diff options
Diffstat (limited to 'lib/PublicInbox/SearchIdx.pm')
| -rw-r--r-- | lib/PublicInbox/SearchIdx.pm | 1239 |
1 files changed, 780 insertions, 459 deletions
diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm index c33a48c3..1cbf6d23 100644 --- a/lib/PublicInbox/SearchIdx.pm +++ b/lib/PublicInbox/SearchIdx.pm @@ -1,6 +1,6 @@ -# Copyright (C) 2015-2020 all contributors <meta@public-inbox.org> +# Copyright (C) all contributors <meta@public-inbox.org> # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt> -# based on notmuch, but with no concept of folders, files or flags +# based on notmuch, but with no concept of folders, files # # Indexes mail with Xapian and our (SQLite-based) ::Msgmap for use # with the web and NNTP interfaces. This index maintains thread @@ -8,31 +8,47 @@ # This writes to the search index. package PublicInbox::SearchIdx; use strict; -use warnings; -use base qw(PublicInbox::Search PublicInbox::Lock); -use PublicInbox::MIME; +use v5.10.1; +use parent qw(PublicInbox::Search PublicInbox::Lock PublicInbox::Umask + Exporter); +use PublicInbox::Eml; +use PublicInbox::Search qw(xap_terms); use PublicInbox::InboxWritable; -use PublicInbox::MID qw/mid_clean mid_mime mids_for_index/; +use PublicInbox::MID qw(mids_for_index mids); use PublicInbox::MsgIter; -use Carp qw(croak); +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); +use PublicInbox::Spawn qw(run_wait popen_rd); 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 $X = \%PublicInbox::Search::X; -my ($DB_CREATE_OR_OPEN, $DB_OPEN); -use constant { - BATCH_BYTES => defined($ENV{XAPIAN_FLUSH_THRESHOLD}) ? - 0x7fffffff : 1_000_000, - DEBUG => !!$ENV{DEBUG}, -}; - +our ($DB_CREATE_OR_OPEN, $DB_OPEN); +our $DB_NO_SYNC = 0; +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-zA-Z0-9\!\#\$\%\&\(\)\*\+\-;<=>\?\@\^_`\{\|\}\~]+/; my $xapianlevels = qr/\A(?:full|medium)\z/; +my $hex = '[a-f0-9]'; +my $OID = $hex .'{40,}'; +my @VMD_MAP = (kw => 'K', L => 'L'); # value order matters +our $INDEXLEVELS = qr/\A(?:full|medium|basic)\z/; +our $PATCHID_BROKEN; sub new { my ($class, $ibx, $creat, $shard) = @_; ref $ibx or die "BUG: expected PublicInbox::Inbox object: $ibx"; - my $levels = qr/\A(?:full|medium|basic)\z/; my $inboxdir = $ibx->{inboxdir}; my $version = $ibx->version; my $indexlevel = 'full'; @@ -42,26 +58,29 @@ sub new { $altid = [ map { PublicInbox::AltId->new($ibx, $_); } @$altid ]; } if ($ibx->{indexlevel}) { - if ($ibx->{indexlevel} =~ $levels) { + if ($ibx->{indexlevel} =~ $INDEXLEVELS) { $indexlevel = $ibx->{indexlevel}; } else { die("Invalid indexlevel $ibx->{indexlevel}\n"); } } + undef $PATCHID_BROKEN; # retry on new instances in case of upgrades $ibx = PublicInbox::InboxWritable->new($ibx); - my $self = bless { - inboxdir => $inboxdir, - -inbox => $ibx, - git => $ibx->git, - -altid => $altid, - ibx_ver => $version, - indexlevel => $indexlevel, - }, $class; - $ibx->umask_prepare; + my $self = PublicInbox::Search->new($ibx); + bless $self, $class; + $self->{ibx} = $ibx; + $self->{-altid} = $altid; + $self->{indexlevel} = $indexlevel; + $self->{-set_indexlevel_once} = 1 if $indexlevel eq 'medium'; + if ($ibx->{-skip_docdata}) { + $self->{-set_skip_docdata_once} = 1; + $self->{-skip_docdata} = 1; + } if ($version == 1) { $self->{lock_path} = "$inboxdir/ssoma.lock"; my $dir = $self->xdir; - $self->{over} = PublicInbox::OverIdx->new("$dir/over.sqlite3"); + $self->{oidx} = PublicInbox::OverIdx->new("$dir/over.sqlite3"); + $self->{oidx}->{-no_fsync} = 1 if $ibx->{-no_fsync}; } elsif ($version == 2) { defined $shard or die "shard is required for v2\n"; # shard is a number @@ -74,21 +93,21 @@ sub new { $self; } -sub need_xapian ($) { $_[0]->{indexlevel} =~ $xapianlevels } +sub need_xapian ($) { ($_[0]->{indexlevel} // 'full') =~ $xapianlevels } -sub _xdb_release { - my ($self) = @_; +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 if $self->{creat}; + $self->lock_release($wake) if $self->{creat}; undef; } 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)) { $X->{$_} = $xap.'::'.$_; @@ -97,10 +116,19 @@ sub load_xapian_writable () { *sortable_serialise = $xap.'::sortable_serialise'; $DB_CREATE_OR_OPEN = eval($xap.'::DB_CREATE_OR_OPEN()'); $DB_OPEN = eval($xap.'::DB_OPEN()'); + my $ver = eval 'v'.join('.', eval($xap.'::major_version()'), + eval($xap.'::minor_version()'), + eval($xap.'::revision()')); + if ($ver ge 1.4) { # new flags in Xapian 1.4 + $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 ge v1.2.21 && $ver le v1.2.24; 1; } -sub _xdb_acquire { +sub idx_acquire { my ($self) = @_; my $flag; my $dir = $self->xdir; @@ -109,21 +137,27 @@ sub _xdb_acquire { load_xapian_writable(); $flag = $self->{creat} ? $DB_CREATE_OR_OPEN : $DB_OPEN; } + my $owner = $self->{ibx} // $self->{eidx} // $self; if ($self->{creat}) { require File::Path; $self->lock_acquire; # don't create empty Xapian directories if we don't need Xapian my $is_shard = defined($self->{shard}); - if (!$is_shard || ($is_shard && need_xapian($self))) { + if (!-d $dir && (!$is_shard || + ($is_shard && need_xapian($self)))) { File::Path::mkpath($dir); + require PublicInbox::Syscall; + PublicInbox::Syscall::nodatacow_dir($dir); + # owner == self for CodeSearchIdx + $self->{-set_has_threadid_once} = 1 if $owner != $self; + $flag |= $DB_DANGEROUS if $owner->{-dangerous}; } } return unless defined $flag; + $flag |= $DB_NO_SYNC if $owner->{-no_fsync}; my $xdb = eval { ($X->{WritableDatabase})->new($dir, $flag) }; - if ($@) { - die "Failed opening $dir: ", $@; - } + croak "Failed opening $dir: $@" if $@; $self->{xdb} = $xdb; } @@ -138,33 +172,52 @@ sub term_generator ($) { # write-only $self->{term_generator} //= do { my $tg = $X->{TermGenerator}->new; - $tg->set_stemmer($self->stemmer); + $tg->set_stemmer(PublicInbox::Search::stemmer($self)); $tg; } } +sub index_phrase ($$$$) { + my ($self, $text, $wdf_inc, $prefix) = @_; + + term_generator($self)->index_text($text, $wdf_inc, $prefix); + $self->{term_generator}->increase_termpos; +} + sub index_text ($$$$) { my ($self, $text, $wdf_inc, $prefix) = @_; - my $tg = term_generator($self); # man Search::Xapian::TermGenerator if ($self->{indexlevel} eq 'full') { - $tg->index_text($text, $wdf_inc, $prefix); - $tg->increase_termpos; + index_phrase($self, $text, $wdf_inc, $prefix); } else { - $tg->index_text_without_positions($text, $wdf_inc, $prefix); + term_generator($self)->index_text_without_positions( + $text, $wdf_inc, $prefix); } } -sub index_users ($$) { +sub index_headers ($$) { my ($self, $smsg) = @_; - - my $from = $smsg->from; - my $to = $smsg->to; - my $cc = $smsg->cc; - - index_text($self, $from, 1, 'A'); # A - author - index_text($self, $to, 1, 'XTO') if $to ne ''; - index_text($self, $cc, 1, 'XCC') if $cc ne ''; + my @x = (from => 'A', to => 'XTO', cc => 'XCC'); # A: Author + while (my ($field, $pfx) = splice(@x, 0, 2)) { + my $val = $smsg->{$field}; + next if $val eq ''; + # include "(comments)" after the address, too, so not using + # PublicInbox::Address::names or pairs + index_text($self, $val, 1, $pfx); + + # we need positional info for email addresses since they + # can be considered phrases + if ($self->{indexlevel} eq 'medium') { + for my $addr (PublicInbox::Address::emails($val)) { + index_phrase($self, $addr, 1, $pfx); + } + } + } + @x = (subject => 'S'); + while (my ($field, $pfx) = splice(@x, 0, 2)) { + my $val = $smsg->{$field}; + index_text($self, $val, 1, $pfx) if $val ne ''; + } } sub index_diff_inc ($$$$) { @@ -173,7 +226,11 @@ sub index_diff_inc ($$$$) { index_text($self, join("\n", @$xnq), 1, 'XNQ'); @$xnq = (); } - index_text($self, $text, 1, $pfx); + if ($pfx eq 'XDFN') { + index_phrase($self, $text, 1, $pfx); + } else { + index_text($self, $text, 1, $pfx); + } } sub index_old_diff_fn { @@ -181,8 +238,8 @@ sub index_old_diff_fn { # no renames or space support for traditional diffs, # find the number of leading common paths to strip: - my @fa = split('/', $fa); - my @fb = split('/', $fb); + my @fa = split(m'/', $fa); + my @fb = split(m'/', $fb); while (scalar(@fa) && scalar(@fb)) { $fa = join('/', @fa); $fb = join('/', @fb); @@ -202,37 +259,59 @@ sub index_diff ($$$) { 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 && ($l[0] =~ /\A$BASE85\h*\z/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); @@ -258,86 +337,150 @@ sub index_diff ($$$) { /^(?:dis)?similarity index / || /^\\ No newline at end of file/ || /^Binary files .* differ/) { - push @xnq, $_; + push @$xnq, $_; } elsif ($_ eq '') { # possible to be in diff context, some mail may be # stripped by MUA or even GNU diff(1). "git apply" # treats a bare "\n" as diff context, too } else { - push @xnq, $_; + push @$xnq, $_; warn "non-diff line: $_\n" if DEBUG && $_ ne ''; $in_diff = undef; } } - index_text($self, join("\n", @xnq), 1, 'XNQ'); + index_text($self, join("\n", @$xnq), 1, 'XNQ'); } -sub index_body ($$$) { - my ($self, $txt, $doc) = @_; - if ($doc) { - # does it look like a diff? - if ($txt =~ /^(?:diff|---|\+\+\+) /ms) { - index_diff($self, $txt, $doc); - } else { - index_text($self, $txt, 1, 'XNQ'); +sub index_body_text { + my ($self, $doc, $sref) = @_; + my $rd; + # start patch-id in parallel + if ($$sref =~ /^(?:diff|---|\+\+\+) /ms && !$PATCHID_BROKEN) { + my $git = ($self->{ibx} // $self->{eidx} // $self)->git; + my $fh = PublicInbox::IO::write_file '+>:utf8', undef, $$sref; + $fh->flush or die "flush: $!"; + sysseek($fh, 0, SEEK_SET); + $rd = popen_rd($git->cmd(qw(patch-id --stable)), undef, + { 0 => $fh }); + } + + # split off quoted and unquoted blocks: + my @sections = PublicInbox::MsgIter::split_quotes($$sref); + undef $$sref; # free memory + for my $txt (@sections) { + if ($txt =~ /\A>/) { + if ($txt =~ /^[>\t ]+GIT binary patch\r?/sm) { + # get rid of Base-85 noise + $txt =~ s/^([>\h]+(?:literal|delta) + \x20[0-9]+\r?\n) + (?:[>\h]+$BASE85\h*\r?\n)+/$1/gsmx; + } + index_text($self, $txt, 0, 'XQUOT'); + } else { # does it look like a diff? + if ($txt =~ /^(?:diff|---|\+\+\+) /ms) { + index_diff($self, \$txt, $doc); + } else { + index_text($self, $txt, 1, 'XNQ'); + } + } + undef $txt; # free memory + } + if (defined $rd) { # reap `git patch-id' + (readline($rd) // '') =~ /\A([a-f0-9]{40,})/ and + $doc->add_term('XDFID'.$1); + if (!$rd->close) { + my $c = 'git patch-id --stable'; + $PATCHID_BROKEN = ($? >> 8) == 129; + $PATCHID_BROKEN ? warn("W: $c requires git v2.1.0+\n") + : warn("W: $c failed: \$?=$? (non-fatal)"); } - } else { - index_text($self, $txt, 0, 'XQUOT'); } } sub index_xapian { # msg_iter callback - my ($part, $depth, @idx) = @{$_[0]}; + my $part = $_[0]->[0]; # ignore $depth and $idx my ($self, $doc) = @{$_[1]}; my $ct = $part->content_type || 'text/plain'; my $fn = $part->filename; if (defined $fn && $fn ne '') { - index_text($self, $fn, 1, 'XFN'); + index_phrase($self, $fn, 1, 'XFN'); + } + if ($part->{is_submsg}) { + my $mids = mids_for_index($part); + index_ids($self, $doc, $part, $mids); + my $smsg = bless {}, 'PublicInbox::Smsg'; + $smsg->populate($part); + index_headers($self, $smsg); } my ($s, undef) = msg_part_text($part, $ct); defined $s or return; + $_[0]->[0] = $part = undef; # free memory + index_body_text($self, $doc, \$s); +} - # split off quoted and unquoted blocks: - my @sections = split(/((?:^>[^\n]*\n)+)/sm, $s); - $part = $s = undef; - index_body($self, $_, /\A>/ ? 0 : $doc) for @sections; +sub index_list_id ($$$) { + my ($self, $doc, $hdr) = @_; + for my $l ($hdr->header_raw('List-Id')) { + $l =~ /<([^>]+)>/ or next; + my $lid = lc $1; + $lid =~ tr/\n\t\r\0//d; # same rules as Message-ID + $doc->add_boolean_term('G' . $lid); + index_phrase($self, $lid, 1, 'XL'); # probabilistic + } +} + +sub index_ids ($$$$) { + my ($self, $doc, $hdr, $mids) = @_; + for my $mid (@$mids) { + index_phrase($self, $mid, 1, 'XM'); + + # because too many Message-IDs are prefixed with + # "Pine.LNX."... + if ($mid =~ /\w{12,}/) { + my @long = ($mid =~ /(\w{3,}+)/g); + index_phrase($self, join(' ', @long), 1, 'XM'); + } + } + $doc->add_boolean_term('Q' . $_) for @$mids; + index_list_id($self, $doc, $hdr); } -sub add_xapian ($$$$$$) { - my ($self, $mime, $num, $oid, $mids, $mid0) = @_; - my $smsg = PublicInbox::SearchMsg->new($mime); +sub eml2doc ($$$;$) { + my ($self, $eml, $smsg, $mids) = @_; + $mids //= mids_for_index($eml); my $doc = $X->{Document}->new; - my $subj = $smsg->subject; - add_val($doc, PublicInbox::Search::TS(), $smsg->ts); - my @ds = gmtime($smsg->ds); + add_val($doc, PublicInbox::Search::TS(), $smsg->{ts}); + my @ds = gmtime($smsg->{ds}); my $yyyymmdd = strftime('%Y%m%d', @ds); add_val($doc, PublicInbox::Search::YYYYMMDD(), $yyyymmdd); my $dt = strftime('%Y%m%d%H%M%S', @ds); add_val($doc, PublicInbox::Search::DT(), $dt); + add_val($doc, PublicInbox::Search::BYTES(), $smsg->{bytes}); + add_val($doc, PublicInbox::Search::UID(), $smsg->{num}); + add_val($doc, PublicInbox::Search::THREADID, $smsg->{tid}); - my $tg = term_generator($self); - - $tg->set_document($doc); - index_text($self, $subj, 1, 'S') if $subj; - index_users($self, $smsg); + term_generator($self)->set_document($doc); + index_headers($self, $smsg); - msg_iter($mime, \&index_xapian, [ $self, $doc ]); - foreach my $mid (@$mids) { - index_text($self, $mid, 1, 'XM'); - - # because too many Message-IDs are prefixed with - # "Pine.LNX."... - if ($mid =~ /\w{12,}/) { - my @long = ($mid =~ /(\w{3,}+)/g); - index_text($self, join(' ', @long), 1, 'XM'); - } + if (defined(my $eidx_key = $smsg->{eidx_key})) { + $doc->add_boolean_term('O'.$eidx_key) if $eidx_key ne '.'; } - $smsg->{to} = $smsg->{cc} = ''; - PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids); - my $data = $smsg->to_doc_data($oid, $mid0); - $doc->set_data($data); + msg_iter($eml, \&index_xapian, [ $self, $doc ]); + index_ids($self, $doc, $eml, $mids); + + # by default, we maintain compatibility with v1.5.0 and earlier + # by writing to docdata.glass, users who never expect to downgrade can + # use --skip-docdata + if (!$self->{-skip_docdata}) { + # WWW doesn't need {to} or {cc}, only NNTP + $smsg->{to} = $smsg->{cc} = ''; + $smsg->parse_references($eml, $mids); + my $data = $smsg->to_doc_data; + $doc->set_data($data); + } + if (my $altid = $self->{-altid}) { foreach my $alt (@$altid) { my $pfx = $alt->{xprefix}; @@ -348,34 +491,61 @@ sub add_xapian ($$$$$$) { } } } - $doc->add_boolean_term('Q' . $_) foreach @$mids; - $self->{xdb}->replace_document($num, $doc); + $doc; +} + +sub add_xapian ($$$$) { + 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_ver} != 1; - $self->{mm} //= eval { + die "BUG: _msgmap_init is only for v1\n" if $self->{ibx}->version != 1; + $self->{mm} //= do { require PublicInbox::Msgmap; - PublicInbox::Msgmap->new($self->{inboxdir}, 1); + PublicInbox::Msgmap->new_file($self->{ibx}, 1); }; } sub add_message { - # mime = Email::MIME object - my ($self, $mime, $bytes, $num, $oid, $mid0) = @_; - my $mids = mids_for_index($mime->header_obj); - $mid0 //= $mids->[0]; # v1 compatibility - $num //= do { # v1 + # mime = PublicInbox::Eml or Email::MIME object + my ($self, $mime, $smsg, $sync) = @_; + begin_txn_lazy($self); + my $mids = mids_for_index($mime); + $smsg //= bless { blob => '' }, 'PublicInbox::Smsg'; # test-only compat + $smsg->{mid} //= $mids->[0]; # v1 compatibility + $smsg->{num} //= do { # v1 _msgmap_init($self); - index_mm($self, $mime); + index_mm($self, $mime, $smsg->{blob}, $sync); }; + + # v1 and tests only: + $smsg->populate($mime, $sync); + $smsg->{bytes} //= length($mime->as_string); + eval { - if (need_xapian($self)) { - add_xapian($self, $mime, $num, $oid, $mids, $mid0); + # order matters, overview stores every possible piece of + # data in doc_data (deflated). Xapian only stores a subset + # of the fields which exist in over.sqlite3. We may stop + # storing doc_data in Xapian sometime after we get multi-inbox + # search working. + if (my $oidx = $self->{oidx}) { # v1 only + $oidx->add_overview($mime, $smsg); } - if (my $over = $self->{over}) { - $over->add_overview($mime, $bytes, $num, $oid, $mid0); + if (need_xapian($self)) { + add_xapian($self, $mime, $smsg, $mids); } }; @@ -383,276 +553,450 @@ sub add_message { warn "failed to index message <".join('> <',@$mids).">: $@\n"; return undef; } - $num; + $smsg->{num}; +} + +sub _get_doc ($$) { + my ($self, $docid) = @_; + $self->get_doc($docid) // do { + warn "E: #$docid missing in Xapian\n"; + undef; + } } -# returns begin and end PostingIterator -sub find_doc_ids { - my ($self, $termval) = @_; - my $db = $self->{xdb}; +sub add_eidx_info { + my ($self, $docid, $eidx_key, $eml) = @_; + begin_txn_lazy($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 $eidx_key ne '.'; - ($db->postlist_begin($termval), $db->postlist_end($termval)); + index_list_id($self, $doc, $eml); + $self->{xdb}->replace_document($docid, $doc); } -# v1 only -sub batch_do { - my ($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++) { - push @ids, $head->get_docid; - } - $cb->(\@ids); - } +sub get_terms { + my ($self, $pfx, $docid) = @_; + begin_txn_lazy($self); + xap_terms($pfx, $self->{xdb}, $docid); } -# v1 only, where $mid is unique -sub remove_message { - my ($self, $mid) = @_; - $mid = mid_clean($mid); +sub remove_eidx_info { + my ($self, $docid, $eidx_key, $eml) = @_; + begin_txn_lazy($self); + my $doc = _get_doc($self, $docid) or return; + eval { $doc->remove_term('O'.$eidx_key) }; + warn "W: ->remove_term O$eidx_key: $@\n" if $@; + for my $l ($eml ? $eml->header_raw('List-Id') : ()) { + $l =~ /<([^>]+)>/ or next; + my $lid = lc $1; + eval { $doc->remove_term('G' . $lid) }; + warn "W: ->remove_term G$lid: $@\n" if $@; + + # nb: we don't remove the XL probabilistic terms + # since terms may overlap if cross-posted. + # + # IOW, a message which has both <foo.example.com> + # and <bar.example.com> would have overlapping + # "XLexample" and "XLcom" as terms and which we + # wouldn't know if they're safe to remove if we just + # unindex <foo.example.com> while preserving + # <bar.example.com>. + # + # In any case, this entire sub is will likely never + # be needed and users using the "l:" prefix are probably + # rarer. + } + $self->{xdb}->replace_document($docid, $doc); +} - if (my $over = $self->{over}) { - my $nr = eval { $over->remove_oid(undef, $mid) }; - if ($@) { - warn "failed to remove <$mid> from overview: $@\n"; - } elsif ($nr == 0) { - warn "<$mid> missing for removal from overview\n"; +sub set_vmd { + my ($self, $docid, $vmd) = @_; + begin_txn_lazy($self); + my $doc = _get_doc($self, $docid) or return; + my ($v, @rm, @add); + my @x = @VMD_MAP; + my ($cur, $end) = ($doc->termlist_begin, $doc->termlist_end); + while (my ($field, $pfx) = splice(@x, 0, 2)) { + my $set = $vmd->{$field} // next; + my %keep = map { $_ => 1 } @$set; + my %add = %keep; + $cur->skip_to($pfx); # works due to @VMD_MAP order + for (; $cur != $end; $cur++) { + $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 need_xapian($self); - my $db = $self->{xdb}; - my $nr = 0; - eval { - batch_do($self, 'Q' . $mid, sub { - my ($ids) = @_; - $db->delete_document($_) for @$ids; - $nr += scalar @$ids; - }); - }; - if ($@) { - warn "failed to remove <$mid> from Xapian: $@\n"; - } elsif ($nr == 0) { - warn "<$mid> missing for removal from Xapian\n"; - } + return unless scalar(@rm) || scalar(@add); + $doc->remove_term($_) for @rm; + $doc->add_boolean_term($_) for @add; + $self->{xdb}->replace_document($docid, $doc); } -# MID is a hint in V2 -sub remove_by_oid { - my ($self, $oid, $mid) = @_; +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; +} - $self->{over}->remove_oid($oid, $mid) if $self->{over}; +sub add_vmd { + my ($self, $docid, $vmd) = @_; + begin_txn_lazy($self); + my $doc = _get_doc($self, $docid) or return; + 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; +} - return unless need_xapian($self); - my $db = $self->{xdb}; +sub remove_vmd { + my ($self, $docid, $vmd) = @_; + begin_txn_lazy($self); + my $doc = _get_doc($self, $docid) or return; + my $replace; + 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; +} - # 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; +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; +} - # there is only ONE element in @delete unless we - # have bugs in our v2writable deduplication check - my @delete; - for (; $head != $tail; $head++) { - my $docid = $head->get_docid; - my $doc = $db->get_document($docid); - my $smsg = PublicInbox::SearchMsg->wrap($mid); - $smsg->load_expand($doc); - if ($smsg->{blob} eq $oid) { - push(@delete, $docid); - } +sub xdb_remove { + my ($self, @docids) = @_; + 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 Xapian? $@\n" if $@; } - $db->delete_document($_) foreach @delete; - scalar(@delete); } +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) = @_; - my $len = length($objid); for (my $len = length($objid); $len >= 7; ) { $doc->add_term($pfx.$objid); $objid = substr($objid, 0, --$len); } } -sub unindex_blob { - my ($self, $mime) = @_; - my $mid = eval { mid_clean(mid_mime($mime)) }; - $self->remove_message($mid) if defined $mid; +# v1 only +sub unindex_eml { + my ($self, $oid, $eml) = @_; + my $mids = mids($eml); + my $nr = 0; + my %tmp; + for my $mid (@$mids) { + my @removed = $self->{oidx}->remove_oid($oid, $mid); + $nr += scalar @removed; + $tmp{$_}++ for @removed; + } + if (!$nr) { + my $m = join('> <', @$mids); + warn "W: <$m> missing for removal from overview\n"; + } + while (my ($num, $nr) = each %tmp) { + warn "BUG: $num appears >1 times ($nr) for $oid\n" if $nr != 1; + } + if ($nr) { + $self->{mm}->num_delete($_) for (keys %tmp); + } else { # just in case msgmap and over.sqlite3 become desynched: + $self->{mm}->mid_delete($mids->[0]); + } + xdb_remove($self, keys %tmp) if need_xapian($self); } sub index_mm { - my ($self, $mime) = @_; - my $mid = mid_clean(mid_mime($mime)); + my ($self, $mime, $oid, $sync) = @_; + my $mids = mids($mime); my $mm = $self->{mm}; - my $num; - - if (defined $self->{regen_down}) { - $num = $mm->num_for($mid) and return $num; - - while (($num = $self->{regen_down}--) > 0) { - if ($mm->mid_set($num, $mid) != 0) { - return $num; - } - } - } elsif (defined $self->{regen_up}) { - $num = $mm->num_for($mid) and return $num; - - # this is to fixup old bugs due to add-remove-add - while (($num = ++$self->{regen_up})) { - if ($mm->mid_set($num, $mid) != 0) { - return $num; - } + if ($sync->{reindex}) { + my $oidx = $self->{oidx}; + for my $mid (@$mids) { + my ($num, undef) = $oidx->num_mid0_for_oid($oid, $mid); + return $num if defined $num; } + $mm->num_for($mids->[0]) // $mm->mid_insert($mids->[0]); + } else { + # fallback to num_for since filters like RubyLang set the number + $mm->mid_insert($mids->[0]) // $mm->num_for($mids->[0]); } - - $num = $mm->mid_insert($mid) and return $num; - - # fallback to num_for since filters like RubyLang set the number - $mm->num_for($mid); } -sub unindex_mm { - my ($self, $mime) = @_; - $self->{mm}->mid_delete(mid_clean(mid_mime($mime))); +sub is_bad_blob ($$$$) { + my ($oid, $type, $size, $expect_oid) = @_; + if ($type ne 'blob') { + carp "W: $expect_oid is not a blob (type=$type)"; + return 1; + } + croak "BUG: $oid != $expect_oid" if $oid ne $expect_oid; + $size == 0 ? 1 : 0; # size == 0 means purged } -sub index_both { - my ($self, $mime, $bytes, $blob) = @_; - my $num = index_mm($self, $mime); - add_message($self, $mime, $bytes, $num, $blob); +sub index_both { # git->cat_async callback + my ($bref, $oid, $type, $size, $sync) = @_; + return if is_bad_blob($oid, $type, $size, $sync->{oid}); + my ($nr, $max) = @$sync{qw(nr max)}; + ++$$nr; + $$max -= $size; + my $smsg = bless { blob => $oid }, 'PublicInbox::Smsg'; + $smsg->set_bytes($$bref, $size); + my $self = $sync->{sidx}; + local $self->{current_info} = "$self->{current_info}: $oid"; + my $eml = PublicInbox::Eml->new($bref); + $smsg->{num} = index_mm($self, $eml, $oid, $sync) or + die "E: could not generate NNTP article number for $oid"; + add_message($self, $eml, $smsg, $sync); + ++$self->{nidx}; + my $cur_cmt = $sync->{cur_cmt} // die 'BUG: {cur_cmt} missing'; + ${$sync->{latest_cmt}} = $cur_cmt; } -sub unindex_both { - my ($self, $mime) = @_; - unindex_blob($self, $mime); - unindex_mm($self, $mime); +sub unindex_both { # git->cat_async callback + my ($bref, $oid, $type, $size, $sync) = @_; + return if is_bad_blob($oid, $type, $size, $sync->{oid}); + my $self = $sync->{sidx}; + local $self->{current_info} = "$self->{current_info}: $oid"; + unindex_eml($self, $oid, PublicInbox::Eml->new($bref)); + # may be undef if leftover + if (defined(my $cur_cmt = $sync->{cur_cmt})) { + ${$sync->{latest_cmt}} = $cur_cmt; + } + ++$self->{nidx}; } -sub do_cat_mail { - my ($git, $blob, $sizeref) = @_; - my $mime = eval { - my $str = $git->cat_file($blob, $sizeref); - # fixup bugs from import: - $$str =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s; - PublicInbox::MIME->new($str); - }; - $@ ? undef : $mime; +sub with_umask { + my $self = shift; + my $owner = $self->{ibx} // $self->{eidx}; + $owner ? $owner->with_umask(@_) : $self->SUPER::with_umask(@_) } # called by public-inbox-index sub index_sync { - my ($self, $opts) = @_; - delete $self->{lock_path} if $opts->{-skip_lock}; - $self->{-inbox}->with_umask(sub { $self->_index_sync($opts) }) + my ($self, $opt) = @_; + delete $self->{lock_path} if $opt->{-skip_lock}; + $self->with_umask(\&_index_sync, $self, $opt); + if ($opt->{reindex} && !$opt->{quit} && + !grep(defined, @$opt{qw(since until)})) { + my %again = %$opt; + delete @again{qw(rethread reindex)}; + index_sync($self, \%again); + $opt->{quit} = $again{quit}; # propagate to caller + } } -sub batch_adjust ($$$$$) { - my ($max, $bytes, $batch_cb, $latest, $nr) = @_; - $$max -= $bytes; - if ($$max <= 0) { - $$max = BATCH_BYTES; - $batch_cb->($nr, $latest); +sub check_size { # check_async cb for -index --max-size=... + my (undef, $oid, $type, $size, $arg) = @_; + ($type // '') eq 'blob' or die "E: bad $oid in $arg->{git}->{git_dir}"; + if ($size <= $arg->{max_size}) { + $arg->{git}->cat_async($oid, $arg->{index_oid}, $arg); + } else { + warn "W: skipping $oid ($size > $arg->{max_size})\n"; + } +} + +sub v1_checkpoint ($$;$) { + my ($self, $sync, $stk) = @_; + $self->{ibx}->git->async_wait_all; + + # $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, $sync, $cur, $newest)) { + $self->{mm}->last_commit($newest); + } + } + ${$sync->{max}} = $self->{batch_bytes}; + + $self->{mm}->{dbh}->commit; + 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, $sync, $cur, $newest)) { + $xdb->set_metadata('last_commit', $newest); + } + } + if ($stk) { # all done if $stk is passed + # let SearchView know a full --reindex was done so it can + # generate ->has_threadid-dependent links + if ($xdb && $sync->{reindex} && !ref($sync->{reindex})) { + my $n = $xdb->get_metadata('has_threadid'); + $xdb->set_metadata('has_threadid', '1') if $n ne '1'; + } + $self->{oidx}->rethread_done($sync->{-opt}); # all done + } + commit_txn_lazy($self); + $sync->{ibx}->git->cleanup; + my $nr = ${$sync->{nr}}; + idx_release($self, $nr); + # let another process do some work... + if (my $pr = $sync->{-opt}->{-progress}) { + $pr->("indexed $nr/$sync->{ntodo}\n") if $nr; + } + if (!$stk && !$sync->{quit}) { # more to come + begin_txn_lazy($self); + $self->{mm}->{dbh}->begin_work; } } # only for v1 -sub read_log { - my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_; - my $hex = '[a-f0-9]'; - my $h40 = $hex .'{40}'; - my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!; - my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!; - my $git = $self->{git}; - my $latest; - my $bytes; - my $max = BATCH_BYTES; - local $/ = "\n"; - my %D; - my $line; - my $newest; +sub process_stack { + my ($self, $sync, $stk) = @_; + my $git = $sync->{ibx}->git; + my $max = $self->{batch_bytes}; my $nr = 0; - while (defined($line = <$log>)) { - if ($line =~ /$addmsg/o) { - my $blob = $1; - if (delete $D{$blob}) { - if (defined $self->{regen_down}) { - my $num = $self->{regen_down}--; - $self->{mm}->num_highwater($num); - } - next; - } - my $mime = do_cat_mail($git, $blob, \$bytes) or next; - batch_adjust(\$max, $bytes, $batch_cb, $latest, ++$nr); - $add_cb->($self, $mime, $bytes, $blob); - } elsif ($line =~ /$delmsg/o) { - my $blob = $1; - $D{$blob} = 1; - } elsif ($line =~ /^commit ($h40)/o) { - $latest = $1; - $newest ||= $latest; + $sync->{nr} = \$nr; + $sync->{max} = \$max; + $sync->{sidx} = $self; + $sync->{latest_cmt} = \(my $latest_cmt); + + $self->{mm}->{dbh}->begin_work; + if (my @leftovers = keys %{delete($sync->{D}) // {}}) { + warn('W: unindexing '.scalar(@leftovers)." leftovers\n"); + for my $oid (@leftovers) { + last if $sync->{quit}; + $oid = unpack('H*', $oid); + $git->cat_async($oid, \&unindex_both, $sync); } } - close($log) or die "git log failed: \$?=$?"; - # get the leftovers - foreach my $blob (keys %D) { - my $mime = do_cat_mail($git, $blob, \$bytes) or next; - $del_cb->($self, $mime); + if ($sync->{max_size} = $sync->{-opt}->{max_size}) { + $sync->{index_oid} = \&index_both; + } + while (my ($f, $at, $ct, $oid, $cur_cmt) = $stk->pop_rec) { + my $arg = { %$sync, cur_cmt => $cur_cmt, oid => $oid }; + last if $sync->{quit}; + if ($f eq 'm') { + $arg->{autime} = $at; + $arg->{cotime} = $ct; + if ($sync->{max_size}) { + $arg->{git} = $git; + $git->check_async($oid, \&check_size, $arg); + } else { + $git->cat_async($oid, \&index_both, $arg); + } + v1_checkpoint($self, $sync) if $max <= 0; + } elsif ($f eq 'd') { + $git->cat_async($oid, \&unindex_both, $arg); + } } - $batch_cb->($nr, $latest, $newest); + v1_checkpoint($self, $sync, $sync->{quit} ? undef : $stk); } -sub _git_log { - my ($self, $opts, $range) = @_; - my $git = $self->{git}; - - if (index($range, '..') < 0) { - # don't show annoying git errrors to users who run -index - # on empty inboxes - $git->qx(qw(rev-parse -q --verify), "$range^0"); - if ($?) { - open my $fh, '<', '/dev/null' or - die "failed to open /dev/null: $!\n"; - return $fh; - } +sub log2stack ($$$) { + my ($sync, $git, $range) = @_; + my $D = $sync->{D}; # OID_BIN => NR (if reindexing, undef otherwise) + my ($add, $del); + if ($sync->{ibx}->version == 1) { + my $path = $hex.'{2}/'.$hex.'{38}'; + $add = qr!\A:000000 100644 \S+ ($OID) A\t$path$!; + $del = qr!\A:100644 000000 ($OID) \S+ D\t$path$!; + } else { + $del = qr!\A:\d{6} 100644 $OID ($OID) [AM]\td$!; + $add = qr!\A:\d{6} 100644 $OID ($OID) [AM]\tm$!; } # Count the new files so they can be added newest to oldest # and still have numbers increasing from oldest to newest - my $fcount = 0; - my $pr = $opts->{-progress}; - $pr->("counting changes\n\t$range ... ") if $pr; - # can't use 'rev-list --count' if we use --diff-filter - my $fh = $git->popen(qw(log --pretty=tformat:%h - --no-notes --no-color --no-renames - --diff-filter=AM), $range); - ++$fcount while <$fh>; - close $fh or die "git log failed: \$?=$?"; - my $high = $self->{mm}->num_highwater; - $pr->("$fcount\n") if $pr; # continue previous line - $self->{ntodo} = $fcount; - - if (index($range, '..') < 0) { - if ($high && $high == $fcount) { - # fix up old bugs in full indexes which caused messages to - # not appear in Msgmap - $self->{regen_up} = $high; - } else { - # normal regen is for for fresh data - $self->{regen_down} = $fcount; + 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 ($l =~ /\A([0-9]+)-([0-9]+)-($OID)$/o) { + ($at, $ct, $cmt) = ($1 + 0, $2 + 0, $3); + $stk //= PublicInbox::IdxStack->new($cmt); + } 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 ($l =~ /$add/) { + my $oid = $1; + if ($D) { + my $oid_bin = pack('H*', $oid); + my $nr = --$D->{$oid_bin}; + delete($D->{$oid_bin}) if $nr <= 0; + # nr < 0 (-1) means it never existed + next if $nr >= 0; + } + $stk->push_rec('m', $at, $ct, $oid, $cmt); } - } else { - # Give oldest messages the smallest numbers - $self->{regen_down} = $high + $fcount; } + $fh->close or die "git log failed: \$?=$?"; + $stk //= PublicInbox::IdxStack->new; + $stk->read_prepare; +} + +sub prepare_stack ($$) { + my ($sync, $range) = @_; + my $git = $sync->{ibx}->git; - $git->popen(qw/log --no-notes --no-color --no-renames - --raw -r --no-abbrev/, $range); + if (index($range, '..') < 0) { + # don't show annoying git errors to users who run -index + # on empty inboxes + $git->qx(qw(rev-parse -q --verify), "$range^0"); + return PublicInbox::IdxStack->new->read_prepare if $?; + } + $sync->{D} = $sync->{reindex} ? {} : undef; # OID_BIN => NR + log2stack($sync, $git, $range); } # --is-ancestor requires git 1.8.0+ @@ -661,15 +1005,20 @@ sub is_ancestor ($$$) { return 0 unless $git->check($cur); my $cmd = [ 'git', "--git-dir=$git->{git_dir}", qw(merge-base --is-ancestor), $cur, $tip ]; - my $pid = spawn($cmd); - waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish'; - $? == 0; + run_wait($cmd) == 0; } -sub need_update ($$$) { - my ($self, $cur, $new) = @_; - my $git = $self->{git}; - return 1 if $cur && !is_ancestor($git, $cur, $new); +sub need_update ($$$$) { + my ($self, $sync, $cur, $new) = @_; + my $git = $self->{ibx}->git; + $cur //= ''; # XS Search::Xapian ->get_metadata doesn't give undef + + # don't rewind if --{since,until,before,after} are in use + return if $cur ne '' && + grep(defined, @{$sync->{-opt}}{qw(since until)}) && + is_ancestor($git, $new, $cur); + + return 1 if $cur ne '' && !is_ancestor($git, $cur, $new); my $range = $cur eq '' ? $new : "$cur..$new"; chomp(my $n = $git->qx(qw(rev-list --count), $range)); ($n eq '' || $n > 0); @@ -688,7 +1037,7 @@ sub _last_x_commit { $lx = $lm; } # Use last_commit from msgmap if it is older or unset - if (!$lm || ($lx && $lm && is_ancestor($self->{git}, $lm, $lx))) { + if (!$lm || ($lx && $lm && is_ancestor($self->{ibx}->git, $lm, $lx))) { $lx = $lm; } $lx; @@ -700,70 +1049,55 @@ sub reindex_from ($$) { ref($reindex) eq 'HASH' ? $reindex->{from} : ''; } +sub quit_cb ($) { + my ($sync) = @_; + sub { + # we set {-opt}->{quit} too, so ->index_sync callers + # can abort multi-inbox loops this way + $sync->{quit} = $sync->{-opt}->{quit} = 1; + warn "gracefully quitting\n"; + } +} + # indexes all unindexed messages (v1 only) sub _index_sync { - my ($self, $opts) = @_; - my $tip = $opts->{ref} || 'HEAD'; - my ($last_commit, $lx, $xlog); - my $git = $self->{git}; - $git->batch_prepare; - my $pr = $opts->{-progress}; - + my ($self, $opt) = @_; + my $tip = $opt->{ref} || 'HEAD'; + my $ibx = $self->{ibx}; + local $self->{current_info} = "$ibx->{inboxdir}"; + $self->{batch_bytes} = $opt->{batch_size} // $BATCH_BYTES; + + if ($X->{CLOEXEC_UNSET}) { + $ibx->git->cat_file($tip); + $ibx->git->check($tip); + } + my $pr = $opt->{-progress}; + my $sync = { reindex => $opt->{reindex}, -opt => $opt, ibx => $ibx }; + my $quit = quit_cb($sync); + local $SIG{QUIT} = $quit; + local $SIG{INT} = $quit; + local $SIG{TERM} = $quit; my $xdb = $self->begin_txn_lazy; + $self->{oidx}->rethread_prepare($opt); my $mm = _msgmap_init($self); - do { - if ($xlog) { - close($xlog) or die "git log failed: \$?=$?"; - $xlog = undef; - } - $last_commit = _last_x_commit($self, $mm); - $lx = reindex_from($opts->{reindex}, $last_commit); - - $self->{over}->rollback_lazy; - $self->{over}->disconnect; - $git->cleanup; - delete $self->{txn}; - $xdb->cancel_transaction if $xdb; - $xdb = _xdb_release($self); - - # ensure we leak no FDs to "git log" with Xapian <= 1.2 - my $range = $lx eq '' ? $tip : "$lx..$tip"; - $xlog = _git_log($self, $opts, $range); - - $xdb = $self->begin_txn_lazy; - } while (_last_x_commit($self, $mm) ne $last_commit); - - my $dbh = $mm->{dbh} if $mm; - my $cb = sub { - my ($nr, $commit, $newest) = @_; - if ($dbh) { - if ($newest) { - my $cur = $mm->last_commit || ''; - if (need_update($self, $cur, $newest)) { - $mm->last_commit($newest); - } - } - $dbh->commit; - } - if ($newest && need_xapian($self)) { - my $cur = $xdb->get_metadata('last_commit'); - if (need_update($self, $cur, $newest)) { - $xdb->set_metadata('last_commit', $newest); - } - } - $self->commit_txn_lazy; - $git->cleanup; - $xdb = _xdb_release($self); - # let another process do some work... < - $pr->("indexed $nr/$self->{ntodo}\n") if $pr && $nr; - if (!$newest) { - $xdb = $self->begin_txn_lazy; - $dbh->begin_work if $dbh; + if ($sync->{reindex}) { + my $last = $mm->last_commit; + if ($last) { + $tip = $last; + } else { + # somebody just blindly added --reindex when indexing + # for the first time, allow it: + undef $sync->{reindex}; } - }; - - $dbh->begin_work; - read_log($self, $xlog, *index_both, *unindex_both, $cb); + } + my $last_commit = _last_x_commit($self, $mm); + my $lx = reindex_from($sync->{reindex}, $last_commit); + my $range = $lx eq '' ? $tip : "$lx..$tip"; + $pr->("counting changes\n\t$range ... ") if $pr; + my $stk = prepare_stack($sync, $range); + $sync->{ntodo} = $stk ? $stk->num_records : 0; + $pr->("$sync->{ntodo}\n") if $pr; # continue previous line + process_stack($self, $sync, $stk) if !$sync->{quit}; } sub DESTROY { @@ -772,80 +1106,67 @@ sub DESTROY { $_[0]->{lockfh} = undef; } -# remote_* subs are only used by SearchIdxPart -sub remote_commit { - my ($self) = @_; - if (my $w = $self->{w}) { - print $w "commit\n" or die "failed to write commit: $!"; - } else { - $self->commit_txn_lazy; - } -} - -sub remote_close { - my ($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 $self->{xdb}; - } -} - -sub remote_remove { - my ($self, $oid, $mid) = @_; - if (my $w = $self->{w}) { - # triggers remove_by_oid in a shard - print $w "D $oid $mid\n" or die "failed to write remove $!"; - } else { - $self->begin_txn_lazy; - $self->remove_by_oid($oid, $mid); - } -} - sub begin_txn_lazy { my ($self) = @_; return if $self->{txn}; - - $self->{-inbox}->with_umask(sub { - my $xdb = $self->{xdb} || $self->_xdb_acquire; - $self->{over}->begin_lazy if $self->{over}; - $xdb->begin_transaction if $xdb; - $self->{txn} = 1; - $xdb; - }); + my $restore = $self->with_umask; + my $xdb = $self->{xdb} || idx_acquire($self); + $self->{oidx}->begin_lazy if $self->{oidx}; + $xdb->begin_transaction if $xdb; + $self->{txn} = 1; + $xdb; } -sub commit_txn_lazy { +# store 'indexlevel=medium' in v2 shard=0 and v1 (only one shard) +# This metadata is read by InboxWritable->detect_indexlevel: +sub set_metadata_once { my ($self) = @_; - delete $self->{txn} or return; - $self->{-inbox}->with_umask(sub { - if (my $xdb = $self->{xdb}) { - - # store 'indexlevel=medium' in v2 shard=0 and - # v1 (only one shard) - # This metadata is read by Admin::detect_indexlevel: - if (!$self->{shard} # undef or 0, not >0 - && $self->{indexlevel} eq 'medium') { - $xdb->set_metadata('indexlevel', 'medium'); - } - $xdb->commit_transaction; + return if $self->{shard}; # only continue if undef or 0, not >0 + my $xdb = $self->{xdb}; + + if (delete($self->{-set_has_threadid_once})) { + $xdb->set_metadata('has_threadid', '1'); + } + if (delete($self->{-set_indexlevel_once})) { + my $level = $xdb->get_metadata('indexlevel'); + if (!$level || $level ne 'medium') { + $xdb->set_metadata('indexlevel', 'medium'); } - $self->{over}->commit_lazy if $self->{over}; - }); + } + if (delete($self->{-set_skip_docdata_once})) { + $xdb->get_metadata('skip_docdata') or + $xdb->set_metadata('skip_docdata', '1'); + } } -sub worker_done { +sub commit_txn_lazy { my ($self) = @_; - if (need_xapian($self)) { - die "$$ $0 xdb not released\n" if $self->{xdb}; + return unless delete($self->{txn}); + my $restore = $self->with_umask; + if (my $eidx = $self->{eidx}) { + $eidx->git->async_wait_all; + $eidx->{transact_bytes} = 0; } - die "$$ $0 still in transaction\n" if $self->{txn}; + if (my $xdb = $self->{xdb}) { + set_metadata_once($self); + $xdb->commit_transaction; + } + $self->{oidx}->commit_lazy if $self->{oidx}; +} + +sub eidx_shard_new { + my ($class, $eidx, $shard) = @_; + my $self = bless { + eidx => $eidx, + xpfx => $eidx->{xpfx}, + indexlevel => $eidx->{indexlevel}, + -skip_docdata => 1, + shard => $shard, + creat => 1, + }, $class; + $self->{-set_indexlevel_once} = 1 if $self->{indexlevel} eq 'medium'; + $self; } 1; |