X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FExtSearchIdx.pm;h=6b29789a2ed89d33d12e654d6a95f9bbe65cc8fa;hb=28d5a8d647e3ab56cc5570af0d6f3ccf75dc91f2;hp=386e1ceecdb0e4b59b0cf5eb1e5f042e270e2757;hpb=672d146577305baa7f508bd2e33212bba6fdb800;p=public-inbox.git diff --git a/lib/PublicInbox/ExtSearchIdx.pm b/lib/PublicInbox/ExtSearchIdx.pm index 386e1cee..6b29789a 100644 --- a/lib/PublicInbox/ExtSearchIdx.pm +++ b/lib/PublicInbox/ExtSearchIdx.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2020 all contributors +# Copyright (C) 2020-2021 all contributors # License: AGPL-3.0+ # Detached/external index cross inbox search indexing support @@ -20,9 +20,10 @@ use parent qw(PublicInbox::ExtSearch PublicInbox::Lock); use Carp qw(croak carp); use Sys::Hostname qw(hostname); use POSIX qw(strftime); +use File::Glob qw(bsd_glob GLOB_NOSORT); +use PublicInbox::MultiGit; use PublicInbox::Search; -use PublicInbox::SearchIdx qw(crlf_adjust prepare_stack is_ancestor - is_bad_blob); +use PublicInbox::SearchIdx qw(prepare_stack is_ancestor is_bad_blob); use PublicInbox::OverIdx; use PublicInbox::MiscIdx; use PublicInbox::MID qw(mids); @@ -30,7 +31,7 @@ use PublicInbox::V2Writable; use PublicInbox::InboxWritable; use PublicInbox::ContentHash qw(content_hash); use PublicInbox::Eml; -use PublicInbox::DS qw(now); +use PublicInbox::DS qw(now add_timer); use DBI qw(:sql_types); # SQL_BLOB sub new { @@ -44,7 +45,8 @@ sub new { topdir => $dir, creat => $opt->{creat}, ibx_map => {}, # (newsgroup//inboxdir) => $ibx - ibx_list => [], + ibx_active => [], # by config section order + ibx_known => [], # by config section order indexlevel => $l, transact_bytes => 0, total_bytes => 0, @@ -52,44 +54,81 @@ sub new { parallel => 1, lock_path => "$dir/ei.lock", }, __PACKAGE__; - $self->{shards} = $self->count_shards || nproc_shards($opt->{creat}); + $self->{shards} = $self->count_shards || + nproc_shards({ nproc => $opt->{jobs} }); my $oidx = PublicInbox::OverIdx->new("$self->{xpfx}/over.sqlite3"); - $oidx->{-no_fsync} = 1 if $opt->{-no_fsync}; + $self->{-no_fsync} = $oidx->{-no_fsync} = 1 if !$opt->{fsync}; $self->{oidx} = $oidx; $self } sub attach_inbox { - my ($self, $ibx) = @_; + my ($self, $ibx, $types) = @_; $self->{ibx_map}->{$ibx->eidx_key} //= do { - push @{$self->{ibx_list}}, $ibx; + delete $self->{-ibx_ary_known}; # invalidate cache + delete $self->{-ibx_ary_active}; # invalidate cache + $types //= [ qw(active known) ]; + for my $t (@$types) { + push @{$self->{"ibx_$t"}}, $ibx; + } $ibx; } } sub _ibx_attach { # each_inbox callback - my ($ibx, $self) = @_; - attach_inbox($self, $ibx); + my ($ibx, $self, $types) = @_; + attach_inbox($self, $ibx, $types); } sub attach_config { - my ($self, $cfg) = @_; + my ($self, $cfg, $ibxs) = @_; $self->{cfg} = $cfg; - $cfg->each_inbox(\&_ibx_attach, $self); + my $types; + if ($ibxs) { + for my $ibx (@$ibxs) { + $self->{ibx_map}->{$ibx->eidx_key} //= do { + push @{$self->{ibx_active}}, $ibx; + push @{$self->{ibx_known}}, $ibx; + } + } + # invalidate cache + delete $self->{-ibx_ary_known}; + delete $self->{-ibx_ary_active}; + $types = [ 'known' ]; + } + $types //= [ qw(known active) ]; + $cfg->each_inbox(\&_ibx_attach, $self, $types); } sub check_batch_limit ($) { my ($req) = @_; my $self = $req->{self}; my $new_smsg = $req->{new_smsg}; - - # {raw_bytes} may be unset, so just use {bytes} my $n = $self->{transact_bytes} += $new_smsg->{bytes}; # set flag for PublicInbox::V2Writable::index_todo: ${$req->{need_checkpoint}} = 1 if $n >= $self->{batch_bytes}; } +sub apply_boost ($$) { + my ($req, $smsg) = @_; + my $id2pos = $req->{id2pos}; # index in ibx_sorted + my $xr3 = $req->{self}->{oidx}->get_xref3($smsg->{num}, 1); + @$xr3 = sort { + $id2pos->{$a->[0]} <=> $id2pos->{$b->[0]} + || + $a->[1] <=> $b->[1] # break ties with {xnum} + } @$xr3; + my $top_blob = unpack('H*', $xr3->[0]->[2]); + my $new_smsg = $req->{new_smsg}; + return if $top_blob ne $new_smsg->{blob}; # loser + + # replace the old smsg with the more boosted one + $new_smsg->{num} = $smsg->{num}; + $new_smsg->populate($req->{eml}, $req); + $req->{self}->{oidx}->add_overview($req->{eml}, $new_smsg); +} + sub do_xpost ($$) { my ($req, $smsg) = @_; my $self = $req->{self}; @@ -102,17 +141,18 @@ sub do_xpost ($$) { if (my $new_smsg = $req->{new_smsg}) { # 'm' on cross-posted message my $xnum = $req->{xnum}; $self->{oidx}->add_xref3($docid, $xnum, $oid, $eidx_key); - $idx->shard_add_eidx_info($docid, $eidx_key, $eml); - check_batch_limit($req); + $idx->ipc_do('add_eidx_info', $docid, $eidx_key, $eml); + apply_boost($req, $smsg) if $req->{boost_in_use}; } else { # 'd' my $rm_eidx_info; my $nr = $self->{oidx}->remove_xref3($docid, $oid, $eidx_key, \$rm_eidx_info); if ($nr == 0) { $self->{oidx}->eidxq_del($docid); - $idx->shard_remove($docid); + $idx->ipc_do('xdb_remove', $docid); } elsif ($rm_eidx_info) { - $idx->shard_remove_eidx_info($docid, $eidx_key, $eml); + $idx->ipc_do('remove_eidx_info', + $docid, $eidx_key, $eml); $self->{oidx}->eidxq_add($docid); # yes, add } } @@ -134,13 +174,13 @@ sub index_unseen ($) { my $oid = $new_smsg->{blob}; my $ibx = delete $req->{ibx} or die 'BUG: {ibx} unset'; $self->{oidx}->add_xref3($docid, $req->{xnum}, $oid, $ibx->eidx_key); - $idx->index_raw(undef, $eml, $new_smsg, $ibx->eidx_key); + $idx->index_eml($eml, $new_smsg, $ibx->eidx_key); check_batch_limit($req); } sub do_finalize ($) { my ($req) = @_; - if (my $indexed = $req->{indexed}) { + if (my $indexed = $req->{indexed}) { # duplicated messages do_xpost($req, $_) for @$indexed; } elsif (exists $req->{new_smsg}) { # totally unseen messsage index_unseen($req); @@ -165,11 +205,10 @@ sub do_step ($) { # main iterator for adding messages to the index \&ck_existing, $req); return; # ck_existing calls do_step } - delete $req->{cur_smsg}; delete $req->{next_arg}; } - my $mid = shift(@{$req->{mids}}); - last unless defined $mid; + die "BUG: {cur_smsg} still set" if $req->{cur_smsg}; + my $mid = shift(@{$req->{mids}}) // last; my ($id, $prev); $req->{next_arg} = [ $mid, \$id, \$prev ]; # loop again @@ -177,9 +216,8 @@ sub do_step ($) { # main iterator for adding messages to the index do_finalize($req); } -sub _blob_missing ($) { # called when req->{cur_smsg}->{blob} is bad - my ($req) = @_; - my $smsg = $req->{cur_smsg} or die 'BUG: {cur_smsg} missing'; +sub _blob_missing ($$) { # called when $smsg->{blob} is bad + my ($req, $smsg) = @_; my $self = $req->{self}; my $xref3 = $self->{oidx}->get_xref3($smsg->{num}); my @keep = grep(!/:$smsg->{blob}\z/, @$xref3); @@ -197,9 +235,9 @@ sub _blob_missing ($) { # called when req->{cur_smsg}->{blob} is bad sub ck_existing { # git->cat_async callback my ($bref, $oid, $type, $size, $req) = @_; - my $smsg = $req->{cur_smsg} or die 'BUG: {cur_smsg} missing'; + my $smsg = delete $req->{cur_smsg} or die 'BUG: {cur_smsg} missing'; if ($type eq 'missing') { - _blob_missing($req); + _blob_missing($req, $smsg); } elsif (!is_bad_blob($oid, $type, $size, $smsg->{blob})) { my $self = $req->{self} // die 'BUG: {self} missing'; local $self->{current_info} = "$self->{current_info} $oid"; @@ -220,8 +258,7 @@ sub cur_ibx_xnum ($$) { $req->{eml} = PublicInbox::Eml->new($bref); $req->{chash} = content_hash($req->{eml}); $req->{mids} = mids($req->{eml}); - my @q = @{$req->{mids}}; # copy - while (defined(my $mid = shift @q)) { + for my $mid (@{$req->{mids}}) { my ($id, $prev); while (my $x = $ibx->over->next_by_mid($mid, \$id, \$prev)) { return $x->{num} if $x->{blob} eq $req->{oid}; @@ -238,7 +275,7 @@ sub index_oid { # git->cat_async callback for 'm' my $new_smsg = $req->{new_smsg} = bless { blob => $oid, }, 'PublicInbox::Smsg'; - $new_smsg->{bytes} = $size + crlf_adjust($$bref); + $new_smsg->set_bytes($$bref, $size); defined($req->{xnum} = cur_ibx_xnum($req, $bref)) or return; ++${$req->{nr}}; do_step($req); @@ -307,27 +344,25 @@ sub _sync_inbox ($$$) { sub gc_unref_doc ($$$$) { my ($self, $ibx_id, $eidx_key, $docid) = @_; - my $dbh = $self->{oidx}->dbh; - + my $remain = 0; # for debug/info purposes, oids may no longer be accessible + my $dbh = $self->{oidx}->dbh; my $sth = $dbh->prepare_cached(<<'', undef, 1); SELECT oidbin FROM xref3 WHERE docid = ? AND ibx_id = ? $sth->execute($docid, $ibx_id); my @oid = map { unpack('H*', $_->[0]) } @{$sth->fetchall_arrayref}; - - $dbh->prepare_cached(<<'')->execute($docid, $ibx_id); -DELETE FROM xref3 WHERE docid = ? AND ibx_id = ? - - my $remain = $self->{oidx}->get_xref3($docid); - if (scalar(@$remain)) { + for my $oid (@oid) { + $remain += $self->{oidx}->remove_xref3($docid, $oid, $eidx_key); + } + if ($remain) { $self->{oidx}->eidxq_add($docid); # enqueue for reindex for my $oid (@oid) { warn "I: unref #$docid $eidx_key $oid\n"; } } else { warn "I: remove #$docid $eidx_key @oid\n"; - $self->idx_shard($docid)->shard_remove($docid); + $self->idx_shard($docid)->ipc_do('xdb_remove', $docid); } } @@ -338,9 +373,11 @@ sub eidx_gc { $self->idx_init($opt); # acquire lock via V2Writable::_idx_init my $dbh = $self->{oidx}->dbh; + $dbh->do('PRAGMA case_sensitive_like = ON'); # only place we use LIKE my $x3_doc = $dbh->prepare('SELECT docid FROM xref3 WHERE ibx_id = ?'); my $ibx_ck = $dbh->prepare('SELECT ibx_id,eidx_key FROM inboxes'); - my $lc_i = $dbh->prepare('SELECT key FROM eidx_meta WHERE key LIKE ?'); + my $lc_i = $dbh->prepare(<<''); +SELECT key FROM eidx_meta WHERE key LIKE ? ESCAPE ? $ibx_ck->execute; while (my ($ibx_id, $eidx_key) = $ibx_ck->fetchrow_array) { @@ -356,8 +393,8 @@ DELETE FROM inboxes WHERE ibx_id = ? # drop last_commit info my $pat = $eidx_key; - $pat =~ s/([_%])/\\$1/g; - $lc_i->execute("lc-%:$pat//%"); + $pat =~ s/([_%\\])/\\$1/g; + $lc_i->execute("lc-%:$pat//%", '\\'); while (my ($key) = $lc_i->fetchrow_array) { next if $key !~ m!\Alc-v[1-9]+:\Q$eidx_key\E//!; warn "I: removing $key\n"; @@ -383,6 +420,11 @@ DELETE FROM xref3 WHERE docid NOT IN (SELECT num FROM over) warn "I: eliminated $nr stale xref3 entries\n" if $nr != 0; + # fixup from old bugs: + $nr = $dbh->do(<<''); +DELETE FROM over WHERE num NOT IN (SELECT docid FROM xref3) + + warn "I: eliminated $nr stale over entries\n" if $nr != 0; done($self); } @@ -390,7 +432,35 @@ sub _ibx_for ($$$) { my ($self, $sync, $smsg) = @_; my $ibx_id = delete($smsg->{ibx_id}) // die '{ibx_id} unset'; my $pos = $sync->{id2pos}->{$ibx_id} // die "$ibx_id no pos"; - $self->{ibx_list}->[$pos] // die "BUG: ibx for $smsg->{blob} not mapped" + $self->{-ibx_ary_known}->[$pos] // + die "BUG: ibx for $smsg->{blob} not mapped" +} + +sub _fd_constrained ($) { + my ($self) = @_; + $self->{-fd_constrained} //= do { + my $soft; + if (eval { require BSD::Resource; 1 }) { + my $NOFILE = BSD::Resource::RLIMIT_NOFILE(); + ($soft, undef) = BSD::Resource::getrlimit($NOFILE); + } else { + chomp($soft = `sh -c 'ulimit -n'`); + } + if (defined($soft)) { + # $want is an estimate + my $want = scalar(@{$self->{ibx_active}}) + 64; + my $ret = $want > $soft; + if ($ret) { + warn <index_raw(undef, $eml, $smsg, $ibx->eidx_key); + $idx->index_eml($eml, $smsg, $ibx->eidx_key); for my $x (reverse @$stable) { $ibx = _ibx_for($self, $sync, $x); my $hdr = delete $x->{hdr} // die 'BUG: no {hdr}'; - $idx->shard_add_eidx_info($docid, $ibx->eidx_key, $hdr); + $idx->ipc_do('add_eidx_info', $docid, $ibx->eidx_key, $hdr); } return if $nr == 1; # likely, all good @@ -429,11 +499,16 @@ sub _reindex_finalize ($$$) { my $x = pop(@$ary) // die "BUG: #$docid {by_chash} empty"; $x->{num} = delete($x->{xnum}) // die '{xnum} unset'; $ibx = _ibx_for($self, $sync, $x); - my $e = $ibx->over->get_art($x->{num}); - $e->{blob} eq $x->{blob} or die <over) { + my $e = $over->get_art($x->{num}); + $e->{blob} eq $x->{blob} or die <{blob} != $e->{blob} (${\$ibx->eidx_key}:$e->{num}); EOF - push @todo, $ibx, $e; + push @todo, $ibx, $e; + $over->dbh_close if _fd_constrained($self); + } else { + die "$ibx->{inboxdir}: over.sqlite3 unusable: $!\n"; + } } undef $by_chash; while (my ($ibx, $e) = splice(@todo, 0, 2)) { @@ -452,19 +527,19 @@ sub _reindex_oid { # git->cat_async callback my $remain = $self->{oidx}->remove_xref3($docid, $expect_oid); if ($remain == 0) { warn "W: #$docid gone or corrupted\n"; - $self->idx_shard($docid)->shard_remove($docid); + $self->idx_shard($docid)->ipc_do('xdb_remove', $docid); } elsif (my $next_oid = $req->{xr3r}->[++$req->{ix}]->[2]) { $self->git->cat_async($next_oid, \&_reindex_oid, $req); } else { warn "BUG: #$docid gone (UNEXPECTED)\n"; - $self->idx_shard($docid)->shard_remove($docid); + $self->idx_shard($docid)->ipc_do('xdb_remove', $docid); } return; } my $ci = $self->{current_info}; local $self->{current_info} = "$ci #$docid $oid"; my $re_smsg = bless { blob => $oid }, 'PublicInbox::Smsg'; - $re_smsg->{bytes} = $size + crlf_adjust($$bref); + $re_smsg->set_bytes($$bref, $size); my $eml = PublicInbox::Eml->new($bref); $re_smsg->populate($eml, { autime => $orig_smsg->{ds}, cotime => $orig_smsg->{ts} }); @@ -491,14 +566,14 @@ sub _reindex_smsg ($$$) { BUG? #$docid $smsg->{blob} is not referenced by inboxes during reindex $self->{oidx}->delete_by_num($docid); - $self->idx_shard($docid)->shard_remove($docid); + $self->idx_shard($docid)->ipc_do('xdb_remove', $docid); return; } - # we sort {xr3r} in the reverse order of {ibx_list} so we can + # we sort {xr3r} in the reverse order of ibx_sorted so we can # hit the common case in _reindex_finalize without rereading # from git (or holding multiple messages in memory). - my $id2pos = $sync->{id2pos}; # index in {ibx_list} + my $id2pos = $sync->{id2pos}; # index in ibx_sorted @$xr3 = sort { $id2pos->{$b->[0]} <=> $id2pos->{$a->[0]} || @@ -592,6 +667,25 @@ EOF undef; } +sub ibx_sorted ($$) { + my ($self, $type) = @_; + $self->{"-ibx_ary_$type"} //= do { + # highest boost first, stable for config-ordering tiebreaker + use sort 'stable'; + [ sort { + ($b->{boost} // 0) <=> ($a->{boost} // 0) + } @{$self->{'ibx_'.$type} // die "BUG: $type unknown"} ]; + } +} + +sub prep_id2pos ($) { + my ($self) = @_; + my %id2pos; + my $pos = 0; + $id2pos{$_->{-ibx_id}} = $pos++ for (@{ibx_sorted($self, 'known')}); + \%id2pos; +} + sub eidxq_process ($$) { # for reindexing my ($self, $sync) = @_; @@ -599,19 +693,14 @@ sub eidxq_process ($$) { # for reindexing my $dbh = $self->{oidx}->dbh; my $tot = $dbh->selectrow_array('SELECT COUNT(*) FROM eidxq') or return; ${$sync->{nr}} = 0; - $sync->{-regen_fmt} = "%u/$tot\n"; + local $sync->{-regen_fmt} = "%u/$tot\n"; my $pr = $sync->{-opt}->{-progress}; if ($pr) { my $min = $dbh->selectrow_array('SELECT MIN(docid) FROM eidxq'); my $max = $dbh->selectrow_array('SELECT MAX(docid) FROM eidxq'); $pr->("Xapian indexing $min..$max (total=$tot)\n"); } - $sync->{id2pos} //= do { - my %id2pos; - my $pos = 0; - $id2pos{$_->{-ibx_id}} = $pos++ for @{$self->{ibx_list}}; - \%id2pos; - }; + $sync->{id2pos} //= prep_id2pos($self); my ($del, $iter); restart: $del = $dbh->prepare('DELETE FROM eidxq WHERE docid = ?'); @@ -644,7 +733,7 @@ sub _reindex_unseen { # git->cat_async callback my $self = $req->{self} // die 'BUG: {self} unset'; local $self->{current_info} = "$self->{current_info} $oid"; my $new_smsg = bless { blob => $oid, }, 'PublicInbox::Smsg'; - $new_smsg->{bytes} = $size + crlf_adjust($$bref); + $new_smsg->set_bytes($$bref, $size); my $eml = $req->{eml} = PublicInbox::Eml->new($bref); $req->{new_smsg} = $new_smsg; $req->{chash} = content_hash($eml); @@ -678,7 +767,8 @@ sub _reindex_check_unseen ($$$) { my $msgs; my $pr = $sync->{-opt}->{-progress}; my $ekey = $ibx->eidx_key; - $sync->{-regen_fmt} = "$ekey checking unseen %u/".$ibx->over->max."\n"; + local $sync->{-regen_fmt} = + "$ekey checking unseen %u/".$ibx->over->max."\n"; ${$sync->{nr}} = 0; while (scalar(@{$msgs = $ibx->over->query_xover($beg, $end)})) { @@ -721,7 +811,7 @@ sub _reindex_check_stale ($$$) { my $pr = $sync->{-opt}->{-progress}; my $fetching; my $ekey = $ibx->eidx_key; - $sync->{-regen_fmt} = + local $sync->{-regen_fmt} = "$ekey check stale/missing %u/".$ibx->over->max."\n"; ${$sync->{nr}} = 0; do { @@ -742,16 +832,16 @@ ORDER BY docid,xnum ASC LIMIT 10000 $fetching = $min = $docid; my $smsg = $ibx->over->get_art($xnum); - my $oidhex = unpack('H*', $oidbin); my $err; if (!$smsg) { $err = 'stale'; - } elsif ($smsg->{blob} ne $oidhex) { + } elsif (pack('H*', $smsg->{blob}) ne $oidbin) { $err = "mismatch (!= $smsg->{blob})"; } else { next; # likely, all good } # current_info already has eidx_key + my $oidhex = unpack('H*', $oidbin); warn "$xnum:$oidhex (#$docid): $err\n"; my $del = $self->{oidx}->dbh->prepare_cached(<<''); DELETE FROM xref3 WHERE ibx_id = ? AND xnum = ? AND oidbin = ? @@ -767,10 +857,10 @@ DELETE FROM xref3 WHERE ibx_id = ? AND xnum = ? AND oidbin = ? if (scalar(@$xr3) == 0) { # all gone $self->{oidx}->delete_by_num($docid); $self->{oidx}->eidxq_del($docid); - $idx->shard_remove($docid); + $idx->ipc_do('xdb_remove', $docid); } else { # enqueue for reindex of remaining messages - $idx->shard_remove_eidx_info($docid, - $ibx->eidx_key); + $idx->ipc_do('remove_eidx_info', + $docid, $ibx->eidx_key); $self->{oidx}->eidxq_add($docid); # yes, add } } @@ -799,7 +889,7 @@ sub eidx_reindex { warn "E: aborting --reindex\n"; return; } - for my $ibx (@{$self->{ibx_list}}) { + for my $ibx (@{ibx_sorted($self, 'active')}) { _reindex_inbox($self, $sync, $ibx); last if $sync->{quit}; } @@ -807,12 +897,124 @@ sub eidx_reindex { eidxq_process($self, $sync) unless $sync->{quit}; } +sub sync_inbox { + my ($self, $sync, $ibx) = @_; + my $err = _sync_inbox($self, $sync, $ibx); + delete @$ibx{qw(mm over)}; + warn $err, "\n" if defined($err); +} + +sub dd_smsg { # git->cat_async callback + my ($bref, $oid, $type, $size, $dd) = @_; + my $smsg = $dd->{smsg} // die 'BUG: dd->{smsg} missing'; + my $self = $dd->{self} // die 'BUG: {self} missing'; + my $per_mid = $dd->{per_mid} // die 'BUG: {per_mid} missing'; + if ($type eq 'missing') { + _blob_missing($dd, $smsg); + } elsif (!is_bad_blob($oid, $type, $size, $smsg->{blob})) { + local $self->{current_info} = "$self->{current_info} $oid"; + my $chash = content_hash(PublicInbox::Eml->new($bref)); + push(@{$per_mid->{dd_chash}->{$chash}}, $smsg); + } + return if $per_mid->{last_smsg} != $smsg; + while (my ($chash, $ary) = each %{$per_mid->{dd_chash}}) { + my $keep = shift @$ary; + next if !scalar(@$ary); + $per_mid->{sync}->{dedupe_cull} += scalar(@$ary); + print STDERR + "# <$keep->{mid}> keeping #$keep->{num}, dropping ", + join(', ', map { "#$_->{num}" } @$ary),"\n"; + next if $per_mid->{sync}->{-opt}->{'dry-run'}; + my $oidx = $self->{oidx}; + for my $smsg (@$ary) { + my $gone = $smsg->{num}; + $oidx->merge_xref3($keep->{num}, $gone, $smsg->{blob}); + $self->idx_shard($gone)->ipc_do('xdb_remove', $gone); + $oidx->delete_by_num($gone); + } + } +} + +sub eidx_dedupe ($$$) { + my ($self, $sync, $msgids) = @_; + $sync->{dedupe_cull} = 0; + my $candidates = 0; + my $nr_mid = 0; + return unless eidxq_lock_acquire($self); + my ($iter, $cur_mid); + my $min_id = 0; + my $idx = 0; + my ($max_id) = $self->{oidx}->dbh->selectrow_array(<{-regen_fmt} = "dedupe %u/$max_id\n"; + + # note: we could write this query more intelligently, + # but that causes lock contention with read-only processes +dedupe_restart: + $cur_mid = $msgids->[$idx]; + if ($cur_mid eq '') { # all Message-IDs + $iter = $self->{oidx}->dbh->prepare(< ? ORDER BY id ASC +EOS + $iter->execute($min_id); + } else { + $iter = $self->{oidx}->dbh->prepare(< ? ORDER BY id ASC +EOS + $iter->execute($cur_mid, $min_id); + } + while (my ($mid, $id) = $iter->fetchrow_array) { + last if $sync->{quit}; + $self->{current_info} = "dedupe $mid"; + ${$sync->{nr}} = $min_id = $id; + my ($prv, @smsg); + while (my $x = $self->{oidx}->next_by_mid($mid, \$id, \$prv)) { + push @smsg, $x; + } + next if scalar(@smsg) < 2; + my $per_mid = { + dd_chash => {}, # chash => [ary of smsgs] + last_smsg => $smsg[-1], + sync => $sync + }; + $nr_mid++; + $candidates += scalar(@smsg) - 1; + for my $smsg (@smsg) { + my $dd = { + per_mid => $per_mid, + smsg => $smsg, + self => $self, + }; + $self->git->cat_async($smsg->{blob}, \&dd_smsg, $dd); + } + # need to wait on every single one @smsg contents can get + # invalidated inside dd_smsg for messages with multiple + # Message-IDs. + $self->git->async_wait_all; + + if (checkpoint_due($sync)) { + undef $iter; + reindex_checkpoint($self, $sync); + goto dedupe_restart; + } + } + goto dedupe_restart if defined($msgids->[++$idx]); + + my $n = delete $sync->{dedupe_cull}; + if (my $pr = $sync->{-opt}->{-progress}) { + $pr->("culled $n/$candidates candidates ($nr_mid msgids)\n"); + } + ${$sync->{nr}} = 0; +} + sub eidx_sync { # main entry point my ($self, $opt) = @_; - my $warn_cb = $SIG{__WARN__} || sub { print STDERR @_ }; + my $warn_cb = $SIG{__WARN__} || \&CORE::warn; local $self->{current_info} = ''; local $SIG{__WARN__} = sub { + return if PublicInbox::Eml::warn_ignore(@_); $warn_cb->($self->{current_info}, ': ', @_); }; $self->idx_init($opt); # acquire lock via V2Writable::_idx_init @@ -833,26 +1035,37 @@ sub eidx_sync { # main entry point local $SIG{QUIT} = $quit; local $SIG{INT} = $quit; local $SIG{TERM} = $quit; - for my $ibx (@{$self->{ibx_list}}) { + for my $ibx (@{ibx_sorted($self, 'known')}) { $ibx->{-ibx_id} //= $self->{oidx}->ibx_id($ibx->eidx_key); } + + if (scalar(grep { defined($_->{boost}) } @{$self->{ibx_known}})) { + $sync->{id2pos} //= prep_id2pos($self); + $sync->{boost_in_use} = 1; + } + + if (my $msgids = delete($opt->{dedupe})) { + local $sync->{checkpoint_unlocks} = 1; + eidx_dedupe($self, $sync, $msgids); + } if (delete($opt->{reindex})) { - $sync->{checkpoint_unlocks} = 1; + local $sync->{checkpoint_unlocks} = 1; eidx_reindex($self, $sync); } # don't use $_ here, it'll get clobbered by reindex_checkpoint - for my $ibx (@{$self->{ibx_list}}) { - last if $sync->{quit}; - my $err = _sync_inbox($self, $sync, $ibx); - delete @$ibx{qw(mm over)}; - warn $err, "\n" if defined($err); + if ($opt->{scan} // 1) { + for my $ibx (@{ibx_sorted($self, 'active')}) { + last if $sync->{quit}; + sync_inbox($self, $sync, $ibx); + } } $self->{oidx}->rethread_done($opt) unless $sync->{quit}; eidxq_process($self, $sync) unless $sync->{quit}; eidxq_release($self); - PublicInbox::V2Writable::done($self); + done($self); + $sync; # for eidx_watch } sub update_last_commit { # overrides V2Writable @@ -886,57 +1099,226 @@ sub update_last_commit { # overrides V2Writable sub _idx_init { # with_umask callback my ($self, $opt) = @_; - PublicInbox::V2Writable::_idx_init($self, $opt); + PublicInbox::V2Writable::_idx_init($self, $opt); # acquires ei.lock $self->{midx} = PublicInbox::MiscIdx->new($self); } +sub symlink_packs ($$) { + my ($ibx, $pd) = @_; + my $ret = 0; + my $glob = "$ibx->{inboxdir}/git/*.git/objects/pack/*.idx"; + for my $idx (bsd_glob($glob, GLOB_NOSORT)) { + my $src = substr($idx, 0, -length('.idx')); + my $dst = $pd . substr($src, rindex($src, '/')); + if (-f "$src.pack" and + symlink("$src.pack", "$dst.pack") and + symlink($idx, "$dst.idx") and + -f $idx) { + ++$ret; + # .promisor, .bitmap, .rev and .keep are optional + # XXX should we symlink .keep here? + for my $s (qw(promisor bitmap rev)) { + symlink("$src.$s", "$dst.$s") if -f "$src.$s"; + } + } elsif (!$!{EEXIST}) { + warn "W: ln -s $src.{pack,idx} => $dst.*: $!\n"; + unlink "$dst.pack", "$dst.idx"; + } + } + $ret; +} + sub idx_init { # similar to V2Writable my ($self, $opt) = @_; return if $self->{idx_shards}; $self->git->cleanup; - - my $ALL = $self->git->{git_dir}; # ALL.git - PublicInbox::Import::init_bare($ALL) unless -d $ALL; - my $info_dir = "$ALL/objects/info"; - my $alt = "$info_dir/alternates"; my $mode = 0644; - my (@old, @new, %seen); # seen: st_dev + st_ino - if (-e $alt) { - open(my $fh, '<', $alt) or die "open $alt: $!"; - $mode = (stat($fh))[2] & 07777; - while (my $line = <$fh>) { - chomp(my $d = $line); - if (my @st = stat($d)) { - next if $seen{"$st[0]\0$st[1]"}++; - } else { - warn "W: stat($d) failed (from $alt): $!\n"; - next if $opt->{-idx_gc}; + my $ALL = $self->git->{git_dir}; # topdir/ALL.git + my ($has_new, $alt, $seen); + if ($opt->{-private}) { # LeiStore + my $local = "$self->{topdir}/local"; # lei/store + $self->{mg} //= PublicInbox::MultiGit->new($self->{topdir}, + 'ALL.git', 'local'); + $mode = 0600; + unless (-d $ALL) { + umask 077; # don't bother restoring for lei + PublicInbox::Import::init_bare($ALL); + $self->git->qx(qw(config core.sharedRepository 0600)); + } + ($alt, $seen) = $self->{mg}->read_alternates(\$mode); + $has_new = $self->{mg}->merge_epochs($alt, $seen); + } else { # extindex has no epochs + $self->{mg} //= PublicInbox::MultiGit->new($self->{topdir}, + 'ALL.git'); + ($alt, $seen) = $self->{mg}->read_alternates(\$mode, + $opt->{-idx_gc}); + PublicInbox::Import::init_bare($ALL); + } + + # git-multi-pack-index(1) can speed up "git cat-file" startup slightly + my $git_midx = 0; + my $pd = "$ALL/objects/pack"; + if (opendir(my $dh, $pd)) { # drop stale symlinks + while (defined(my $dn = readdir($dh))) { + if ($dn =~ /\.(?:idx|pack|promisor|bitmap|rev)\z/) { + my $f = "$pd/$dn"; + unlink($f) if -l $f && !-e $f; } - push @old, $line; } + } elsif ($!{ENOENT}) { + mkdir($pd) or die "mkdir($pd): $!"; + } else { + die "opendir($pd): $!"; } - for my $ibx (@{$self->{ibx_list}}) { - my $line = $ibx->git->{git_dir} . "/objects\n"; - chomp(my $d = $line); + my $new = ''; + for my $ibx (@{ibx_sorted($self, 'active')}) { + # create symlinks for multi-pack-index + $git_midx += symlink_packs($ibx, $pd); + # add new lines to our alternates file + my $d = $ibx->git->{git_dir} . '/objects'; + next if exists $alt->{$d}; if (my @st = stat($d)) { - next if $seen{"$st[0]\0$st[1]"}++; + next if $seen->{"$st[0]\0$st[1]"}++; } else { warn "W: stat($d) failed (from $ibx->{inboxdir}): $!\n"; next if $opt->{-idx_gc}; } - push @new, $line; - } - if (scalar @new) { - push @old, @new; - my $o = \@old; - PublicInbox::V2Writable::write_alternates($info_dir, $mode, $o); + $new .= "$d\n"; } + ($has_new || $new ne '') and + $self->{mg}->write_alternates($mode, $alt, $new); + $git_midx and $self->with_umask(sub { + my @cmd = ('multi-pack-index'); + push @cmd, '--no-progress' if ($opt->{quiet}//0) > 1; + my $lk = $self->lock_for_scope; + system('git', "--git-dir=$ALL", @cmd, 'write'); + # ignore errors, fairly new command, may not exist + }); $self->parallel_init($self->{indexlevel}); $self->with_umask(\&_idx_init, $self, $opt); $self->{oidx}->begin_lazy; $self->{oidx}->eidx_prep; - $self->{midx}->begin_txn; + $self->{midx}->create_xdb if $new ne ''; +} + +sub _watch_commit { # PublicInbox::DS::add_timer callback + my ($self) = @_; + delete $self->{-commit_timer}; + eidxq_process($self, $self->{-watch_sync}); + eidxq_release($self); + my $fmt = delete $self->{-watch_sync}->{-regen_fmt}; + reindex_checkpoint($self, $self->{-watch_sync}); + $self->{-watch_sync}->{-regen_fmt} = $fmt; + + # call event_step => done unless commit_timer is armed + PublicInbox::DS::requeue($self); +} + +sub on_inbox_unlock { # called by PublicInbox::InboxIdle + my ($self, $ibx) = @_; + my $opt = $self->{-watch_sync}->{-opt}; + my $pr = $opt->{-progress}; + my $ekey = $ibx->eidx_key; + local $0 = "sync $ekey"; + $pr->("indexing $ekey\n") if $pr; + $self->idx_init($opt); + sync_inbox($self, $self->{-watch_sync}, $ibx); + $self->{-commit_timer} //= add_timer($opt->{'commit-interval'} // 10, + \&_watch_commit, $self); +} + +sub eidx_reload { # -extindex --watch SIGHUP handler + my ($self, $idler) = @_; + if ($self->{cfg}) { + my $pr = $self->{-watch_sync}->{-opt}->{-progress}; + $pr->('reloading ...') if $pr; + delete $self->{-resync_queue}; + delete $self->{-ibx_ary_known}; + delete $self->{-ibx_ary_active}; + $self->{ibx_known} = []; + $self->{ibx_active} = []; + %{$self->{ibx_map}} = (); + delete $self->{-watch_sync}->{id2pos}; + my $cfg = PublicInbox::Config->new; + attach_config($self, $cfg); + $idler->refresh($cfg); + $pr->(" done\n") if $pr; + } else { + warn "reload not supported without --all\n"; + } +} + +sub eidx_resync_start ($) { # -extindex --watch SIGUSR1 handler + my ($self) = @_; + $self->{-resync_queue} //= [ @{ibx_sorted($self, 'active')} ]; + PublicInbox::DS::requeue($self); # trigger our ->event_step +} + +sub event_step { # PublicInbox::DS::requeue callback + my ($self) = @_; + if (my $resync_queue = $self->{-resync_queue}) { + if (my $ibx = shift(@$resync_queue)) { + on_inbox_unlock($self, $ibx); + PublicInbox::DS::requeue($self); + } else { + delete $self->{-resync_queue}; + _watch_commit($self); + } + } else { + done($self) unless $self->{-commit_timer}; + } +} + +sub eidx_watch { # public-inbox-extindex --watch main loop + my ($self, $opt) = @_; + local @SIG{keys %SIG} = values %SIG; + for my $sig (qw(HUP USR1 TSTP QUIT INT TERM)) { + $SIG{$sig} = sub { warn "SIG$sig ignored while scanning\n" }; + } + require PublicInbox::InboxIdle; + require PublicInbox::DS; + require PublicInbox::Syscall; + require PublicInbox::Sigfd; + my $idler = PublicInbox::InboxIdle->new($self->{cfg}); + if (!$self->{cfg}) { + $idler->watch_inbox($_) for (@{ibx_sorted($self, 'active')}); + } + for my $ibx (@{ibx_sorted($self, 'active')}) { + $ibx->subscribe_unlock(__PACKAGE__, $self) + } + my $pr = $opt->{-progress}; + $pr->("performing initial scan ...\n") if $pr; + my $sync = eidx_sync($self, $opt); # initial sync + return if $sync->{quit}; + my $oldset = PublicInbox::DS::block_signals(); + local $self->{current_info} = ''; + my $cb = $SIG{__WARN__} || \&CORE::warn; + local $SIG{__WARN__} = sub { + return if PublicInbox::Eml::warn_ignore(@_); + $cb->($self->{current_info}, ': ', @_); + }; + my $sig = { + HUP => sub { eidx_reload($self, $idler) }, + USR1 => sub { eidx_resync_start($self) }, + TSTP => sub { kill('STOP', $$) }, + }; + my $quit = PublicInbox::SearchIdx::quit_cb($sync); + $sig->{QUIT} = $sig->{INT} = $sig->{TERM} = $quit; + my $sigfd = PublicInbox::Sigfd->new($sig, + $PublicInbox::Syscall::SFD_NONBLOCK); + @SIG{keys %$sig} = values(%$sig) if !$sigfd; + local $self->{-watch_sync} = $sync; # for ->on_inbox_unlock + if (!$sigfd) { + # wake up every second to accept signals if we don't + # have signalfd or IO::KQueue: + PublicInbox::DS::sig_setmask($oldset); + PublicInbox::DS->SetLoopTimeout(1000); + } + PublicInbox::DS->SetPostLoopCallback(sub { !$sync->{quit} }); + $pr->("initial scan complete, entering event loop\n") if $pr; + PublicInbox::DS->EventLoop; # calls InboxIdle->event_step + done($self); } no warnings 'once'; @@ -950,5 +1332,6 @@ no warnings 'once'; *atfork_child = \&PublicInbox::V2Writable::atfork_child; *idx_shard = \&PublicInbox::V2Writable::idx_shard; *reindex_checkpoint = \&PublicInbox::V2Writable::reindex_checkpoint; +*checkpoint = \&PublicInbox::V2Writable::checkpoint; 1;