filters like RubyLang set the number
524 my ($self, $mime) = @_;
525 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
529 my ($self, $mime, $bytes, $blob) = @_;
530 my $num = index_mm($self, $mime);
531 add_message($self, $mime, $bytes, $num, $blob);
535 my ($self, $mime) = @_;
536 unindex_blob($self, $mime);
537 unindex_mm($self, $mime);
541 my ($git, $blob, $sizeref) = @_;
543 my $str = $git->cat_file($blob, $sizeref);
544 # fixup bugs from import:
545 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
546 PublicInbox::MIME->new($str);
552 my ($self, $opts) = @_;
553 $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) })
556 sub batch_adjust ($$$$) {
557 my ($max, $bytes, $batch_cb, $latest) = @_;
561 $batch_cb->($latest);
567 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
568 my $hex = '[a-f0-9]';
569 my $h40 = $hex .'{40}';
570 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
571 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
572 my $git = $self->{git};
575 my $max = BATCH_BYTES;
580 while (defined($line = <$log>)) {
581 if ($line =~ /$addmsg/o) {
583 if (delete $D{$blob}) {
584 if (defined $self->{regen_down}) {
585 $self->{regen_down}--;
589 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
590 batch_adjust(\$max, $bytes, $batch_cb, $latest);
591 $add_cb->($self, $mime, $bytes, $blob);
592 } elsif ($line =~ /$delmsg/o) {
595 } elsif ($line =~ /^commit ($h40)/o) {
601 foreach my $blob (keys %D) {
602 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
603 $del_cb->($self, $mime);
605 $batch_cb->($latest, $newest);
610 die "BUG: _msgmap_init is only for v1\n" if $self->{version} != 1;
611 $self->{mm} ||= eval {
612 require PublicInbox::Msgmap;
613 PublicInbox::Msgmap->new($self->{mainrepo}, 1);
618 my ($self, $range) = @_;
619 my $git = $self->{git};
621 if (index($range, '..') < 0) {
623 # can't use 'rev-list --count' if we use --diff-filter
624 my $fh = $git->popen(qw(log --pretty=tformat:%h
625 --no-notes --no-color --no-renames
626 --diff-filter=AM), $range);
627 ++$regen_max while <$fh>;
628 my (undef, $max) = $self->{mm}->minmax;
630 if ($max && $max == $regen_max) {
631 # fix up old bugs in full indexes which caused messages to
632 # not appear in Msgmap
633 $self->{regen_up} = $max;
635 # normal regen is for for fresh data
636 $self->{regen_down} = $regen_max;
640 $git->popen(qw/log --no-notes --no-color --no-renames
641 --raw -r --no-abbrev/, $range);
644 sub is_ancestor ($$$) {
645 my ($git, $cur, $tip) = @_;
646 return 0 unless $git->check($cur);
647 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
648 qw(merge-base --is-ancestor), $cur, $tip ];
649 my $pid = spawn($cmd);
650 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
651 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
655 sub need_update ($$$) {
656 my ($self, $cur, $new) = @_;
657 my $git = $self->{git};
658 return 1 if $cur && !is_ancestor($git, $cur, $new);
659 my $range = $cur eq '' ? $new : "$cur..$new";
660 chomp(my $n = $git->qx(qw(rev-list --count), $range));
661 ($n eq '' || $n > 0);
664 # indexes all unindexed messages (v1 only)
666 my ($self, $opts) = @_;
667 my $tip = $opts->{ref} || 'HEAD';
668 my $reindex = $opts->{reindex};
669 my ($mkey, $last_commit, $lx, $xlog);
670 my $git = $self->{git};
673 my $xdb = $self->begin_txn_lazy;
674 my $mm = _msgmap_init($self);
677 $mkey = 'last_commit';
678 $last_commit = $xdb->get_metadata('last_commit');
682 $mkey = undef if $last_commit ne '';
685 # use last_commit from msgmap if it is older or unset
686 my $lm = $mm->last_commit || '';
687 if (!$lm || ($lm && $lx && is_ancestor($git, $lm, $lx))) {
691 $self->{over}->rollback_lazy;
692 $self->{over}->disconnect;
695 $xdb->cancel_transaction;
696 $xdb = _xdb_release($self);
698 # ensure we leak no FDs to "git log" with Xapian <= 1.2
699 my $range = $lx eq '' ? $tip : "$lx..$tip";
700 $xlog = _git_log($self, $range);
702 $xdb = $self->begin_txn_lazy;
703 } while ($xdb->get_metadata('last_commit') ne $last_commit);
705 my $dbh = $mm->{dbh} if $mm;
707 my ($commit, $newest) = @_;
710 my $cur = $mm->last_commit || '';
711 if (need_update($self, $cur, $newest)) {
712 $mm->last_commit($newest);
717 if ($mkey && $newest) {
718 my $cur = $xdb->get_metadata($mkey);
719 if (need_update($self, $cur, $newest)) {
720 $xdb->set_metadata($mkey, $newest);
723 $self->commit_txn_lazy;
725 $xdb = _xdb_release($self);
726 # let another process do some work... <
728 $xdb = $self->begin_txn_lazy;
729 $dbh->begin_work if $dbh;
734 read_log($self, $xlog, *index_both, *unindex_both, $cb);
738 # order matters for unlocking
739 $_[0]->{xdb} = undef;
740 $_[0]->{lockfh} = undef;
743 # remote_* subs are only used by SearchIdxPart
746 if (my $w = $self->{w}) {
747 print $w "commit\n" or die "failed to write commit: $!";
749 $self->commit_txn_lazy;
755 if (my $w = delete $self->{w}) {
756 my $pid = delete $self->{pid} or die "no process to wait on\n";
757 print $w "close\n" or die "failed to write to pid:$pid: $!\n";
758 close $w or die "failed to close pipe for pid:$pid: $!\n";
759 waitpid($pid, 0) == $pid or die "remote process did not finish";
760 $? == 0 or die ref($self)." pid:$pid exited with: $?";
762 die "transaction in progress $self\n" if $self->{txn};
763 $self->_xdb_release if $self->{xdb};
768 my ($self, $oid, $mid) = @_;
769 if (my $w = $self->{w}) {
770 # triggers remove_by_oid in a partition
771 print $w "D $oid $mid\n" or die "failed to write remove $!";
773 $self->begin_txn_lazy;
774 $self->remove_by_oid($oid, $mid);
780 return if $self->{txn};
782 $self->{-inbox}->with_umask(sub {
783 my $xdb = $self->{xdb} || $self->_xdb_acquire;
784 $self->{over}->begin_lazy if $self->{over};
785 $xdb->begin_transaction;
791 sub commit_txn_lazy {
793 delete $self->{txn} or return;
794 $self->{-inbox}->with_umask(sub {
795 $self->{xdb}->commit_transaction;
796 $self->{over}->commit_lazy if $self->{over};
802 die "$$ $0 xdb not released\n" if $self->{xdb};
803 die "$$ $0 still in transaction\n" if $self->{txn};