X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FSearchIdx.pm;h=36f97b36d7bb8170237884bd27742cc6afed0fd1;hb=0321a1a9e7ae9c9d878d547ee67659ef8aa95689;hp=a70e1ebf649ffbe48b1d993ea0ce47d34b8ba1f7;hpb=ae68bf5da734189549bbac3a525845a58e45d77f;p=public-inbox.git diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm index a70e1ebf..36f97b36 100644 --- a/lib/PublicInbox/SearchIdx.pm +++ b/lib/PublicInbox/SearchIdx.pm @@ -9,9 +9,9 @@ package PublicInbox::SearchIdx; use strict; use warnings; -use Fcntl qw(:flock :DEFAULT); +use base qw(PublicInbox::Search PublicInbox::Lock); use PublicInbox::MIME; -use base qw(PublicInbox::Search); +use PublicInbox::InboxWritable; use PublicInbox::MID qw/mid_clean id_compress mid_mime mids references/; use PublicInbox::MsgIter; use Carp qw(croak); @@ -19,11 +19,6 @@ use POSIX qw(strftime); require PublicInbox::Git; use constant { - PERM_UMASK => 0, - OLD_PERM_GROUP => 1, - OLD_PERM_EVERYBODY => 2, - PERM_GROUP => 0660, - PERM_EVERYBODY => 0664, BATCH_BYTES => 1_000_000, DEBUG => !!$ENV{DEBUG}, }; @@ -63,20 +58,19 @@ sub new { PublicInbox::AltId->new($ibx, $_); } @$altid ]; } - $git = $ibx->git; - } else { - $git = PublicInbox::Git->new($git_dir); # v1 only + } else { # v1 + $ibx = { mainrepo => $git_dir, version => 1 }; } + $ibx = PublicInbox::InboxWritable->new($ibx); require Search::Xapian::WritableDatabase; my $self = bless { mainrepo => $mainrepo, - git => $git, + -inbox => $ibx, + git => $ibx->git, -altid => $altid, version => $version, }, $class; - my $perm = $self->_git_config_perm; - my $umask = _umask_for($perm); - $self->{umask} = $umask; + $ibx->umask_prepare; if ($version == 1) { $self->{lock_path} = "$mainrepo/ssoma.lock"; } elsif ($version == 2) { @@ -96,7 +90,7 @@ sub _xdb_release { my ($self) = @_; my $xdb = delete $self->{xdb} or croak 'not acquired'; $xdb->close; - _lock_release($self) if $self->{creat}; + $self->lock_release if $self->{creat}; undef; } @@ -107,33 +101,13 @@ sub _xdb_acquire { my $flag = Search::Xapian::DB_OPEN; if ($self->{creat}) { require File::Path; - _lock_acquire($self); + $self->lock_acquire; File::Path::mkpath($dir); $flag = Search::Xapian::DB_CREATE_OR_OPEN; } $self->{xdb} = Search::Xapian::WritableDatabase->new($dir, $flag); } -# we only acquire the flock if creating or reindexing; -# PublicInbox::Import already has the lock on its own. -sub _lock_acquire { - my ($self) = @_; - croak 'already locked' if $self->{lockfh}; - my $lock_path = $self->{lock_path} or return; - sysopen(my $lockfh, $lock_path, O_WRONLY|O_CREAT) or - die "failed to open lock $lock_path: $!\n"; - flock($lockfh, LOCK_EX) or die "lock failed: $!\n"; - $self->{lockfh} = $lockfh; -} - -sub _lock_release { - my ($self) = @_; - return unless $self->{lock_path}; - my $lockfh = delete $self->{lockfh} or croak 'not locked'; - flock($lockfh, LOCK_UN) or die "unlock failed: $!\n"; - close $lockfh or die "close failed: $!\n"; -} - sub add_val ($$$) { my ($doc, $col, $num) = @_; $num = Search::Xapian::sortable_serialise($num); @@ -155,7 +129,9 @@ sub add_values ($$) { my $lines = $values->[PublicInbox::Search::LINES]; add_val($doc, PublicInbox::Search::LINES, $lines); - my $yyyymmdd = strftime('%Y%m%d', gmtime($ts)); + my $ds = $values->[PublicInbox::Search::DS]; + add_val($doc, PublicInbox::Search::DS, $ds); + my $yyyymmdd = strftime('%Y%m%d', gmtime($ds)); add_val($doc, PublicInbox::Search::YYYYMMDD, $yyyymmdd); } @@ -302,7 +278,8 @@ sub index_body ($$$) { } sub add_message { - my ($self, $mime, $bytes, $num, $blob) = @_; # mime = Email::MIME object + # mime = Email::MIME object + my ($self, $mime, $bytes, $num, $oid, $mid0) = @_; my $doc_id; my $mids = mids($mime->header_obj); my $skel = $self->{skeleton}; @@ -318,7 +295,7 @@ sub add_message { } my $lines = $mime->body_raw =~ tr!\n!\n!; - my @values = ($smsg->ts, $num, $bytes, $lines); + my @values = ($smsg->ds, $num, $bytes, $lines, $smsg->ts); add_values($doc, \@values); my $tg = $self->term_generator; @@ -370,7 +347,8 @@ sub add_message { # populates smsg->references for smsg->to_doc_data my $refs = parse_references($smsg); - my $data = $smsg->to_doc_data($blob); + $mid0 = $mids->[0] unless defined $mid0; # v1 compatibility + my $data = $smsg->to_doc_data($oid, $mid0); foreach my $mid (@$mids) { $tg->index_text($mid, 1, 'XM'); } @@ -386,10 +364,12 @@ sub add_message { } } + $self->delete_article($num) if defined $num; # for reindexing if ($skel) { push @values, $mids, $xpath, $data; $skel->index_skeleton(\@values); $doc->add_boolean_term('Q' . $_) foreach @$mids; + $doc->add_boolean_term('XNUM' . $num) if defined $num; $doc_id = $self->{xdb}->add_document($doc); } else { $doc_id = link_and_save($self, $doc, $mids, $refs, @@ -404,29 +384,81 @@ sub add_message { $doc_id; } -# returns deleted doc_id on success, undef on missing +# returns begin and end PostingIterator +sub find_doc_ids { + my ($self, $termval) = @_; + my $db = $self->{xdb}; + + ($db->postlist_begin($termval), $db->postlist_end($termval)); +} + +sub batch_do { + my ($self, $termval, $cb) = @_; + my $batch_size = 1000; # don't let @ids grow too large to avoid OOM + while (1) { + my ($head, $tail) = $self->find_doc_ids($termval); + return if $head == $tail; + my @ids; + for (; $head != $tail && @ids < $batch_size; $head->inc) { + push @ids, $head->get_docid; + } + $cb->(\@ids); + } +} + sub remove_message { my ($self, $mid) = @_; my $db = $self->{xdb}; - my $doc_id; + my $called; $mid = mid_clean($mid); eval { - my ($head, $tail) = $self->find_doc_ids('Q' . $mid); - if ($head->equal($tail)) { - warn "cannot remove non-existent <$mid>\n"; - } - for (; $head != $tail; $head->inc) { - my $docid = $head->get_docid; - $db->delete_document($docid); - } + batch_do($self, 'Q' . $mid, sub { + my ($ids) = @_; + $db->delete_document($_) for @$ids; + $called = 1; + }); }; - if ($@) { warn "failed to remove message <$mid>: $@\n"; - return undef; + } elsif (!$called) { + warn "cannot remove non-existent <$mid>\n"; } - $doc_id; +} + +sub delete_article { + my ($self, $num) = @_; + my $ndel = 0; + batch_do($self, 'XNUM' . $num, sub { + my ($ids) = @_; + $ndel += scalar @$ids; + $self->{xdb}->delete_document($_) for @$ids; + }); +} + +# MID is a hint in V2 +sub remove_by_oid { + my ($self, $oid, $mid) = @_; + my $db = $self->{xdb}; + + # XXX careful, we cannot use batch_do here since we conditionally + # delete documents based on other factors, so we cannot call + # find_doc_ids twice. + my ($head, $tail) = $self->find_doc_ids('Q' . $mid); + return if $head == $tail; + + # there is only ONE element in @delete unless we + # have bugs in our v2writable deduplication check + my @delete; + for (; $head != $tail; $head->inc) { + my $docid = $head->get_docid; + my $doc = $db->get_document($docid); + my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid); + $smsg->load_expand; + push(@delete, $docid) if $smsg->{blob} eq $oid; + } + $db->delete_document($_) foreach @delete; + scalar(@delete); } sub term_generator { # write-only @@ -507,8 +539,9 @@ sub link_and_save { $doc->add_boolean_term('XPATH' . $xpath) if defined $xpath; $doc->add_boolean_term('Q' . $_) foreach @$mids; - my $vivified = 0; + $self->{skel} and die "Should not have read-only skel here\n";; foreach my $mid (@$mids) { + my $vivified = 0; $self->each_smsg_by_mid($mid, sub { my ($cur) = @_; my $type = $cur->type; @@ -530,6 +563,8 @@ sub link_and_save { ++$vivified; 1; }); + $vivified > 1 and warn + "BUG: vivified multiple ($vivified) ghosts for $mid\n"; } # not really important, but we return any vivified ghost docid, here: return $doc_id if defined $doc_id; @@ -606,7 +641,7 @@ sub do_cat_mail { sub index_sync { my ($self, $opts) = @_; - with_umask($self, sub { $self->_index_sync($opts) }); + $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) }) } sub batch_adjust ($$$$) { @@ -793,103 +828,79 @@ sub merge_threads { my ($self, $winner_tid, $loser_tid) = @_; return if $winner_tid == $loser_tid; my $db = $self->{xdb}; - - my $batch_size = 1000; # don't let @ids grow too large to avoid OOM - while (1) { - my ($head, $tail) = $self->find_doc_ids('G' . $loser_tid); - return if $head == $tail; - my @ids; - for (; $head != $tail && @ids < $batch_size; $head->inc) { - push @ids, $head->get_docid; - } - foreach my $docid (@ids) { + batch_do($self, 'G' . $loser_tid, sub { + my ($ids) = @_; + foreach my $docid (@$ids) { my $doc = $db->get_document($docid); $doc->remove_term('G' . $loser_tid); $doc->add_boolean_term('G' . $winner_tid); $db->replace_document($docid, $doc); } - } + }); } -sub _read_git_config_perm { - my ($self) = @_; - my @cmd = qw(config); - if ($self->{version} == 2) { - push @cmd, "--file=$self->{mainrepo}/inbox-config"; - } - my $fh = $self->{git}->popen(@cmd, 'core.sharedRepository'); - local $/ = "\n"; - my $perm = <$fh>; - chomp $perm if defined $perm; - $perm; -} - -sub _git_config_perm { - my $self = shift; - my $perm = scalar @_ ? $_[0] : _read_git_config_perm($self); - return PERM_GROUP if (!defined($perm) || $perm eq ''); - return PERM_UMASK if ($perm eq 'umask'); - return PERM_GROUP if ($perm eq 'group'); - if ($perm =~ /\A(?:all|world|everybody)\z/) { - return PERM_EVERYBODY; - } - return PERM_GROUP if ($perm =~ /\A(?:true|yes|on|1)\z/); - return PERM_UMASK if ($perm =~ /\A(?:false|no|off|0)\z/); - - my $i = oct($perm); - return PERM_UMASK if ($i == PERM_UMASK); - return PERM_GROUP if ($i == OLD_PERM_GROUP); - return PERM_EVERYBODY if ($i == OLD_PERM_EVERYBODY); +sub DESTROY { + # order matters for unlocking + $_[0]->{xdb} = undef; + $_[0]->{lockfh} = undef; +} - if (($i & 0600) != 0600) { - die "core.sharedRepository mode invalid: ". - sprintf('%.3o', $i) . "\nOwner must have permissions\n"; +# remote_* subs are only used by SearchIdxPart and SearchIdxSkeleton +sub remote_commit { + my ($self) = @_; + if (my $w = $self->{w}) { + print $w "commit\n" or die "failed to write commit: $!"; + } else { + $self->commit_txn_lazy; + if (my $skel = $self->{skeleton}) { + $skel->commit_txn_lazy; + } } - ($i & 0666); } -sub _umask_for { - my ($perm) = @_; # _git_config_perm return value - my $rv = $perm; - return umask if $rv == 0; - - # set +x bit if +r or +w were set - $rv |= 0100 if ($rv & 0600); - $rv |= 0010 if ($rv & 0060); - $rv |= 0001 if ($rv & 0006); - (~$rv & 0777); +sub remote_close { + my ($self) = @_; + if (my $w = delete $self->{w}) { + my $pid = delete $self->{pid} or die "no process to wait on\n"; + print $w "close\n" or die "failed to write to pid:$pid: $!\n"; + close $w or die "failed to close pipe for pid:$pid: $!\n"; + waitpid($pid, 0) == $pid or die "remote process did not finish"; + $? == 0 or die ref($self)." pid:$pid exited with: $?"; + } else { + die "transaction in progress $self\n" if $self->{txn}; + $self->_xdb_release if $self->{xdb}; + } } -sub with_umask { - my ($self, $cb) = @_; - my $old = umask $self->{umask}; - my $rv = eval { $cb->() }; - my $err = $@; - umask $old; - die $err if $err; - $rv; +sub remote_remove { + my ($self, $oid, $mid) = @_; + if (my $w = $self->{w}) { + # triggers remove_by_oid in partition or skeleton + print $w "D $oid $mid\n" or die "failed to write remove $!"; + } else { + $self->begin_txn_lazy; + $self->remove_by_oid($oid, $mid); + } } -sub DESTROY { - # order matters for unlocking - $_[0]->{xdb} = undef; - $_[0]->{lockfh} = undef; +sub begin_txn_lazy { + my ($self) = @_; + return if $self->{txn}; + my $xdb = $self->{xdb} || $self->_xdb_acquire; + $xdb->begin_transaction; + $self->{txn} = 1; } -# remote_* subs are only used by SearchIdxPart and SearchIdxSkeleton -sub remote_commit { +sub commit_txn_lazy { my ($self) = @_; - print { $self->{w} } "commit\n" or die "failed to write commit: $!"; + delete $self->{txn} or return; + $self->{xdb}->commit_transaction; } -sub remote_close { +sub worker_done { my ($self) = @_; - my $pid = delete $self->{pid} or die "no process to wait on\n"; - my $w = delete $self->{w} or die "no pipe to write to\n"; - print $w "close\n" or die "failed to write to pid:$pid: $!\n"; - close $w or die "failed to close pipe for pid:$pid: $!\n"; - waitpid($pid, 0) == $pid or die "remote process did not finish"; - $? == 0 or die ref($self)." pid:$pid exited with: $?"; + die "$$ $0 xdb not released\n" if $self->{xdb}; + die "$$ $0 still in transaction\n" if $self->{txn}; } 1;