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 = Search::Xapian::Document->new;
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 my ($s, undef) = msg_part_text($part, $ct);
309 defined $s or return;
312 my @lines = split(/\n/, $s);
313 while (defined(my $l = shift @lines)) {
315 $self->index_body(\@orig, $doc) if @orig;
318 $self->index_body(\@quot, 0) if @quot;
322 $self->index_body(\@quot, 0) if @quot;
323 $self->index_body(\@orig, $doc) if @orig;
326 foreach my $mid (@$mids) {
327 $self->index_text($mid, 1, 'XM');
329 # because too many Message-IDs are prefixed with
331 if ($mid =~ /\w{12,}/) {
332 my @long = ($mid =~ /(\w{3,}+)/g);
333 $self->index_text(join(' ', @long), 1, 'XM');
336 $smsg->{to} = $smsg->{cc} = '';
337 PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids);
338 my $data = $smsg->to_doc_data($oid, $mid0);
339 $doc->set_data($data);
340 if (my $altid = $self->{-altid}) {
341 foreach my $alt (@$altid) {
342 my $pfx = $alt->{xprefix};
343 foreach my $mid (@$mids) {
344 my $id = $alt->mid2alt($mid);
345 next unless defined $id;
346 $doc->add_boolean_term($pfx . $id);
350 $doc->add_boolean_term('Q' . $_) foreach @$mids;
351 $self->{xdb}->replace_document($num, $doc);
355 # mime = Email::MIME object
356 my ($self, $mime, $bytes, $num, $oid, $mid0) = @_;
357 my $mids = mids($mime->header_obj);
358 $mid0 = $mids->[0] unless defined $mid0; # v1 compatibility
359 unless (defined $num) { # v1
361 $num = index_mm($self, $mime);
364 if ($self->{indexlevel} =~ $xapianlevels) {
365 $self->add_xapian($mime, $num, $oid, $mids, $mid0)
367 if (my $over = $self->{over}) {
368 $over->add_overview($mime, $bytes, $num, $oid, $mid0);
373 warn "failed to index message <".join('> <',@$mids).">: $@\n";
379 # returns begin and end PostingIterator
381 my ($self, $termval) = @_;
382 my $db = $self->{xdb};
384 ($db->postlist_begin($termval), $db->postlist_end($termval));
388 my ($self, $termval, $cb) = @_;
389 my $batch_size = 1000; # don't let @ids grow too large to avoid OOM
391 my ($head, $tail) = $self->find_doc_ids($termval);
392 return if $head == $tail;
394 for (; $head != $tail && @ids < $batch_size; $head->inc) {
395 push @ids, $head->get_docid;
402 my ($self, $mid) = @_;
403 my $db = $self->{xdb};
405 $mid = mid_clean($mid);
406 my $over = $self->{over};
409 batch_do($self, 'Q' . $mid, sub {
411 $db->delete_document($_) for @$ids;
412 $over->delete_articles($ids) if $over;
417 warn "failed to remove message <$mid>: $@\n";
419 warn "cannot remove non-existent <$mid>\n";
423 # MID is a hint in V2
425 my ($self, $oid, $mid) = @_;
426 my $db = $self->{xdb};
428 $self->{over}->remove_oid($oid, $mid) if $self->{over};
430 # XXX careful, we cannot use batch_do here since we conditionally
431 # delete documents based on other factors, so we cannot call
432 # find_doc_ids twice.
433 my ($head, $tail) = $self->find_doc_ids('Q' . $mid);
434 return if $head == $tail;
436 # there is only ONE element in @delete unless we
437 # have bugs in our v2writable deduplication check
439 for (; $head != $tail; $head->inc) {
440 my $docid = $head->get_docid;
441 my $doc = $db->get_document($docid);
442 my $smsg = PublicInbox::SearchMsg->wrap($mid);
443 $smsg->load_expand($doc);
444 if ($smsg->{blob} eq $oid) {
445 push(@delete, $docid);
448 $db->delete_document($_) foreach @delete;
452 sub term_generator { # write-only
455 my $tg = $self->{term_generator};
458 $tg = Search::Xapian::TermGenerator->new;
459 $tg->set_stemmer($self->stemmer);
461 $self->{term_generator} = $tg;
464 sub index_git_blob_id {
465 my ($doc, $pfx, $objid) = @_;
467 my $len = length($objid);
468 for (my $len = length($objid); $len >= 7; ) {
469 $doc->add_term($pfx.$objid);
470 $objid = substr($objid, 0, --$len);
475 my ($self, $mime) = @_;
476 my $mid = eval { mid_clean(mid_mime($mime)) };
477 $self->remove_message($mid) if defined $mid;
481 my ($self, $mime) = @_;
482 my $mid = mid_clean(mid_mime($mime));
483 my $mm = $self->{mm};
486 if (defined $self->{regen_down}) {
487 $num = $mm->num_for($mid) and return $num;
489 while (($num = $self->{regen_down}--) > 0) {
490 if ($mm->mid_set($num, $mid) != 0) {
494 } elsif (defined $self->{regen_up}) {
495 $num = $mm->num_for($mid) and return $num;
497 # this is to fixup old bugs due to add-remove-add
498 while (($num = ++$self->{regen_up})) {
499 if ($mm->mid_set($num, $mid) != 0) {
505 $num = $mm->mid_insert($mid) and return $num;
507 # fallback to num_for since filters like RubyLang set the number
512 my ($self, $mime) = @_;
513 $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
517 my ($self, $mime, $bytes, $blob) = @_;
518 my $num = index_mm($self, $mime);
519 add_message($self, $mime, $bytes, $num, $blob);
523 my ($self, $mime) = @_;
524 unindex_blob($self, $mime);
525 unindex_mm($self, $mime);
529 my ($git, $blob, $sizeref) = @_;
531 my $str = $git->cat_file($blob, $sizeref);
532 # fixup bugs from import:
533 $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
534 PublicInbox::MIME->new($str);
540 my ($self, $opts) = @_;
541 $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) })
544 sub batch_adjust ($$$$) {
545 my ($max, $bytes, $batch_cb, $latest) = @_;
549 $batch_cb->($latest);
555 my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
556 my $hex = '[a-f0-9]';
557 my $h40 = $hex .'{40}';
558 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
559 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
560 my $git = $self->{git};
563 my $max = BATCH_BYTES;
568 while (defined($line = <$log>)) {
569 if ($line =~ /$addmsg/o) {
571 if (delete $D{$blob}) {
572 if (defined $self->{regen_down}) {
573 my $num = $self->{regen_down}--;
574 $self->{mm}->num_highwater($num);
578 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
579 batch_adjust(\$max, $bytes, $batch_cb, $latest);
580 $add_cb->($self, $mime, $bytes, $blob);
581 } elsif ($line =~ /$delmsg/o) {
584 } elsif ($line =~ /^commit ($h40)/o) {
590 foreach my $blob (keys %D) {
591 my $mime = do_cat_mail($git, $blob, \$bytes) or next;
592 $del_cb->($self, $mime);
594 $batch_cb->($latest, $newest);
599 die "BUG: _msgmap_init is only for v1\n" if $self->{version} != 1;
600 $self->{mm} ||= eval {
601 require PublicInbox::Msgmap;
602 PublicInbox::Msgmap->new($self->{mainrepo}, 1);
607 my ($self, $range) = @_;
608 my $git = $self->{git};
610 if (index($range, '..') < 0) {
611 # don't show annoying git errrors to users who run -index
613 $git->qx(qw(rev-parse -q --verify), "$range^0");
615 open my $fh, '<', '/dev/null' or
616 die "failed to open /dev/null: $!\n";
621 # Count the new files so they can be added newest to oldest
622 # and still have numbers increasing from oldest to newest
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 ++$fcount while <$fh>;
629 my $high = $self->{mm}->num_highwater;
631 if (index($range, '..') < 0) {
632 if ($high && $high == $fcount) {
633 # fix up old bugs in full indexes which caused messages to
634 # not appear in Msgmap
635 $self->{regen_up} = $high;
637 # normal regen is for for fresh data
638 $self->{regen_down} = $fcount;
641 # Give oldest messages the smallest numbers
642 $self->{regen_down} = $high + $fcount;
645 $git->popen(qw/log --no-notes --no-color --no-renames
646 --raw -r --no-abbrev/, $range);
649 # --is-ancestor requires git 1.8.0+
650 sub is_ancestor ($$$) {
651 my ($git, $cur, $tip) = @_;
652 return 0 unless $git->check($cur);
653 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
654 qw(merge-base --is-ancestor), $cur, $tip ];
655 my $pid = spawn($cmd);
656 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
657 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
661 sub need_update ($$$) {
662 my ($self, $cur, $new) = @_;
663 my $git = $self->{git};
664 return 1 if $cur && !is_ancestor($git, $cur, $new);
665 my $range = $cur eq '' ? $new : "$cur..$new";
666 chomp(my $n = $git->qx(qw(rev-list --count), $range));
667 ($n eq '' || $n > 0);
670 # indexes all unindexed messages (v1 only)
672 my ($self, $opts) = @_;
673 my $tip = $opts->{ref} || 'HEAD';
674 my $reindex = $opts->{reindex};
675 my ($mkey, $last_commit, $lx, $xlog);
676 my $git = $self->{git};
679 my $xdb = $self->begin_txn_lazy;
680 my $mm = _msgmap_init($self);
683 $mkey = 'last_commit';
684 $last_commit = $xdb->get_metadata('last_commit');
688 $mkey = undef if $last_commit ne '';
691 # use last_commit from msgmap if it is older or unset
692 my $lm = $mm->last_commit || '';
693 if (!$lm || ($lm && $lx && is_ancestor($git, $lm, $lx))) {
697 $self->{over}->rollback_lazy;
698 $self->{over}->disconnect;
701 $xdb->cancel_transaction;
702 $xdb = _xdb_release($self);
704 # ensure we leak no FDs to "git log" with Xapian <= 1.2
705 my $range = $lx eq '' ? $tip : "$lx..$tip";
706 $xlog = _git_log($self, $range);
708 $xdb = $self->begin_txn_lazy;
709 } while ($xdb->get_metadata('last_commit') ne $last_commit);
711 my $dbh = $mm->{dbh} if $mm;
713 my ($commit, $newest) = @_;
716 my $cur = $mm->last_commit || '';
717 if (need_update($self, $cur, $newest)) {
718 $mm->last_commit($newest);
723 if ($mkey && $newest && $self->{indexlevel} =~ $xapianlevels) {
724 my $cur = $xdb->get_metadata($mkey);
725 if (need_update($self, $cur, $newest)) {
726 $xdb->set_metadata($mkey, $newest);
729 $self->commit_txn_lazy;
731 $xdb = _xdb_release($self);
732 # let another process do some work... <
734 $xdb = $self->begin_txn_lazy;
735 $dbh->begin_work if $dbh;
740 read_log($self, $xlog, *index_both, *unindex_both, $cb);
744 # order matters for unlocking
745 $_[0]->{xdb} = undef;
746 $_[0]->{lockfh} = undef;
749 # remote_* subs are only used by SearchIdxPart
752 if (my $w = $self->{w}) {
753 print $w "commit\n" or die "failed to write commit: $!";
755 $self->commit_txn_lazy;
761 if (my $w = delete $self->{w}) {
762 my $pid = delete $self->{pid} or die "no process to wait on\n";
763 print $w "close\n" or die "failed to write to pid:$pid: $!\n";
764 close $w or die "failed to close pipe for pid:$pid: $!\n";
765 waitpid($pid, 0) == $pid or die "remote process did not finish";
766 $? == 0 or die ref($self)." pid:$pid exited with: $?";
768 die "transaction in progress $self\n" if $self->{txn};
769 $self->_xdb_release if $self->{xdb};
774 my ($self, $oid, $mid) = @_;
775 if (my $w = $self->{w}) {
776 # triggers remove_by_oid in a partition
777 print $w "D $oid $mid\n" or die "failed to write remove $!";
779 $self->begin_txn_lazy;
780 $self->remove_by_oid($oid, $mid);
786 return if $self->{txn};
788 $self->{-inbox}->with_umask(sub {
789 my $xdb = $self->{xdb} || $self->_xdb_acquire;
790 $self->{over}->begin_lazy if $self->{over};
791 $xdb->begin_transaction;
797 sub commit_txn_lazy {
799 delete $self->{txn} or return;
800 $self->{-inbox}->with_umask(sub {
801 $self->{xdb}->commit_transaction;
802 $self->{over}->commit_lazy if $self->{over};
808 die "$$ $0 xdb not released\n" if $self->{xdb};
809 die "$$ $0 still in transaction\n" if $self->{txn};