1 # Copyright (C) 2018-2020 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # for XOVER, OVER in NNTP, and feeds/homepage/threads in PSGI
5 # Unlike Msgmap, this is an _UNSTABLE_ cache which can be
6 # tweaked/updated over time and rebuilt.
8 # Ghost messages (messages which are only referenced in References/In-Reply-To)
9 # are denoted by a negative NNTP article number.
10 package PublicInbox::OverIdx;
13 use parent qw(PublicInbox::Over);
15 use DBI qw(:sql_types); # SQL_BLOB
16 use PublicInbox::MID qw/id_compress mids_for_index references/;
17 use PublicInbox::Smsg qw(subject_normalized);
18 use Compress::Zlib qw(compress);
23 my $dbh = $self->SUPER::dbh_new($self->{-no_fsync} ? 2 : 1);
25 # 80000 pages (80MiB on SQLite <3.12.0, 320MiB on 3.12.0+)
26 # was found to be good in 2018 during the large LKML import
27 # at the time. This ought to be configurable based on HW
28 # and inbox size; I suspect it's overkill for many inboxes.
29 $dbh->do('PRAGMA cache_size = 80000');
37 my $self = $class->SUPER::new($f);
42 sub get_counter ($$) {
44 my $sth = $dbh->prepare_cached(<<'', undef, 1);
45 SELECT val FROM counter WHERE key = ? LIMIT 1
51 sub adj_counter ($$$) {
52 my ($self, $key, $op) = @_;
53 my $dbh = $self->{dbh};
54 my $sth = $dbh->prepare_cached(<<"");
55 UPDATE counter SET val = val $op 1 WHERE key = ?
59 get_counter($dbh, $key);
62 sub next_tid { adj_counter($_[0], 'thread', '+') }
63 sub next_ghost_num { adj_counter($_[0], 'ghost', '-') }
66 my ($self, $tbl, $id_col, $val_col, $val) = @_;
67 my $dbh = $self->{dbh};
68 my $in = $dbh->prepare_cached(<<"")->execute($val);
69 INSERT OR IGNORE INTO $tbl ($val_col) VALUES (?)
72 my $sth = $dbh->prepare_cached(<<"", undef, 1);
73 SELECT $id_col FROM $tbl WHERE $val_col = ? LIMIT 1
78 $dbh->last_insert_id(undef, undef, $tbl, $id_col);
83 my ($self, $path) = @_;
84 return unless defined $path && $path ne '';
85 id_for($self, 'subject', 'sid', 'path' => $path);
89 my ($self, $mid) = @_;
90 id_for($self, 'msgid', 'id', 'mid' => $mid);
94 my ($self, $num, $tid_ref) = @_;
95 my $dbh = $self->{dbh};
97 my $sth = $dbh->prepare_cached(<<'', undef, 1);
98 SELECT tid FROM over WHERE num = ? LIMIT 1
101 $$tid_ref = $sth->fetchrow_array; # may be undef
103 foreach (qw(over id2num)) {
104 $dbh->prepare_cached(<<"")->execute($num);
105 DELETE FROM $_ WHERE num = ?
110 # this includes ghosts
112 my ($self, $mid, $cols, $cb, @arg) = @_;
113 my $dbh = $self->{dbh};
116 I originally wanted to stuff everything into a single query:
118 SELECT over.* FROM over
119 LEFT JOIN id2num ON over.num = id2num.num
120 LEFT JOIN msgid ON msgid.id = id2num.id
121 WHERE msgid.mid = ? AND over.num >= ?
122 ORDER BY over.num ASC
125 But it's faster broken out (and we're always in a
126 transaction for subroutines in this file)
129 my $sth = $dbh->prepare_cached(<<'', undef, 1);
130 SELECT id FROM msgid WHERE mid = ? LIMIT 1
133 my $id = $sth->fetchrow_array;
134 defined $id or return;
137 $cols = join(',', map { $_ } @$cols);
139 my $prev = get_counter($dbh, 'ghost');
141 $sth = $dbh->prepare_cached(<<"", undef, 1);
142 SELECT num FROM id2num WHERE id = ? AND num >= ?
146 $sth->execute($id, $prev);
147 my $nums = $sth->fetchall_arrayref;
148 my $nr = scalar(@$nums) or return;
149 $prev = $nums->[-1]->[0];
151 $sth = $dbh->prepare_cached(<<"", undef, 1);
152 SELECT $cols FROM over WHERE over.num = ? LIMIT 1
155 $sth->execute($_->[0]);
156 my $smsg = $sth->fetchrow_hashref;
157 $smsg = PublicInbox::Over::load_from_row($smsg);
158 $cb->($self, $smsg, @arg) or return;
160 return if $nr != $lim;
164 sub _resolve_mid_to_tid {
165 my ($self, $smsg, $tid) = @_;
166 my $cur_tid = $smsg->{tid};
168 merge_threads($self, $$tid, $cur_tid);
169 } elsif ($cur_tid > $self->{min_tid}) {
171 } else { # rethreading, queue up dead ghosts
172 $$tid = next_tid($self);
173 my $n = $smsg->{num};
175 $self->{dbh}->prepare_cached(<<'')->execute($$tid, $n);
176 UPDATE over SET tid = ? WHERE num = ?
179 push(@{$self->{-ghosts_to_delete}}, $n);
185 # this will create a ghost as necessary
186 sub resolve_mid_to_tid {
187 my ($self, $mid) = @_;
189 each_by_mid($self, $mid, ['tid'], \&_resolve_mid_to_tid, \$tid);
190 if (my $del = delete $self->{-ghosts_to_delete}) {
191 delete_by_num($self, $_) for @$del;
193 $tid // do { # create a new ghost
194 my $id = mid2id($self, $mid);
195 my $num = next_ghost_num($self);
196 $num < 0 or die "ghost num is non-negative: $num\n";
197 $tid = next_tid($self);
198 my $dbh = $self->{dbh};
199 $dbh->prepare_cached(<<'')->execute($num, $tid);
200 INSERT INTO over (num, tid) VALUES (?,?)
202 $dbh->prepare_cached(<<'')->execute($id, $num);
203 INSERT INTO id2num (id, num) VALUES (?,?)
210 my ($self, $winner_tid, $loser_tid) = @_;
211 return if $winner_tid == $loser_tid;
212 my $dbh = $self->{dbh};
213 $dbh->prepare_cached(<<'')->execute($winner_tid, $loser_tid);
214 UPDATE over SET tid = ? WHERE tid = ?
219 my ($self, $refs, $old_tid) = @_;
223 # first ref *should* be the thread root,
224 # but we can never trust clients to do the right thing
225 my $ref = $refs->[0];
226 $tid = resolve_mid_to_tid($self, $ref);
227 merge_threads($self, $tid, $old_tid) if defined $old_tid;
229 # the rest of the refs should point to this tid:
230 foreach my $i (1..$#$refs) {
232 my $ptid = resolve_mid_to_tid($self, $ref);
233 merge_threads($self, $tid, $ptid);
236 $tid = $old_tid // next_tid($self);
241 sub parse_references ($$$) {
242 my ($smsg, $hdr, $mids) = @_;
243 my $refs = references($hdr);
244 push(@$refs, @$mids) if scalar(@$mids) > 1;
245 return $refs if scalar(@$refs) == 0;
247 # prevent circular references here:
248 my %seen = ( $smsg->{mid} => 1 );
250 foreach my $ref (@$refs) {
251 if (length($ref) > PublicInbox::MID::MAX_MID_SIZE) {
252 warn "References: <$ref> too long, ignoring\n";
255 push(@keep, $ref) unless $seen{$ref}++;
257 $smsg->{references} = '<'.join('> <', @keep).'>' if @keep;
261 # normalize subjects so they are suitable as pathnames for URLs
262 # XXX: consider for removal
263 sub subject_path ($) {
265 $subj = subject_normalized($subj);
266 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
272 my $dd = $smsg->to_doc_data;
278 my ($self, $eml, $smsg) = @_;
279 $smsg->{lines} = $eml->body_raw =~ tr!\n!\n!;
280 my $mids = mids_for_index($eml);
281 my $refs = parse_references($smsg, $eml, $mids);
282 my $subj = $smsg->{subject};
285 $xpath = subject_path($subj);
286 $xpath = id_compress($xpath);
288 add_over($self, $smsg, $mids, $refs, $xpath, ddd_for($smsg));
292 my ($self, $smsg, $mid, $refs, $old_tid, $v) = @_;
293 my $cur_tid = $smsg->{tid};
294 my $n = $smsg->{num};
295 die "num must not be zero for $mid" if !$n;
296 my $cur_valid = $cur_tid > $self->{min_tid};
298 if ($n > 0) { # regular mail
300 $$old_tid //= $cur_tid;
301 merge_threads($self, $$old_tid, $cur_tid);
303 $$old_tid //= next_tid($self);
305 } elsif ($n < 0) { # ghost
306 $$old_tid //= $cur_valid ? $cur_tid : next_tid($self);
307 $$old_tid = link_refs($self, $refs, $$old_tid);
308 delete_by_num($self, $n);
315 my ($self, $smsg, $mids, $refs, $xpath, $ddd) = @_;
318 my $num = $smsg->{num};
321 delete_by_num($self, $num, \$old_tid);
322 $old_tid = undef if ($old_tid // 0) <= $self->{min_tid};
323 foreach my $mid (@$mids) {
325 each_by_mid($self, $mid, ['tid'], \&_add_over,
326 $mid, $refs, \$old_tid, \$v);
327 $v > 1 and warn "BUG: vivified multiple ($v) ghosts for $mid\n";
330 $smsg->{tid} = $vivified ? $old_tid : link_refs($self, $refs, $old_tid);
331 $smsg->{sid} = sid($self, $xpath);
332 my $dbh = $self->{dbh};
333 my $sth = $dbh->prepare_cached(<<'');
334 INSERT INTO over (num, tid, sid, ts, ds, ddd)
338 $sth->bind_param($nc, $num);
339 $sth->bind_param(++$nc, $smsg->{$_}) for (qw(tid sid ts ds));
340 $sth->bind_param(++$nc, $ddd, SQL_BLOB);
342 $sth = $dbh->prepare_cached(<<'');
343 INSERT INTO id2num (id, num) VALUES (?,?)
345 foreach my $mid (@$mids) {
346 my $id = mid2id($self, $mid);
347 $sth->execute($id, $num);
352 my ($self, $smsg, $oid, $removed) = @_;
353 if (!defined($oid) || $smsg->{blob} eq $oid) {
354 delete_by_num($self, $smsg->{num});
355 push @$removed, $smsg->{num};
360 # returns number of removed messages in scalar context,
361 # array of removed article numbers in array context.
362 # $oid may be undef to match only on $mid
364 my ($self, $oid, $mid) = @_;
367 each_by_mid($self, $mid, ['ddd'], \&_remove_oid, $oid, $removed);
371 sub _num_mid0_for_oid {
372 my ($self, $smsg, $oid, $res) = @_;
373 my $blob = $smsg->{blob};
374 return 1 if (!defined($blob) || $blob ne $oid); # continue;
375 @$res = ($smsg->{num}, $smsg->{mid});
379 sub num_mid0_for_oid {
380 my ($self, $oid, $mid) = @_;
383 each_by_mid($self, $mid, ['ddd'], \&_num_mid0_for_oid, $oid, $res);
384 @$res, # ($num, $mid0);
391 CREATE TABLE IF NOT EXISTS over (
392 num INTEGER NOT NULL, /* NNTP article number == IMAP UID */
393 tid INTEGER NOT NULL, /* THREADID (IMAP REFERENCES threading, JMAP) */
394 sid INTEGER, /* Subject ID (IMAP ORDEREDSUBJECT "threading") */
395 ts INTEGER, /* IMAP INTERNALDATE (Received: header, git commit time) */
396 ds INTEGER, /* RFC-2822 sent Date: header, git author time */
397 ddd VARBINARY, /* doc-data-deflated (->to_doc_data, ->load_from_data) */
401 $dbh->do('CREATE INDEX IF NOT EXISTS idx_tid ON over (tid)');
402 $dbh->do('CREATE INDEX IF NOT EXISTS idx_sid ON over (sid)');
403 $dbh->do('CREATE INDEX IF NOT EXISTS idx_ts ON over (ts)');
404 $dbh->do('CREATE INDEX IF NOT EXISTS idx_ds ON over (ds)');
407 CREATE TABLE IF NOT EXISTS counter (
408 key VARCHAR(8) PRIMARY KEY NOT NULL,
409 val INTEGER DEFAULT 0,
413 $dbh->do("INSERT OR IGNORE INTO counter (key) VALUES ('thread')");
414 $dbh->do("INSERT OR IGNORE INTO counter (key) VALUES ('ghost')");
417 CREATE TABLE IF NOT EXISTS subject (
418 sid INTEGER PRIMARY KEY AUTOINCREMENT,
419 path VARCHAR(40) NOT NULL, /* SHA-1 of normalized subject */
424 CREATE TABLE IF NOT EXISTS id2num (
425 id INTEGER NOT NULL, /* <=> msgid.id */
426 num INTEGER NOT NULL,
430 # performance critical:
431 $dbh->do('CREATE INDEX IF NOT EXISTS idx_inum ON id2num (num)');
432 $dbh->do('CREATE INDEX IF NOT EXISTS idx_id ON id2num (id)');
435 CREATE TABLE IF NOT EXISTS msgid (
436 id INTEGER PRIMARY KEY AUTOINCREMENT, /* <=> id2num.id */
437 mid VARCHAR(244) NOT NULL,
445 delete $self->{txn} or return;
446 $self->{dbh}->commit;
451 return if $self->{txn};
452 my $dbh = $self->dbh or return;
454 # $dbh->{Profile} = 2;
460 delete $self->{txn} or return;
461 $self->{dbh}->rollback;
466 die "in transaction" if $self->{txn};
467 $self->SUPER::dbh_close;
472 unless (-r $self->{filename}) {
474 require File::Basename;
475 File::Path::mkpath(File::Basename::dirname($self->{filename}));
478 PublicInbox::Over::dbh($self);
482 sub rethread_prepare {
483 my ($self, $opt) = @_;
484 return unless $opt->{rethread};
486 my $min = $self->{min_tid} = get_counter($self->{dbh}, 'thread') // 0;
487 my $pr = $opt->{-progress};
488 $pr->("rethread min THREADID ".($min + 1)."\n") if $pr && $min;
492 my ($self, $opt) = @_;
493 return unless $opt->{rethread} && $self->{txn};
494 defined(my $min = $self->{min_tid}) or croak('BUG: no min_tid');
495 my $dbh = $self->{dbh} or croak('BUG: no dbh');
496 my $rows = $dbh->selectall_arrayref(<<'', { Slice => {} }, $min);
497 SELECT num,tid FROM over WHERE num < 0 AND tid < ?
499 my $show_id = $dbh->prepare('SELECT id FROM id2num WHERE num = ?');
500 my $show_mid = $dbh->prepare('SELECT mid FROM msgid WHERE id = ?');
501 my $pr = $opt->{-progress};
505 $show_id->execute($r->{num});
506 while (defined(my $id = $show_id->fetchrow_array)) {
508 $show_mid->execute($id);
509 my $mid = $show_mid->fetchrow_array;
510 if (!defined($mid)) {
512 E: ghost NUM=$r->{num} ID=$id THREADID=$r->{tid} has no Message-ID
517 I: ghost $r->{num} <$mid> THREADID=$r->{tid} culled
520 delete_by_num($self, $r->{num});
522 $pr->("I: rethread culled $total ghosts\n") if $pr && $total;
525 # used for cross-inbox search
528 $self->{-eidx_prep} //= do {
529 my $dbh = $self->dbh;
531 INSERT OR IGNORE INTO counter (key) VALUES ('eidx_docid')
534 CREATE TABLE IF NOT EXISTS inboxes (
535 ibx_id INTEGER PRIMARY KEY AUTOINCREMENT,
536 eidx_key VARCHAR(255) NOT NULL, /* {newsgroup} // {inboxdir} */
541 CREATE TABLE IF NOT EXISTS xref3 (
542 docid INTEGER NOT NULL, /* <=> over.num */
543 ibx_id INTEGER NOT NULL, /* <=> inboxes.ibx_id */
544 xnum INTEGER NOT NULL, /* NNTP article number in ibx */
545 oidbin VARBINARY NOT NULL, /* 20-byte SHA-1 or 32-byte SHA-256 */
546 UNIQUE (docid, ibx_id, xnum, oidbin)
549 $dbh->do('CREATE INDEX IF NOT EXISTS idx_docid ON xref3 (docid)');
551 # performance critical, this is not UNIQUE since we may need to
552 # tolerate some old bugs from indexing mirrors
553 $dbh->do('CREATE INDEX IF NOT EXISTS idx_nntp ON '.
554 'xref3 (oidbin,xnum,ibx_id)');
557 CREATE TABLE IF NOT EXISTS eidx_meta (
558 key VARCHAR(255) PRIMARY KEY,
559 val VARCHAR(255) NOT NULL
566 sub eidx_meta { # requires transaction
567 my ($self, $key, $val) = @_;
569 my $sql = 'SELECT val FROM eidx_meta WHERE key = ? LIMIT 1';
570 my $dbh = $self->{dbh};
571 defined($val) or return $dbh->selectrow_array($sql, undef, $key);
573 my $prev = $dbh->selectrow_array($sql, undef, $key);
575 $sql = 'UPDATE eidx_meta SET val = ? WHERE key = ?';
576 $dbh->do($sql, undef, $val, $key);
578 $sql = 'INSERT INTO eidx_meta (key,val) VALUES (?,?)';
579 $dbh->do($sql, undef, $key, $val);
586 get_counter($self->{dbh}, 'eidx_docid');
590 my ($self, $docid, $xnum, $oidhex, $eidx_key) = @_;
592 my $ibx_id = id_for($self, 'inboxes', 'ibx_id', eidx_key => $eidx_key);
593 my $oidbin = pack('H*', $oidhex);
594 my $sth = $self->{dbh}->prepare_cached(<<'');
595 INSERT OR IGNORE INTO xref3 (docid, ibx_id, xnum, oidbin) VALUES (?, ?, ?, ?)
597 $sth->bind_param(1, $docid);
598 $sth->bind_param(2, $ibx_id);
599 $sth->bind_param(3, $xnum);
600 $sth->bind_param(4, $oidbin, SQL_BLOB);
604 # returns remaining reference count to $docid
606 my ($self, $docid, $oidhex, $eidx_key, $rm_eidx_info) = @_;
608 my $oidbin = pack('H*', $oidhex);
610 if (defined $eidx_key) {
611 $ibx_id = id_for($self, 'inboxes', 'ibx_id',
612 eidx_key => $eidx_key);
613 $sth = $self->{dbh}->prepare_cached(<<'');
614 DELETE FROM xref3 WHERE docid = ? AND ibx_id = ? AND oidbin = ?
616 $sth->bind_param(1, $docid);
617 $sth->bind_param(2, $ibx_id);
618 $sth->bind_param(3, $oidbin, SQL_BLOB);
620 $sth = $self->{dbh}->prepare_cached(<<'');
621 DELETE FROM xref3 WHERE docid = ? AND oidbin = ?
623 $sth->bind_param(1, $docid);
624 $sth->bind_param(2, $oidbin, SQL_BLOB);
627 $sth = $self->{dbh}->prepare_cached(<<'', undef, 1);
628 SELECT COUNT(*) FROM xref3 WHERE docid = ?
630 $sth->execute($docid);
631 my $nr = $sth->fetchrow_array;
633 delete_by_num($self, $docid);
634 } elsif (defined($ibx_id) && $rm_eidx_info) {
635 # if deduplication rules in ContentHash change, it's
636 # possible a docid can have multiple rows with the
637 # same ibx_id. This governs whether or not we call
638 # ->shard_remove_eidx_info in ExtSearchIdx.
639 $sth = $self->{dbh}->prepare_cached(<<'', undef, 1);
640 SELECT COUNT(*) FROM xref3 WHERE docid = ? AND ibx_id = ?
642 $sth->execute($docid, $ibx_id);
643 my $count = $sth->fetchrow_array;
644 $$rm_eidx_info = ($count == 0);
649 # for when an xref3 goes missing, this does NOT update {ts}
651 my ($self, $smsg, $oidhex) = @_;
652 my $sth = $self->{dbh}->prepare(<<'');
653 UPDATE over SET ddd = ? WHERE num = ?
655 $smsg->{blob} = $oidhex;
656 $sth->bind_param(1, ddd_for($smsg), SQL_BLOB);
657 $sth->bind_param(2, $smsg->{num});