X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FSearchIdx.pm;h=bdb84fc73dacf7bbfb1d463d386cdf466f1749ce;hb=840785917bc74c8e7df226463144185294047d75;hp=772f5a644a5b0f580e531c78d3c4971d6915d2bf;hpb=64b557420689476493d752968d99ab8ae62bad9a;p=public-inbox.git diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm index 772f5a64..bdb84fc7 100644 --- a/lib/PublicInbox/SearchIdx.pm +++ b/lib/PublicInbox/SearchIdx.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2015-2021 all contributors +# Copyright (C) all contributors # License: AGPL-3.0+ # based on notmuch, but with no concept of folders, files # @@ -11,29 +11,37 @@ use strict; use v5.10.1; use parent qw(PublicInbox::Search PublicInbox::Lock Exporter); use PublicInbox::Eml; +use PublicInbox::Search qw(xap_terms); use PublicInbox::InboxWritable; use PublicInbox::MID qw(mids_for_index mids); use PublicInbox::MsgIter; use PublicInbox::IdxStack; use Carp qw(croak carp); use POSIX qw(strftime); +use Fcntl qw(SEEK_SET); use Time::Local qw(timegm); use PublicInbox::OverIdx; -use PublicInbox::Spawn qw(spawn nodatacow_dir); +use PublicInbox::Spawn qw(spawn); use PublicInbox::Git qw(git_unquote); use PublicInbox::MsgTime qw(msg_timestamp msg_datestamp); use PublicInbox::Address; +use Config; our @EXPORT_OK = qw(log2stack is_ancestor check_size prepare_stack index_text term_generator add_val is_bad_blob); my $X = \%PublicInbox::Search::X; our ($DB_CREATE_OR_OPEN, $DB_OPEN); our $DB_NO_SYNC = 0; -our $BATCH_BYTES = $ENV{XAPIAN_FLUSH_THRESHOLD} ? 0x7fffffff : 1_000_000; +our $DB_DANGEROUS = 0; +our $BATCH_BYTES = $ENV{XAPIAN_FLUSH_THRESHOLD} ? 0x7fffffff : + # assume a typical 64-bit system has 8x more RAM than a + # typical 32-bit system: + (($Config{ptrsize} >= 8 ? 8192 : 1024) * 1024); use constant DEBUG => !!$ENV{DEBUG}; - +my $BASE85 = qr/\A[a-zA-Z0-9\!\#\$\%\&\(\)\*\+\-;<=>\?\@\^_`\{\|\}\~]+\z/; my $xapianlevels = qr/\A(?:full|medium)\z/; my $hex = '[a-f0-9]'; my $OID = $hex .'{40,}'; +my @VMD_MAP = (kw => 'K', L => 'L'); our $INDEXLEVELS = qr/\A(?:full|medium|basic)\z/; sub new { @@ -87,7 +95,7 @@ sub need_xapian ($) { $_[0]->{indexlevel} =~ $xapianlevels } sub idx_release { my ($self, $wake) = @_; if (need_xapian($self)) { - my $xdb = delete $self->{xdb} or croak 'not acquired'; + my $xdb = delete $self->{xdb} or croak '{xdb} not acquired'; $xdb->close; } $self->lock_release($wake) if $self->{creat}; @@ -96,7 +104,7 @@ sub idx_release { sub load_xapian_writable () { return 1 if $X->{WritableDatabase}; - PublicInbox::Search::load_xapian() or return; + PublicInbox::Search::load_xapian() or die "failed to load Xapian: $@\n"; my $xap = $PublicInbox::Search::Xap; for (qw(Document TermGenerator WritableDatabase)) { $X->{$_} = $xap.'::'.$_; @@ -108,7 +116,10 @@ sub load_xapian_writable () { my $ver = (eval($xap.'::major_version()') << 16) | (eval($xap.'::minor_version()') << 8) | eval($xap.'::revision()'); - $DB_NO_SYNC = 0x4 if $ver >= 0x10400; + if ($ver >= 0x10400) { + $DB_NO_SYNC = 0x4; + $DB_DANGEROUS = 0x10; + } # Xapian v1.2.21..v1.2.24 were missing close-on-exec on OFD locks $X->{CLOEXEC_UNSET} = 1 if $ver >= 0x010215 && $ver <= 0x010218; 1; @@ -132,8 +143,12 @@ sub idx_acquire { if (!-d $dir && (!$is_shard || ($is_shard && need_xapian($self)))) { File::Path::mkpath($dir); - nodatacow_dir($dir); + require PublicInbox::Syscall; + PublicInbox::Syscall::nodatacow_dir($dir); $self->{-set_has_threadid_once} = 1; + if (($self->{ibx} // $self->{eidx})->{-dangerous}) { + $flag |= $DB_DANGEROUS; + } } } return unless defined $flag; @@ -221,8 +236,8 @@ sub index_old_diff_fn { # no renames or space support for traditional diffs, # find the number of leading common paths to strip: - my @fa = split('/', $fa); - my @fb = split('/', $fb); + my @fa = split(m'/', $fa); + my @fb = split(m'/', $fb); while (scalar(@fa) && scalar(@fb)) { $fa = join('/', @fa); $fb = join('/', @fb); @@ -242,37 +257,59 @@ sub index_diff ($$$) { my ($self, $txt, $doc) = @_; my %seen; my $in_diff; - my @xnq; - my $xnq = \@xnq; - foreach (split(/\n/, $txt)) { - if ($in_diff && s/^ //) { # diff context + my $xnq = []; + my @l = split(/\n/, $$txt); + undef $$txt; + while (defined($_ = shift @l)) { + if ($in_diff && /^GIT binary patch/) { + push @$xnq, $_; + while (@l && $l[0] =~ /^(?:literal|delta) /) { + # TODO allow searching by size range? + # allows searching by exact size via: + # "literal $SIZE" or "delta $SIZE" + push @$xnq, shift(@l); + + # skip base85 and empty lines + while (@l && ($l[0] =~ /$BASE85/o || + $l[0] !~ /\S/)) { + shift @l; + } + # loop hits trailing "literal 0\nHcmV?d00001\n" + } + } elsif ($in_diff && s/^ //) { # diff context index_diff_inc($self, $_, 'XDFCTX', $xnq); } elsif (/^-- $/) { # email signature begins $in_diff = undef; - } elsif (m!^diff --git "?[^/]+/.+ "?[^/]+/.+\z!) { - # wait until "---" and "+++" to capture filenames + } elsif (m!^diff --git ("?[^/]+/.+) ("?[^/]+/.+)\z!) { + # capture filenames here for binary diffs: + my ($fa, $fb) = ($1, $2); + push @$xnq, $_; $in_diff = 1; + $fa = (split(m'/', git_unquote($fa), 2))[1]; + $fb = (split(m'/', git_unquote($fb), 2))[1]; + $seen{$fa}++ or index_diff_inc($self, $fa, 'XDFN', $xnq); + $seen{$fb}++ or index_diff_inc($self, $fb, 'XDFN', $xnq); # traditional diff: } elsif (m/^diff -(.+) (\S+) (\S+)$/) { my ($opt, $fa, $fb) = ($1, $2, $3); - push @xnq, $_; + push @$xnq, $_; # only support unified: next unless $opt =~ /[uU]/; $in_diff = index_old_diff_fn($self, \%seen, $fa, $fb, $xnq); } elsif (m!^--- ("?[^/]+/.+)!) { my $fn = $1; - $fn = (split('/', git_unquote($fn), 2))[1]; + $fn = (split(m'/', git_unquote($fn), 2))[1]; $seen{$fn}++ or index_diff_inc($self, $fn, 'XDFN', $xnq); $in_diff = 1; } elsif (m!^\+\+\+ ("?[^/]+/.+)!) { my $fn = $1; - $fn = (split('/', git_unquote($fn), 2))[1]; + $fn = (split(m'/', git_unquote($fn), 2))[1]; $seen{$fn}++ or index_diff_inc($self, $fn, 'XDFN', $xnq); $in_diff = 1; } elsif (/^--- (\S+)/) { - $in_diff = $1; - push @xnq, $_; + $in_diff = $1; # old diff filename + push @$xnq, $_; } elsif (defined $in_diff && /^\+\+\+ (\S+)/) { $in_diff = index_old_diff_fn($self, \%seen, $in_diff, $1, $xnq); @@ -298,19 +335,19 @@ sub index_diff ($$$) { /^(?:dis)?similarity index / || /^\\ No newline at end of file/ || /^Binary files .* differ/) { - push @xnq, $_; + push @$xnq, $_; } elsif ($_ eq '') { # possible to be in diff context, some mail may be # stripped by MUA or even GNU diff(1). "git apply" # treats a bare "\n" as diff context, too } else { - push @xnq, $_; + push @$xnq, $_; warn "non-diff line: $_\n" if DEBUG && $_ ne ''; $in_diff = undef; } } - index_text($self, join("\n", @xnq), 1, 'XNQ'); + index_text($self, join("\n", @$xnq), 1, 'XNQ'); } sub index_xapian { # msg_iter callback @@ -333,6 +370,20 @@ sub index_xapian { # msg_iter callback defined $s or return; $_[0]->[0] = $part = undef; # free memory + if ($s =~ /^(?:diff|---|\+\+\+) /ms) { + open(my $fh, '+>:utf8', undef) or die "open: $!"; + open(my $eh, '+>', undef) or die "open: $!"; + $fh->autoflush(1); + print $fh $s or die "print: $!"; + sysseek($fh, 0, SEEK_SET) or die "sysseek: $!"; + my $id = ($self->{ibx} // $self->{eidx})->git->qx( + [qw(patch-id --stable)], + {}, { 0 => $fh, 2 => $eh }); + $id =~ /\A([a-f0-9]{40,})/ and $doc->add_term('XDFID'.$1); + seek($eh, 0, SEEK_SET) or die "seek: $!"; + while (<$eh>) { warn $_ } + } + # split off quoted and unquoted blocks: my @sections = PublicInbox::MsgIter::split_quotes($s); undef $s; # free memory @@ -342,7 +393,7 @@ sub index_xapian { # msg_iter callback } else { # does it look like a diff? if ($txt =~ /^(?:diff|---|\+\+\+) /ms) { - index_diff($self, $txt, $doc); + index_diff($self, \$txt, $doc); } else { index_text($self, $txt, 1, 'XNQ'); } @@ -428,17 +479,25 @@ sub eml2doc ($$$;$) { sub add_xapian ($$$$) { my ($self, $eml, $smsg, $mids) = @_; begin_txn_lazy($self); + my $merge_vmd = delete $smsg->{-merge_vmd}; my $doc = eml2doc($self, $eml, $smsg, $mids); + if (my $old = $merge_vmd ? _get_doc($self, $smsg->{num}) : undef) { + my @x = @VMD_MAP; + while (my ($field, $pfx) = splice(@x, 0, 2)) { + for my $term (xap_terms($pfx, $old)) { + $doc->add_boolean_term($pfx.$term); + } + } + } $self->{xdb}->replace_document($smsg->{num}, $doc); } sub _msgmap_init ($) { my ($self) = @_; die "BUG: _msgmap_init is only for v1\n" if $self->{ibx}->version != 1; - $self->{mm} //= eval { + $self->{mm} //= do { require PublicInbox::Msgmap; - my $rw = $self->{ibx}->{-no_fsync} ? 2 : 1; - PublicInbox::Msgmap->new($self->{ibx}->{inboxdir}, $rw); + PublicInbox::Msgmap->new_file($self->{ibx}, 1); }; } @@ -494,11 +553,20 @@ sub add_eidx_info { begin_txn_lazy($self); my $doc = _get_doc($self, $docid) or return; term_generator($self)->set_document($doc); + + # '.' is special for lei_store $doc->add_boolean_term('O'.$eidx_key) if $eidx_key ne '.'; + index_list_id($self, $doc, $eml); $self->{xdb}->replace_document($docid, $doc); } +sub get_terms { + my ($self, $pfx, $docid) = @_; + begin_txn_lazy($self); + xap_terms($pfx, $self->{xdb}, $docid); +} + sub remove_eidx_info { my ($self, $docid, $eidx_key, $eml) = @_; begin_txn_lazy($self); @@ -528,57 +596,114 @@ sub remove_eidx_info { $self->{xdb}->replace_document($docid, $doc); } -sub set_keywords { - my ($self, $docid, @kw) = @_; +sub set_vmd { + my ($self, $docid, $vmd) = @_; begin_txn_lazy($self); my $doc = _get_doc($self, $docid) or return; - my %keep = map { $_ => 1 } @kw; - my %add = %keep; - my @rm; - my $end = $doc->termlist_end; - for (my $cur = $doc->termlist_begin; $cur != $end; $cur++) { - $cur->skip_to('K'); - last if $cur == $end; - my $kw = $cur->get_termname; - $kw =~ s/\AK//s or next; - $keep{$kw} ? delete($add{$kw}) : push(@rm, $kw); - } - return unless (scalar(@rm) + scalar(keys %add)); - $doc->remove_term('K'.$_) for @rm; - $doc->add_boolean_term('K'.$_) for (keys %add); + my ($end, @rm, @add); + my @x = @VMD_MAP; + while (my ($field, $pfx) = splice(@x, 0, 2)) { + my $set = $vmd->{$field} // next; + my %keep = map { $_ => 1 } @$set; + my %add = %keep; + $end //= $doc->termlist_end; + for (my $cur = $doc->termlist_begin; $cur != $end; $cur++) { + $cur->skip_to($pfx); + last if $cur == $end; + my $v = $cur->get_termname; + $v =~ s/\A$pfx//s or next; + $keep{$v} ? delete($add{$v}) : push(@rm, $pfx.$v); + } + push(@add, map { $pfx.$_ } keys %add); + } + return unless scalar(@rm) || scalar(@add); + $doc->remove_term($_) for @rm; + $doc->add_boolean_term($_) for @add; $self->{xdb}->replace_document($docid, $doc); } -sub add_keywords { - my ($self, $docid, @kw) = @_; +sub apply_vmd_mod ($$) { + my ($doc, $vmd_mod) = @_; + my $updated = 0; + my @x = @VMD_MAP; + while (my ($field, $pfx) = splice(@x, 0, 2)) { + # field: "L" or "kw" + for my $val (@{$vmd_mod->{"-$field"} // []}) { + eval { + $doc->remove_term($pfx . $val); + ++$updated; + }; + } + for my $val (@{$vmd_mod->{"+$field"} // []}) { + $doc->add_boolean_term($pfx . $val); + ++$updated; + } + } + $updated; +} + +sub add_vmd { + my ($self, $docid, $vmd) = @_; begin_txn_lazy($self); my $doc = _get_doc($self, $docid) or return; - $doc->add_boolean_term('K'.$_) for @kw; - $self->{xdb}->replace_document($docid, $doc); + my @x = @VMD_MAP; + my $updated = 0; + while (my ($field, $pfx) = splice(@x, 0, 2)) { + my $add = $vmd->{$field} // next; + $doc->add_boolean_term($pfx . $_) for @$add; + $updated += scalar(@$add); + } + $updated += apply_vmd_mod($doc, $vmd); + $self->{xdb}->replace_document($docid, $doc) if $updated; } -sub remove_keywords { - my ($self, $docid, @kw) = @_; +sub remove_vmd { + my ($self, $docid, $vmd) = @_; begin_txn_lazy($self); my $doc = _get_doc($self, $docid) or return; my $replace; - eval { - $doc->remove_term('K'.$_); - $replace = 1 - } for @kw; + my @x = @VMD_MAP; + while (my ($field, $pfx) = splice(@x, 0, 2)) { + my $rm = $vmd->{$field} // next; + for (@$rm) { + eval { + $doc->remove_term($pfx . $_); + $replace = 1; + }; + } + } $self->{xdb}->replace_document($docid, $doc) if $replace; } +sub update_vmd { + my ($self, $docid, $vmd_mod) = @_; + begin_txn_lazy($self); + my $doc = _get_doc($self, $docid) or return; + my $updated = apply_vmd_mod($doc, $vmd_mod); + $self->{xdb}->replace_document($docid, $doc) if $updated; + $updated; +} + sub xdb_remove { my ($self, @docids) = @_; - $self->begin_txn_lazy; - my $xdb = $self->{xdb} or return; + begin_txn_lazy($self); + my $xdb = $self->{xdb} // die 'BUG: missing {xdb}'; for my $docid (@docids) { eval { $xdb->delete_document($docid) }; warn "E: #$docid not in in Xapian? $@\n" if $@; } } +sub xdb_remove_quiet { + my ($self, $docid) = @_; + begin_txn_lazy($self); + my $xdb = $self->{xdb} // die 'BUG: missing {xdb}'; + eval { $xdb->delete_document($docid) }; + ++$self->{-quiet_rm} unless $@; +} + +sub nr_quiet_rm { delete($_[0]->{-quiet_rm}) // 0 } + sub index_git_blob_id { my ($doc, $pfx, $objid) = @_; @@ -684,7 +809,8 @@ sub index_sync { my ($self, $opt) = @_; delete $self->{lock_path} if $opt->{-skip_lock}; $self->with_umask(\&_index_sync, $self, $opt); - if ($opt->{reindex} && !$opt->{quit}) { + if ($opt->{reindex} && !$opt->{quit} && + !grep(defined, @$opt{qw(since until)})) { my %again = %$opt; delete @again{qw(rethread reindex)}; index_sync($self, \%again); @@ -709,18 +835,19 @@ sub v1_checkpoint ($$;$) { # $newest may be undef my $newest = $stk ? $stk->{latest_cmt} : ${$sync->{latest_cmt}}; if (defined($newest)) { - my $cur = $self->{mm}->last_commit || ''; - if (need_update($self, $cur, $newest)) { + my $cur = $self->{mm}->last_commit; + if (need_update($self, $sync, $cur, $newest)) { $self->{mm}->last_commit($newest); } } ${$sync->{max}} = $self->{batch_bytes}; $self->{mm}->{dbh}->commit; - my $xdb = need_xapian($self) ? $self->{xdb} : undef; + eval { $self->{mm}->{dbh}->do('PRAGMA optimize') }; + my $xdb = $self->{xdb}; if ($newest && $xdb) { my $cur = $xdb->get_metadata('last_commit'); - if (need_update($self, $cur, $newest)) { + if (need_update($self, $sync, $cur, $newest)) { $xdb->set_metadata('last_commit', $newest); } } @@ -804,23 +931,28 @@ sub log2stack ($$$) { # Count the new files so they can be added newest to oldest # and still have numbers increasing from oldest to newest - my $fh = $git->popen(qw(log --raw -r --pretty=tformat:%at-%ct-%H - --no-notes --no-color --no-renames --no-abbrev), - $range); - my ($at, $ct, $stk, $cmt); - while (<$fh>) { + my @cmd = qw(log --raw -r --pretty=tformat:%at-%ct-%H + --no-notes --no-color --no-renames --no-abbrev); + for my $k (qw(since until)) { + my $v = $sync->{-opt}->{$k} // next; + next if !$sync->{-opt}->{reindex}; + push @cmd, "--$k=$v"; + } + my $fh = $git->popen(@cmd, $range); + my ($at, $ct, $stk, $cmt, $l); + while (defined($l = <$fh>)) { return if $sync->{quit}; - if (/\A([0-9]+)-([0-9]+)-($OID)$/o) { + if ($l =~ /\A([0-9]+)-([0-9]+)-($OID)$/o) { ($at, $ct, $cmt) = ($1 + 0, $2 + 0, $3); $stk //= PublicInbox::IdxStack->new($cmt); - } elsif (/$del/) { + } elsif ($l =~ /$del/) { my $oid = $1; if ($D) { # reindex case $D->{pack('H*', $oid)}++; } else { # non-reindex case: $stk->push_rec('d', $at, $ct, $oid, $cmt); } - } elsif (/$add/) { + } elsif ($l =~ /$add/) { my $oid = $1; if ($D) { my $oid_bin = pack('H*', $oid); @@ -862,10 +994,17 @@ sub is_ancestor ($$$) { $? == 0; } -sub need_update ($$$) { - my ($self, $cur, $new) = @_; +sub need_update ($$$$) { + my ($self, $sync, $cur, $new) = @_; my $git = $self->{ibx}->git; - return 1 if $cur && !is_ancestor($git, $cur, $new); + $cur //= ''; # XS Search::Xapian ->get_metadata doesn't give undef + + # don't rewind if --{since,until,before,after} are in use + return if $cur ne '' && + grep(defined, @{$sync->{-opt}}{qw(since until)}) && + is_ancestor($git, $new, $cur); + + return 1 if $cur ne '' && !is_ancestor($git, $cur, $new); my $range = $cur eq '' ? $new : "$cur..$new"; chomp(my $n = $git->qx(qw(rev-list --count), $range)); ($n eq '' || $n > 0); @@ -1019,25 +1158,4 @@ sub eidx_shard_new { $self; } -# ensure there's no stale Xapian docs by treating $over as canonical -sub over_check { - my ($self, $over) = @_; - begin_txn_lazy($self); - my $sth = $over->dbh->prepare(<<''); -SELECT COUNT(*) FROM over WHERE num = ? - - my $xdb = $self->{xdb}; - my $cur = $xdb->postlist_begin(''); - my $end = $xdb->postlist_end(''); - my $xdir = $self->xdir; - for (; $cur != $end; $cur++) { - my $docid = $cur->get_docid; - $sth->execute($docid); - my $x = $sth->fetchrow_array; - next if $x > 0; - warn "I: removing $xdir #$docid, not in `over'\n"; - $xdb->delete_document($docid); - } -} - 1;