1 # Copyright (C) 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
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 parent qw(PublicInbox::Search PublicInbox::Lock Exporter);
14 use PublicInbox::Search qw(xap_terms);
15 use PublicInbox::InboxWritable;
16 use PublicInbox::MID qw(mids_for_index mids);
17 use PublicInbox::MsgIter;
18 use PublicInbox::IdxStack;
19 use Carp qw(croak carp);
20 use POSIX qw(strftime);
21 use Time::Local qw(timegm);
22 use PublicInbox::OverIdx;
23 use PublicInbox::Spawn qw(spawn);
24 use PublicInbox::Git qw(git_unquote);
25 use PublicInbox::MsgTime qw(msg_timestamp msg_datestamp);
26 use PublicInbox::Address;
28 our @EXPORT_OK = qw(log2stack is_ancestor check_size prepare_stack
29 index_text term_generator add_val is_bad_blob);
30 my $X = \%PublicInbox::Search::X;
31 our ($DB_CREATE_OR_OPEN, $DB_OPEN);
33 our $BATCH_BYTES = $ENV{XAPIAN_FLUSH_THRESHOLD} ? 0x7fffffff :
34 # assume a typical 64-bit system has 8x more RAM than a
35 # typical 32-bit system:
36 (($Config{ptrsize} >= 8 ? 8192 : 1024) * 1024);
38 use constant DEBUG => !!$ENV{DEBUG};
40 my $xapianlevels = qr/\A(?:full|medium)\z/;
42 my $OID = $hex .'{40,}';
43 my @VMD_MAP = (kw => 'K', L => 'L');
44 our $INDEXLEVELS = qr/\A(?:full|medium|basic)\z/;
47 my ($class, $ibx, $creat, $shard) = @_;
48 ref $ibx or die "BUG: expected PublicInbox::Inbox object: $ibx";
49 my $inboxdir = $ibx->{inboxdir};
50 my $version = $ibx->version;
51 my $indexlevel = 'full';
52 my $altid = $ibx->{altid};
54 require PublicInbox::AltId;
55 $altid = [ map { PublicInbox::AltId->new($ibx, $_); } @$altid ];
57 if ($ibx->{indexlevel}) {
58 if ($ibx->{indexlevel} =~ $INDEXLEVELS) {
59 $indexlevel = $ibx->{indexlevel};
61 die("Invalid indexlevel $ibx->{indexlevel}\n");
64 $ibx = PublicInbox::InboxWritable->new($ibx);
65 my $self = PublicInbox::Search->new($ibx);
68 $self->{-altid} = $altid;
69 $self->{indexlevel} = $indexlevel;
70 $self->{-set_indexlevel_once} = 1 if $indexlevel eq 'medium';
71 if ($ibx->{-skip_docdata}) {
72 $self->{-set_skip_docdata_once} = 1;
73 $self->{-skip_docdata} = 1;
76 $self->{lock_path} = "$inboxdir/ssoma.lock";
77 my $dir = $self->xdir;
78 $self->{oidx} = PublicInbox::OverIdx->new("$dir/over.sqlite3");
79 $self->{oidx}->{-no_fsync} = 1 if $ibx->{-no_fsync};
80 } elsif ($version == 2) {
81 defined $shard or die "shard is required for v2\n";
83 $self->{shard} = $shard;
84 $self->{lock_path} = undef;
86 die "unsupported inbox version=$version\n";
88 $self->{creat} = ($creat || 0) == 1;
92 sub need_xapian ($) { $_[0]->{indexlevel} =~ $xapianlevels }
95 my ($self, $wake) = @_;
96 if (need_xapian($self)) {
97 my $xdb = delete $self->{xdb} or croak '{xdb} not acquired';
100 $self->lock_release($wake) if $self->{creat};
104 sub load_xapian_writable () {
105 return 1 if $X->{WritableDatabase};
106 PublicInbox::Search::load_xapian() or die "failed to load Xapian: $@\n";
107 my $xap = $PublicInbox::Search::Xap;
108 for (qw(Document TermGenerator WritableDatabase)) {
109 $X->{$_} = $xap.'::'.$_;
111 eval 'require '.$X->{WritableDatabase} or die;
112 *sortable_serialise = $xap.'::sortable_serialise';
113 $DB_CREATE_OR_OPEN = eval($xap.'::DB_CREATE_OR_OPEN()');
114 $DB_OPEN = eval($xap.'::DB_OPEN()');
115 my $ver = (eval($xap.'::major_version()') << 16) |
116 (eval($xap.'::minor_version()') << 8) |
117 eval($xap.'::revision()');
118 $DB_NO_SYNC = 0x4 if $ver >= 0x10400;
119 # Xapian v1.2.21..v1.2.24 were missing close-on-exec on OFD locks
120 $X->{CLOEXEC_UNSET} = 1 if $ver >= 0x010215 && $ver <= 0x010218;
127 my $dir = $self->xdir;
128 if (need_xapian($self)) {
129 croak 'already acquired' if $self->{xdb};
130 load_xapian_writable();
131 $flag = $self->{creat} ? $DB_CREATE_OR_OPEN : $DB_OPEN;
133 if ($self->{creat}) {
137 # don't create empty Xapian directories if we don't need Xapian
138 my $is_shard = defined($self->{shard});
139 if (!-d $dir && (!$is_shard ||
140 ($is_shard && need_xapian($self)))) {
141 File::Path::mkpath($dir);
142 require PublicInbox::Syscall;
143 PublicInbox::Syscall::nodatacow_dir($dir);
144 $self->{-set_has_threadid_once} = 1;
147 return unless defined $flag;
148 $flag |= $DB_NO_SYNC if ($self->{ibx} // $self->{eidx})->{-no_fsync};
149 my $xdb = eval { ($X->{WritableDatabase})->new($dir, $flag) };
150 croak "Failed opening $dir: $@" if $@;
155 my ($doc, $col, $num) = @_;
156 $num = sortable_serialise($num);
157 $doc->add_value($col, $num);
160 sub term_generator ($) { # write-only
163 $self->{term_generator} //= do {
164 my $tg = $X->{TermGenerator}->new;
165 $tg->set_stemmer(PublicInbox::Search::stemmer($self));
170 sub index_phrase ($$$$) {
171 my ($self, $text, $wdf_inc, $prefix) = @_;
173 my $tg = term_generator($self);
174 $tg->index_text($text, $wdf_inc, $prefix);
175 $tg->increase_termpos;
178 sub index_text ($$$$) {
179 my ($self, $text, $wdf_inc, $prefix) = @_;
181 if ($self->{indexlevel} eq 'full') {
182 index_phrase($self, $text, $wdf_inc, $prefix);
184 my $tg = term_generator($self);
185 $tg->index_text_without_positions($text, $wdf_inc, $prefix);
189 sub index_headers ($$) {
190 my ($self, $smsg) = @_;
191 my @x = (from => 'A', to => 'XTO', cc => 'XCC'); # A: Author
192 while (my ($field, $pfx) = splice(@x, 0, 2)) {
193 my $val = $smsg->{$field};
195 # include "(comments)" after the address, too, so not using
196 # PublicInbox::Address::names or pairs
197 index_text($self, $val, 1, $pfx);
199 # we need positional info for email addresses since they
200 # can be considered phrases
201 if ($self->{indexlevel} eq 'medium') {
202 for my $addr (PublicInbox::Address::emails($val)) {
203 index_phrase($self, $addr, 1, $pfx);
207 @x = (subject => 'S');
208 while (my ($field, $pfx) = splice(@x, 0, 2)) {
209 my $val = $smsg->{$field};
210 index_text($self, $val, 1, $pfx) if $val ne '';
214 sub index_diff_inc ($$$$) {
215 my ($self, $text, $pfx, $xnq) = @_;
217 index_text($self, join("\n", @$xnq), 1, 'XNQ');
220 if ($pfx eq 'XDFN') {
221 index_phrase($self, $text, 1, $pfx);
223 index_text($self, $text, 1, $pfx);
227 sub index_old_diff_fn {
228 my ($self, $seen, $fa, $fb, $xnq) = @_;
230 # no renames or space support for traditional diffs,
231 # find the number of leading common paths to strip:
232 my @fa = split('/', $fa);
233 my @fb = split('/', $fb);
234 while (scalar(@fa) && scalar(@fb)) {
235 $fa = join('/', @fa);
236 $fb = join('/', @fb);
238 unless ($seen->{$fa}++) {
239 index_diff_inc($self, $fa, 'XDFN', $xnq);
249 sub index_diff ($$$) {
250 my ($self, $txt, $doc) = @_;
255 foreach (split(/\n/, $txt)) {
256 if ($in_diff && s/^ //) { # diff context
257 index_diff_inc($self, $_, 'XDFCTX', $xnq);
258 } elsif (/^-- $/) { # email signature begins
260 } elsif (m!^diff --git "?[^/]+/.+ "?[^/]+/.+\z!) {
261 # wait until "---" and "+++" to capture filenames
265 } elsif (m/^diff -(.+) (\S+) (\S+)$/) {
266 my ($opt, $fa, $fb) = ($1, $2, $3);
268 # only support unified:
269 next unless $opt =~ /[uU]/;
270 $in_diff = index_old_diff_fn($self, \%seen, $fa, $fb,
272 } elsif (m!^--- ("?[^/]+/.+)!) {
274 $fn = (split('/', git_unquote($fn), 2))[1];
275 $seen{$fn}++ or index_diff_inc($self, $fn, 'XDFN', $xnq);
277 } elsif (m!^\+\+\+ ("?[^/]+/.+)!) {
279 $fn = (split('/', git_unquote($fn), 2))[1];
280 $seen{$fn}++ or index_diff_inc($self, $fn, 'XDFN', $xnq);
282 } elsif (/^--- (\S+)/) {
285 } elsif (defined $in_diff && /^\+\+\+ (\S+)/) {
286 $in_diff = index_old_diff_fn($self, \%seen, $in_diff,
288 } elsif ($in_diff && s/^\+//) { # diff added
289 index_diff_inc($self, $_, 'XDFB', $xnq);
290 } elsif ($in_diff && s/^-//) { # diff removed
291 index_diff_inc($self, $_, 'XDFA', $xnq);
292 } elsif (m!^index ([a-f0-9]+)\.\.([a-f0-9]+)!) {
293 my ($ba, $bb) = ($1, $2);
294 index_git_blob_id($doc, 'XDFPRE', $ba);
295 index_git_blob_id($doc, 'XDFPOST', $bb);
297 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*$/) {
298 # traditional diff w/o -p
299 } elsif (/^@@ (?:\S+) (?:\S+) @@\s*(\S+.*)$/) {
300 # hunk header context
301 index_diff_inc($self, $1, 'XDFHH', $xnq);
302 # ignore the following lines:
303 } elsif (/^(?:dis)similarity index/ ||
304 /^(?:old|new) mode/ ||
305 /^(?:deleted|new) file mode/ ||
306 /^(?:copy|rename) (?:from|to) / ||
307 /^(?:dis)?similarity index / ||
308 /^\\ No newline at end of file/ ||
309 /^Binary files .* differ/) {
312 # possible to be in diff context, some mail may be
313 # stripped by MUA or even GNU diff(1). "git apply"
314 # treats a bare "\n" as diff context, too
317 warn "non-diff line: $_\n" if DEBUG && $_ ne '';
322 index_text($self, join("\n", @xnq), 1, 'XNQ');
325 sub index_xapian { # msg_iter callback
326 my $part = $_[0]->[0]; # ignore $depth and $idx
327 my ($self, $doc) = @{$_[1]};
328 my $ct = $part->content_type || 'text/plain';
329 my $fn = $part->filename;
330 if (defined $fn && $fn ne '') {
331 index_phrase($self, $fn, 1, 'XFN');
333 if ($part->{is_submsg}) {
334 my $mids = mids_for_index($part);
335 index_ids($self, $doc, $part, $mids);
336 my $smsg = bless {}, 'PublicInbox::Smsg';
337 $smsg->populate($part);
338 index_headers($self, $smsg);
341 my ($s, undef) = msg_part_text($part, $ct);
342 defined $s or return;
343 $_[0]->[0] = $part = undef; # free memory
345 # split off quoted and unquoted blocks:
346 my @sections = PublicInbox::MsgIter::split_quotes($s);
347 undef $s; # free memory
348 for my $txt (@sections) {
350 index_text($self, $txt, 0, 'XQUOT');
352 # does it look like a diff?
353 if ($txt =~ /^(?:diff|---|\+\+\+) /ms) {
354 index_diff($self, $txt, $doc);
356 index_text($self, $txt, 1, 'XNQ');
359 undef $txt; # free memory
363 sub index_list_id ($$$) {
364 my ($self, $doc, $hdr) = @_;
365 for my $l ($hdr->header_raw('List-Id')) {
366 $l =~ /<([^>]+)>/ or next;
368 $doc->add_boolean_term('G' . $lid);
369 index_phrase($self, $lid, 1, 'XL'); # probabilistic
373 sub index_ids ($$$$) {
374 my ($self, $doc, $hdr, $mids) = @_;
375 for my $mid (@$mids) {
376 index_phrase($self, $mid, 1, 'XM');
378 # because too many Message-IDs are prefixed with
380 if ($mid =~ /\w{12,}/) {
381 my @long = ($mid =~ /(\w{3,}+)/g);
382 index_phrase($self, join(' ', @long), 1, 'XM');
385 $doc->add_boolean_term('Q' . $_) for @$mids;
386 index_list_id($self, $doc, $hdr);
389 sub eml2doc ($$$;$) {
390 my ($self, $eml, $smsg, $mids) = @_;
391 $mids //= mids_for_index($eml);
392 my $doc = $X->{Document}->new;
393 add_val($doc, PublicInbox::Search::TS(), $smsg->{ts});
394 my @ds = gmtime($smsg->{ds});
395 my $yyyymmdd = strftime('%Y%m%d', @ds);
396 add_val($doc, PublicInbox::Search::YYYYMMDD(), $yyyymmdd);
397 my $dt = strftime('%Y%m%d%H%M%S', @ds);
398 add_val($doc, PublicInbox::Search::DT(), $dt);
399 add_val($doc, PublicInbox::Search::BYTES(), $smsg->{bytes});
400 add_val($doc, PublicInbox::Search::UID(), $smsg->{num});
401 add_val($doc, PublicInbox::Search::THREADID, $smsg->{tid});
403 my $tg = term_generator($self);
404 $tg->set_document($doc);
405 index_headers($self, $smsg);
407 if (defined(my $eidx_key = $smsg->{eidx_key})) {
408 $doc->add_boolean_term('O'.$eidx_key) if $eidx_key ne '.';
410 msg_iter($eml, \&index_xapian, [ $self, $doc ]);
411 index_ids($self, $doc, $eml, $mids);
413 # by default, we maintain compatibility with v1.5.0 and earlier
414 # by writing to docdata.glass, users who never exect to downgrade can
416 if (!$self->{-skip_docdata}) {
417 # WWW doesn't need {to} or {cc}, only NNTP
418 $smsg->{to} = $smsg->{cc} = '';
419 $smsg->parse_references($eml, $mids);
420 my $data = $smsg->to_doc_data;
421 $doc->set_data($data);
424 if (my $altid = $self->{-altid}) {
425 foreach my $alt (@$altid) {
426 my $pfx = $alt->{xprefix};
427 foreach my $mid (@$mids) {
428 my $id = $alt->mid2alt($mid);
429 next unless defined $id;
430 $doc->add_boolean_term($pfx . $id);
437 sub add_xapian ($$$$) {
438 my ($self, $eml, $smsg, $mids) = @_;
439 begin_txn_lazy($self);
440 my $merge_vmd = delete $smsg->{-merge_vmd};
441 my $doc = eml2doc($self, $eml, $smsg, $mids);
442 if (my $old = $merge_vmd ? _get_doc($self, $smsg->{num}) : undef) {
444 while (my ($field, $pfx) = splice(@x, 0, 2)) {
445 for my $term (xap_terms($pfx, $old)) {
446 $doc->add_boolean_term($pfx.$term);
450 $self->{xdb}->replace_document($smsg->{num}, $doc);
453 sub _msgmap_init ($) {
455 die "BUG: _msgmap_init is only for v1\n" if $self->{ibx}->version != 1;
457 require PublicInbox::Msgmap;
458 PublicInbox::Msgmap->new_file($self->{ibx}, 1);
463 # mime = PublicInbox::Eml or Email::MIME object
464 my ($self, $mime, $smsg, $sync) = @_;
465 begin_txn_lazy($self);
466 my $mids = mids_for_index($mime);
467 $smsg //= bless { blob => '' }, 'PublicInbox::Smsg'; # test-only compat
468 $smsg->{mid} //= $mids->[0]; # v1 compatibility
469 $smsg->{num} //= do { # v1
471 index_mm($self, $mime, $smsg->{blob}, $sync);
475 $smsg->populate($mime, $sync);
476 $smsg->{bytes} //= length($mime->as_string);
479 # order matters, overview stores every possible piece of
480 # data in doc_data (deflated). Xapian only stores a subset
481 # of the fields which exist in over.sqlite3. We may stop
482 # storing doc_data in Xapian sometime after we get multi-inbox
484 if (my $oidx = $self->{oidx}) { # v1 only
485 $oidx->add_overview($mime, $smsg);
487 if (need_xapian($self)) {
488 add_xapian($self, $mime, $smsg, $mids);
493 warn "failed to index message <".join('> <',@$mids).">: $@\n";
500 my ($self, $docid) = @_;
501 my $doc = eval { $self->{xdb}->get_document($docid) };
503 warn "E: $@\n" if $@;
504 warn "E: #$docid missing in Xapian\n";
510 my ($self, $docid, $eidx_key, $eml) = @_;
511 begin_txn_lazy($self);
512 my $doc = _get_doc($self, $docid) or return;
513 term_generator($self)->set_document($doc);
515 # '.' is special for lei_store
516 $doc->add_boolean_term('O'.$eidx_key) if $eidx_key ne '.';
518 index_list_id($self, $doc, $eml);
519 $self->{xdb}->replace_document($docid, $doc);
523 my ($self, $pfx, $docid) = @_;
524 begin_txn_lazy($self);
525 xap_terms($pfx, $self->{xdb}, $docid);
528 sub remove_eidx_info {
529 my ($self, $docid, $eidx_key, $eml) = @_;
530 begin_txn_lazy($self);
531 my $doc = _get_doc($self, $docid) or return;
532 eval { $doc->remove_term('O'.$eidx_key) };
533 warn "W: ->remove_term O$eidx_key: $@\n" if $@;
534 for my $l ($eml ? $eml->header_raw('List-Id') : ()) {
535 $l =~ /<([^>]+)>/ or next;
537 eval { $doc->remove_term('G' . $lid) };
538 warn "W: ->remove_term G$lid: $@\n" if $@;
540 # nb: we don't remove the XL probabilistic terms
541 # since terms may overlap if cross-posted.
543 # IOW, a message which has both <foo.example.com>
544 # and <bar.example.com> would have overlapping
545 # "XLexample" and "XLcom" as terms and which we
546 # wouldn't know if they're safe to remove if we just
547 # unindex <foo.example.com> while preserving
550 # In any case, this entire sub is will likely never
551 # be needed and users using the "l:" prefix are probably
554 $self->{xdb}->replace_document($docid, $doc);
558 my ($self, $docid, $vmd) = @_;
559 begin_txn_lazy($self);
560 my $doc = _get_doc($self, $docid) or return;
561 my ($end, @rm, @add);
563 while (my ($field, $pfx) = splice(@x, 0, 2)) {
564 my $set = $vmd->{$field} // next;
565 my %keep = map { $_ => 1 } @$set;
567 $end //= $doc->termlist_end;
568 for (my $cur = $doc->termlist_begin; $cur != $end; $cur++) {
570 last if $cur == $end;
571 my $v = $cur->get_termname;
572 $v =~ s/\A$pfx//s or next;
573 $keep{$v} ? delete($add{$v}) : push(@rm, $pfx.$v);
575 push(@add, map { $pfx.$_ } keys %add);
577 return unless scalar(@rm) || scalar(@add);
578 $doc->remove_term($_) for @rm;
579 $doc->add_boolean_term($_) for @add;
580 $self->{xdb}->replace_document($docid, $doc);
583 sub apply_vmd_mod ($$) {
584 my ($doc, $vmd_mod) = @_;
587 while (my ($field, $pfx) = splice(@x, 0, 2)) {
589 for my $val (@{$vmd_mod->{"-$field"} // []}) {
591 $doc->remove_term($pfx . $val);
595 for my $val (@{$vmd_mod->{"+$field"} // []}) {
596 $doc->add_boolean_term($pfx . $val);
604 my ($self, $docid, $vmd) = @_;
605 begin_txn_lazy($self);
606 my $doc = _get_doc($self, $docid) or return;
609 while (my ($field, $pfx) = splice(@x, 0, 2)) {
610 my $add = $vmd->{$field} // next;
611 $doc->add_boolean_term($pfx . $_) for @$add;
612 $updated += scalar(@$add);
614 $updated += apply_vmd_mod($doc, $vmd);
615 $self->{xdb}->replace_document($docid, $doc) if $updated;
619 my ($self, $docid, $vmd) = @_;
620 begin_txn_lazy($self);
621 my $doc = _get_doc($self, $docid) or return;
624 while (my ($field, $pfx) = splice(@x, 0, 2)) {
625 my $rm = $vmd->{$field} // next;
628 $doc->remove_term($pfx . $_);
633 $self->{xdb}->replace_document($docid, $doc) if $replace;
637 my ($self, $docid, $vmd_mod) = @_;
638 begin_txn_lazy($self);
639 my $doc = _get_doc($self, $docid) or return;
640 my $updated = apply_vmd_mod($doc, $vmd_mod);
641 $self->{xdb}->replace_document($docid, $doc) if $updated;
646 my ($self, @docids) = @_;
647 begin_txn_lazy($self);
648 my $xdb = $self->{xdb} // die 'BUG: missing {xdb}';
649 for my $docid (@docids) {
650 eval { $xdb->delete_document($docid) };
651 warn "E: #$docid not in in Xapian? $@\n" if $@;
655 sub xdb_remove_quiet {
656 my ($self, $docid) = @_;
657 begin_txn_lazy($self);
658 my $xdb = $self->{xdb} // die 'BUG: missing {xdb}';
659 eval { $xdb->delete_document($docid) };
660 ++$self->{-quiet_rm} unless $@;
663 sub nr_quiet_rm { delete($_[0]->{-quiet_rm}) // 0 }
665 sub index_git_blob_id {
666 my ($doc, $pfx, $objid) = @_;
668 my $len = length($objid);
669 for (my $len = length($objid); $len >= 7; ) {
670 $doc->add_term($pfx.$objid);
671 $objid = substr($objid, 0, --$len);
677 my ($self, $oid, $eml) = @_;
678 my $mids = mids($eml);
681 for my $mid (@$mids) {
682 my @removed = $self->{oidx}->remove_oid($oid, $mid);
683 $nr += scalar @removed;
684 $tmp{$_}++ for @removed;
687 my $m = join('> <', @$mids);
688 warn "W: <$m> missing for removal from overview\n";
690 while (my ($num, $nr) = each %tmp) {
691 warn "BUG: $num appears >1 times ($nr) for $oid\n" if $nr != 1;
694 $self->{mm}->num_delete($_) for (keys %tmp);
695 } else { # just in case msgmap and over.sqlite3 become desynched:
696 $self->{mm}->mid_delete($mids->[0]);
698 xdb_remove($self, keys %tmp) if need_xapian($self);
702 my ($self, $mime, $oid, $sync) = @_;
703 my $mids = mids($mime);
704 my $mm = $self->{mm};
705 if ($sync->{reindex}) {
706 my $oidx = $self->{oidx};
707 for my $mid (@$mids) {
708 my ($num, undef) = $oidx->num_mid0_for_oid($oid, $mid);
709 return $num if defined $num;
711 $mm->num_for($mids->[0]) // $mm->mid_insert($mids->[0]);
713 # fallback to num_for since filters like RubyLang set the number
714 $mm->mid_insert($mids->[0]) // $mm->num_for($mids->[0]);
718 sub is_bad_blob ($$$$) {
719 my ($oid, $type, $size, $expect_oid) = @_;
720 if ($type ne 'blob') {
721 carp "W: $expect_oid is not a blob (type=$type)";
724 croak "BUG: $oid != $expect_oid" if $oid ne $expect_oid;
725 $size == 0 ? 1 : 0; # size == 0 means purged
728 sub index_both { # git->cat_async callback
729 my ($bref, $oid, $type, $size, $sync) = @_;
730 return if is_bad_blob($oid, $type, $size, $sync->{oid});
731 my ($nr, $max) = @$sync{qw(nr max)};
734 my $smsg = bless { blob => $oid }, 'PublicInbox::Smsg';
735 $smsg->set_bytes($$bref, $size);
736 my $self = $sync->{sidx};
737 local $self->{current_info} = "$self->{current_info}: $oid";
738 my $eml = PublicInbox::Eml->new($bref);
739 $smsg->{num} = index_mm($self, $eml, $oid, $sync) or
740 die "E: could not generate NNTP article number for $oid";
741 add_message($self, $eml, $smsg, $sync);
743 my $cur_cmt = $sync->{cur_cmt} // die 'BUG: {cur_cmt} missing';
744 ${$sync->{latest_cmt}} = $cur_cmt;
747 sub unindex_both { # git->cat_async callback
748 my ($bref, $oid, $type, $size, $sync) = @_;
749 return if is_bad_blob($oid, $type, $size, $sync->{oid});
750 my $self = $sync->{sidx};
751 local $self->{current_info} = "$self->{current_info}: $oid";
752 unindex_eml($self, $oid, PublicInbox::Eml->new($bref));
753 # may be undef if leftover
754 if (defined(my $cur_cmt = $sync->{cur_cmt})) {
755 ${$sync->{latest_cmt}} = $cur_cmt;
762 ($self->{ibx} // $self->{eidx})->with_umask(@_);
765 # called by public-inbox-index
767 my ($self, $opt) = @_;
768 delete $self->{lock_path} if $opt->{-skip_lock};
769 $self->with_umask(\&_index_sync, $self, $opt);
770 if ($opt->{reindex} && !$opt->{quit} &&
771 !grep(defined, @$opt{qw(since until)})) {
773 delete @again{qw(rethread reindex)};
774 index_sync($self, \%again);
775 $opt->{quit} = $again{quit}; # propagate to caller
779 sub check_size { # check_async cb for -index --max-size=...
780 my ($oid, $type, $size, $arg, $git) = @_;
781 (($type // '') eq 'blob') or die "E: bad $oid in $git->{git_dir}";
782 if ($size <= $arg->{max_size}) {
783 $git->cat_async($oid, $arg->{index_oid}, $arg);
785 warn "W: skipping $oid ($size > $arg->{max_size})\n";
789 sub v1_checkpoint ($$;$) {
790 my ($self, $sync, $stk) = @_;
791 $self->{ibx}->git->async_wait_all;
793 # $newest may be undef
794 my $newest = $stk ? $stk->{latest_cmt} : ${$sync->{latest_cmt}};
795 if (defined($newest)) {
796 my $cur = $self->{mm}->last_commit;
797 if (need_update($self, $sync, $cur, $newest)) {
798 $self->{mm}->last_commit($newest);
801 ${$sync->{max}} = $self->{batch_bytes};
803 $self->{mm}->{dbh}->commit;
804 eval { $self->{mm}->{dbh}->do('PRAGMA optimize') };
805 my $xdb = $self->{xdb};
806 if ($newest && $xdb) {
807 my $cur = $xdb->get_metadata('last_commit');
808 if (need_update($self, $sync, $cur, $newest)) {
809 $xdb->set_metadata('last_commit', $newest);
812 if ($stk) { # all done if $stk is passed
813 # let SearchView know a full --reindex was done so it can
814 # generate ->has_threadid-dependent links
815 if ($xdb && $sync->{reindex} && !ref($sync->{reindex})) {
816 my $n = $xdb->get_metadata('has_threadid');
817 $xdb->set_metadata('has_threadid', '1') if $n ne '1';
819 $self->{oidx}->rethread_done($sync->{-opt}); # all done
821 commit_txn_lazy($self);
822 $sync->{ibx}->git->cleanup;
823 my $nr = ${$sync->{nr}};
824 idx_release($self, $nr);
825 # let another process do some work...
826 if (my $pr = $sync->{-opt}->{-progress}) {
827 $pr->("indexed $nr/$sync->{ntodo}\n") if $nr;
829 if (!$stk && !$sync->{quit}) { # more to come
830 begin_txn_lazy($self);
831 $self->{mm}->{dbh}->begin_work;
837 my ($self, $sync, $stk) = @_;
838 my $git = $sync->{ibx}->git;
839 my $max = $self->{batch_bytes};
842 $sync->{max} = \$max;
843 $sync->{sidx} = $self;
844 $sync->{latest_cmt} = \(my $latest_cmt);
846 $self->{mm}->{dbh}->begin_work;
847 if (my @leftovers = keys %{delete($sync->{D}) // {}}) {
848 warn('W: unindexing '.scalar(@leftovers)." leftovers\n");
849 for my $oid (@leftovers) {
850 last if $sync->{quit};
851 $oid = unpack('H*', $oid);
852 $git->cat_async($oid, \&unindex_both, $sync);
855 if ($sync->{max_size} = $sync->{-opt}->{max_size}) {
856 $sync->{index_oid} = \&index_both;
858 while (my ($f, $at, $ct, $oid, $cur_cmt) = $stk->pop_rec) {
859 my $arg = { %$sync, cur_cmt => $cur_cmt, oid => $oid };
860 last if $sync->{quit};
862 $arg->{autime} = $at;
863 $arg->{cotime} = $ct;
864 if ($sync->{max_size}) {
865 $git->check_async($oid, \&check_size, $arg);
867 $git->cat_async($oid, \&index_both, $arg);
869 v1_checkpoint($self, $sync) if $max <= 0;
870 } elsif ($f eq 'd') {
871 $git->cat_async($oid, \&unindex_both, $arg);
874 v1_checkpoint($self, $sync, $sync->{quit} ? undef : $stk);
877 sub log2stack ($$$) {
878 my ($sync, $git, $range) = @_;
879 my $D = $sync->{D}; # OID_BIN => NR (if reindexing, undef otherwise)
881 if ($sync->{ibx}->version == 1) {
882 my $path = $hex.'{2}/'.$hex.'{38}';
883 $add = qr!\A:000000 100644 \S+ ($OID) A\t$path$!;
884 $del = qr!\A:100644 000000 ($OID) \S+ D\t$path$!;
886 $del = qr!\A:\d{6} 100644 $OID ($OID) [AM]\td$!;
887 $add = qr!\A:\d{6} 100644 $OID ($OID) [AM]\tm$!;
890 # Count the new files so they can be added newest to oldest
891 # and still have numbers increasing from oldest to newest
892 my @cmd = qw(log --raw -r --pretty=tformat:%at-%ct-%H
893 --no-notes --no-color --no-renames --no-abbrev);
894 for my $k (qw(since until)) {
895 my $v = $sync->{-opt}->{$k} // next;
896 next if !$sync->{-opt}->{reindex};
897 push @cmd, "--$k=$v";
899 my $fh = $git->popen(@cmd, $range);
900 my ($at, $ct, $stk, $cmt, $l);
901 while (defined($l = <$fh>)) {
902 return if $sync->{quit};
903 if ($l =~ /\A([0-9]+)-([0-9]+)-($OID)$/o) {
904 ($at, $ct, $cmt) = ($1 + 0, $2 + 0, $3);
905 $stk //= PublicInbox::IdxStack->new($cmt);
906 } elsif ($l =~ /$del/) {
908 if ($D) { # reindex case
909 $D->{pack('H*', $oid)}++;
910 } else { # non-reindex case:
911 $stk->push_rec('d', $at, $ct, $oid, $cmt);
913 } elsif ($l =~ /$add/) {
916 my $oid_bin = pack('H*', $oid);
917 my $nr = --$D->{$oid_bin};
918 delete($D->{$oid_bin}) if $nr <= 0;
919 # nr < 0 (-1) means it never existed
922 $stk->push_rec('m', $at, $ct, $oid, $cmt);
925 close $fh or die "git log failed: \$?=$?";
926 $stk //= PublicInbox::IdxStack->new;
930 sub prepare_stack ($$) {
931 my ($sync, $range) = @_;
932 my $git = $sync->{ibx}->git;
934 if (index($range, '..') < 0) {
935 # don't show annoying git errors to users who run -index
937 $git->qx(qw(rev-parse -q --verify), "$range^0");
938 return PublicInbox::IdxStack->new->read_prepare if $?;
940 $sync->{D} = $sync->{reindex} ? {} : undef; # OID_BIN => NR
941 log2stack($sync, $git, $range);
944 # --is-ancestor requires git 1.8.0+
945 sub is_ancestor ($$$) {
946 my ($git, $cur, $tip) = @_;
947 return 0 unless $git->check($cur);
948 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
949 qw(merge-base --is-ancestor), $cur, $tip ];
950 my $pid = spawn($cmd);
951 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
955 sub need_update ($$$$) {
956 my ($self, $sync, $cur, $new) = @_;
957 my $git = $self->{ibx}->git;
958 $cur //= ''; # XS Search::Xapian ->get_metadata doesn't give undef
960 # don't rewind if --{since,until,before,after} are in use
961 return if $cur ne '' &&
962 grep(defined, @{$sync->{-opt}}{qw(since until)}) &&
963 is_ancestor($git, $new, $cur);
965 return 1 if $cur ne '' && !is_ancestor($git, $cur, $new);
966 my $range = $cur eq '' ? $new : "$cur..$new";
967 chomp(my $n = $git->qx(qw(rev-list --count), $range));
968 ($n eq '' || $n > 0);
971 # The last git commit we indexed with Xapian or SQLite (msgmap)
972 # This needs to account for cases where Xapian or SQLite is
973 # out-of-date with respect to the other.
975 my ($self, $mm) = @_;
976 my $lm = $mm->last_commit || '';
978 if (need_xapian($self)) {
979 $lx = $self->{xdb}->get_metadata('last_commit') || '';
983 # Use last_commit from msgmap if it is older or unset
984 if (!$lm || ($lx && $lm && is_ancestor($self->{ibx}->git, $lm, $lx))) {
990 sub reindex_from ($$) {
991 my ($reindex, $last_commit) = @_;
992 return $last_commit unless $reindex;
993 ref($reindex) eq 'HASH' ? $reindex->{from} : '';
999 # we set {-opt}->{quit} too, so ->index_sync callers
1000 # can abort multi-inbox loops this way
1001 $sync->{quit} = $sync->{-opt}->{quit} = 1;
1002 warn "gracefully quitting\n";
1006 # indexes all unindexed messages (v1 only)
1008 my ($self, $opt) = @_;
1009 my $tip = $opt->{ref} || 'HEAD';
1010 my $ibx = $self->{ibx};
1011 local $self->{current_info} = "$ibx->{inboxdir}";
1012 $self->{batch_bytes} = $opt->{batch_size} // $BATCH_BYTES;
1013 $ibx->git->batch_prepare;
1014 my $pr = $opt->{-progress};
1015 my $sync = { reindex => $opt->{reindex}, -opt => $opt, ibx => $ibx };
1016 my $quit = quit_cb($sync);
1017 local $SIG{QUIT} = $quit;
1018 local $SIG{INT} = $quit;
1019 local $SIG{TERM} = $quit;
1020 my $xdb = $self->begin_txn_lazy;
1021 $self->{oidx}->rethread_prepare($opt);
1022 my $mm = _msgmap_init($self);
1023 if ($sync->{reindex}) {
1024 my $last = $mm->last_commit;
1028 # somebody just blindly added --reindex when indexing
1029 # for the first time, allow it:
1030 undef $sync->{reindex};
1033 my $last_commit = _last_x_commit($self, $mm);
1034 my $lx = reindex_from($sync->{reindex}, $last_commit);
1035 my $range = $lx eq '' ? $tip : "$lx..$tip";
1036 $pr->("counting changes\n\t$range ... ") if $pr;
1037 my $stk = prepare_stack($sync, $range);
1038 $sync->{ntodo} = $stk ? $stk->num_records : 0;
1039 $pr->("$sync->{ntodo}\n") if $pr; # continue previous line
1040 process_stack($self, $sync, $stk) if !$sync->{quit};
1044 # order matters for unlocking
1045 $_[0]->{xdb} = undef;
1046 $_[0]->{lockfh} = undef;
1051 my $xdb = $self->{xdb} || idx_acquire($self);
1052 $self->{oidx}->begin_lazy if $self->{oidx};
1053 $xdb->begin_transaction if $xdb;
1058 sub begin_txn_lazy {
1060 $self->with_umask(\&_begin_txn, $self) if !$self->{txn};
1063 # store 'indexlevel=medium' in v2 shard=0 and v1 (only one shard)
1064 # This metadata is read by Admin::detect_indexlevel:
1065 sub set_metadata_once {
1068 return if $self->{shard}; # only continue if undef or 0, not >0
1069 my $xdb = $self->{xdb};
1071 if (delete($self->{-set_has_threadid_once})) {
1072 $xdb->set_metadata('has_threadid', '1');
1074 if (delete($self->{-set_indexlevel_once})) {
1075 my $level = $xdb->get_metadata('indexlevel');
1076 if (!$level || $level ne 'medium') {
1077 $xdb->set_metadata('indexlevel', 'medium');
1080 if (delete($self->{-set_skip_docdata_once})) {
1081 $xdb->get_metadata('skip_docdata') or
1082 $xdb->set_metadata('skip_docdata', '1');
1088 if (my $eidx = $self->{eidx}) {
1089 $eidx->git->async_wait_all;
1090 $eidx->{transact_bytes} = 0;
1092 if (my $xdb = $self->{xdb}) {
1093 set_metadata_once($self);
1094 $xdb->commit_transaction;
1096 $self->{oidx}->commit_lazy if $self->{oidx};
1099 sub commit_txn_lazy {
1101 delete($self->{txn}) and
1102 $self->with_umask(\&_commit_txn, $self);
1105 sub eidx_shard_new {
1106 my ($class, $eidx, $shard) = @_;
1109 xpfx => $eidx->{xpfx},
1110 indexlevel => $eidx->{indexlevel},
1115 $self->{-set_indexlevel_once} = 1 if $self->{indexlevel} eq 'medium';