use base qw(PublicInbox::Search PublicInbox::Lock);
use PublicInbox::MIME;
use PublicInbox::InboxWritable;
-use PublicInbox::MID qw/mid_clean id_compress mid_mime mids references/;
+use PublicInbox::MID qw/mid_clean id_compress mid_mime mids/;
use PublicInbox::MsgIter;
use Carp qw(croak);
use POSIX qw(strftime);
use PublicInbox::OverIdx;
+use PublicInbox::Spawn qw(spawn);
require PublicInbox::Git;
use Compress::Zlib qw(compress);
my $mids = mids($mime->header_obj);
$mid0 = $mids->[0] unless defined $mid0; # v1 compatibility
unless (defined $num) { # v1
- my $mm = $self->_msgmap_init;
- $num = $mm->mid_insert($mid0) || $mm->num_for($mid0);
+ $self->_msgmap_init;
+ $num = index_mm($self, $mime);
}
eval {
my $smsg = PublicInbox::SearchMsg->new($mime);
my $doc = $smsg->{doc};
my $subj = $smsg->subject;
- my $xpath;
- if ($subj ne '') {
- $xpath = $self->subject_path($subj);
- $xpath = id_compress($xpath);
- }
-
- $smsg->{lines} = $mime->body_raw =~ tr!\n!\n!;
- defined $bytes or $bytes = length($mime->as_string);
- $smsg->{bytes} = $bytes;
-
add_val($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(), $dt);
- my @vals = ($smsg->{ts}, $smsg->{ds});
my $tg = $self->term_generator;
index_body($tg, \@orig, $doc) if @orig;
});
- # populates smsg->references for smsg->to_doc_data
- my $refs = parse_references($smsg, $mid0, $mids);
- my $data = $smsg->to_doc_data($oid, $mid0);
foreach my $mid (@$mids) {
$tg->index_text($mid, 1, 'XM');
}
+ $smsg->{to} = $smsg->{cc} = '';
+ PublicInbox::OverIdx::parse_references($smsg, $mid0, $mids);
+ my $data = $smsg->to_doc_data($oid, $mid0);
$doc->set_data($data);
if (my $altid = $self->{-altid}) {
foreach my $alt (@$altid) {
}
}
- $self->delete_article($num) if defined $num; # for reindexing
-
- utf8::encode($data);
- $data = compress($data);
- push @vals, $num, $mids, $refs, $xpath, $data;
- $self->{over}->add_over(\@vals);
+ if (my $over = $self->{over}) {
+ $over->add_overview($mime, $bytes, $num, $oid, $mid0);
+ }
$doc->add_boolean_term('Q' . $_) foreach @$mids;
- $doc->add_boolean_term('XNUM' . $num) if defined $num;
- $doc_id = $self->{xdb}->add_document($doc);
+ $self->{xdb}->replace_document($doc_id = $num, $doc);
};
if ($@) {
my $db = $self->{xdb};
my $called;
$mid = mid_clean($mid);
+ my $over = $self->{over};
eval {
batch_do($self, 'Q' . $mid, sub {
my ($ids) = @_;
$db->delete_document($_) for @$ids;
+ $over->delete_articles($ids) if $over;
$called = 1;
});
};
}
}
-sub delete_article {
- my ($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};
+ $self->{over}->remove_oid($oid, $mid) if $self->{over};
+
# 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.
# there is only ONE element in @delete unless we
# have bugs in our v2writable deduplication check
my @delete;
- my @over_del;
for (; $head != $tail; $head->inc) {
my $docid = $head->get_docid;
my $doc = $db->get_document($docid);
$smsg->load_expand;
if ($smsg->{blob} eq $oid) {
push(@delete, $docid);
- push(@over_del, $smsg->num);
}
}
$db->delete_document($_) foreach @delete;
- $self->{over}->remove_oid($oid, $mid);
scalar(@delete);
}
$self->{term_generator} = $tg;
}
-sub parse_references ($$$) {
- my ($smsg, $mid0, $mids) = @_;
- my $mime = $smsg->{mime};
- my $hdr = $mime->header_obj;
- my $refs = references($hdr);
- push(@$refs, @$mids) if scalar(@$mids) > 1;
- return $refs if scalar(@$refs) == 0;
-
- # prevent circular references here:
- my %seen = ( $mid0 => 1 );
- my @keep;
- foreach my $ref (@$refs) {
- if (length($ref) > PublicInbox::MID::MAX_MID_SIZE) {
- warn "References: <$ref> too long, ignoring\n";
- next;
- }
- next if $seen{$ref}++;
- push @keep, $ref;
- }
- $smsg->{references} = '<'.join('> <', @keep).'>' if @keep;
- \@keep;
-}
-
sub index_git_blob_id {
my ($doc, $pfx, $objid) = @_;
my ($self, $mime) = @_;
my $mid = mid_clean(mid_mime($mime));
my $mm = $self->{mm};
- my $num = $mm->mid_insert($mid);
- return $num if defined $num;
+ my $num;
+
+ if (defined $self->{regen_down}) {
+ $num = $mm->num_for($mid) and return $num;
+
+ while (($num = $self->{regen_down}--) > 0) {
+ if ($mm->mid_set($num, $mid) != 0) {
+ return $num;
+ }
+ }
+ } elsif (defined $self->{regen_up}) {
+ $num = $mm->num_for($mid) and return $num;
+
+ # this is to fixup old bugs due to add-remove-add
+ while (($num = ++$self->{regen_up})) {
+ if ($mm->mid_set($num, $mid) != 0) {
+ return $num;
+ }
+ }
+ }
+
+ $num = $mm->mid_insert($mid) and return $num;
# fallback to num_for since filters like RubyLang set the number
$mm->num_for($mid);
$self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
}
-sub index_mm2 {
- my ($self, $mime, $bytes, $blob) = @_;
- my $num = $self->{mm}->num_for(mid_clean(mid_mime($mime)));
- add_message($self, $mime, $bytes, $num, $blob);
-}
-
-sub unindex_mm2 {
- my ($self, $mime) = @_;
- $self->{mm}->mid_delete(mid_clean(mid_mime($mime)));
- unindex_blob($self, $mime);
-}
-
sub index_both {
my ($self, $mime, $bytes, $blob) = @_;
my $num = index_mm($self, $mime);
$$max -= $bytes;
if ($$max <= 0) {
$$max = BATCH_BYTES;
- $batch_cb->($latest, 1);
+ $batch_cb->($latest);
}
}
# only for v1
-sub rlog {
+sub read_log {
my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
my $hex = '[a-f0-9]';
my $h40 = $hex .'{40}';
my $bytes;
my $max = BATCH_BYTES;
local $/ = "\n";
+ my %D;
my $line;
+ my $newest;
+ my $mid = '20170114215743.5igbjup6qpsh3jfg@genre.crustytoothpaste.net';
while (defined($line = <$log>)) {
if ($line =~ /$addmsg/o) {
my $blob = $1;
+ delete $D{$blob} and next;
my $mime = do_cat_mail($git, $blob, \$bytes) or next;
+ my $mids = mids($mime->header_obj);
+ foreach (@$mids) {
+ warn "ADD $mid\n" if ($_ eq $mid);
+ }
batch_adjust(\$max, $bytes, $batch_cb, $latest);
$add_cb->($self, $mime, $bytes, $blob);
} elsif ($line =~ /$delmsg/o) {
my $blob = $1;
- my $mime = do_cat_mail($git, $blob, \$bytes) or next;
- batch_adjust(\$max, $bytes, $batch_cb, $latest);
- $del_cb->($self, $mime);
+ $D{$blob} = 1;
} elsif ($line =~ /^commit ($h40)/o) {
$latest = $1;
+ $newest ||= $latest;
}
}
- $batch_cb->($latest, 0);
+ # get the leftovers
+ foreach my $blob (keys %D) {
+ my $mime = do_cat_mail($git, $blob, \$bytes) or next;
+ my $mids = mids($mime->header_obj);
+ foreach (@$mids) {
+ warn "DEL $mid\n" if ($_ eq $mid);
+ }
+ $del_cb->($self, $mime);
+ }
+ $batch_cb->($latest, $newest);
}
sub _msgmap_init {
sub _git_log {
my ($self, $range) = @_;
- $self->{git}->popen(qw/log --reverse --no-notes --no-color
+ my $git = $self->{git};
+
+ if (index($range, '..') < 0) {
+ my $regen_max = 0;
+ # 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), $range);
+ ++$regen_max while <$fh>;
+ my (undef, $max) = $self->{mm}->minmax;
+
+ if ($max && $max == $regen_max) {
+ # fix up old bugs in full indexes which caused messages to
+ # not appear in Msgmap
+ $self->{regen_up} = $max;
+ } else {
+ # normal regen is for for fresh data
+ $self->{regen_down} = $regen_max;
+ }
+ }
+
+ $git->popen(qw/log --no-notes --no-color --no-renames
--raw -r --no-abbrev/, $range);
}
-# indexes all unindexed messages
+sub is_ancestor ($$$) {
+ my ($git, $cur, $tip) = @_;
+ return 0 unless $git->check($cur);
+ my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
+ qw(merge-base --is-ancestor), $cur, $tip ];
+ my $pid = spawn($cmd);
+ defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
+ waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
+ $? == 0;
+}
+
+sub need_update ($$$) {
+ my ($self, $cur, $new) = @_;
+ my $git = $self->{git};
+ return 1 if $cur && !is_ancestor($git, $cur, $new);
+ my $range = $cur eq '' ? $new : "$cur..$new";
+ chomp(my $n = $git->qx(qw(rev-list --count), $range));
+ ($n eq '' || $n > 0);
+}
+
+# indexes all unindexed messages (v1 only)
sub _index_sync {
my ($self, $opts) = @_;
my $tip = $opts->{ref} || 'HEAD';
my $reindex = $opts->{reindex};
my ($mkey, $last_commit, $lx, $xlog);
- $self->{git}->batch_prepare;
+ my $git = $self->{git};
+ $git->batch_prepare;
+
my $xdb = $self->begin_txn_lazy;
+ my $mm = _msgmap_init($self);
do {
$xlog = undef;
$mkey = 'last_commit';
$lx = '';
$mkey = undef if $last_commit ne '';
}
+
+ # use last_commit from msgmap if it is older or unset
+ my $lm = $mm->last_commit || '';
+ if (!$lm || ($lm && $lx && is_ancestor($git, $lm, $lx))) {
+ $lx = $lm;
+ }
+
$self->{over}->rollback_lazy;
$self->{over}->disconnect;
delete $self->{txn};
$xdb->cancel_transaction;
$xdb = _xdb_release($self);
- # ensure we leak no FDs to "git log"
+ # ensure we leak no FDs to "git log" with Xapian <= 1.2
my $range = $lx eq '' ? $tip : "$lx..$tip";
$xlog = _git_log($self, $range);
$xdb = $self->begin_txn_lazy;
} while ($xdb->get_metadata('last_commit') ne $last_commit);
- my $mm = _msgmap_init($self);
my $dbh = $mm->{dbh} if $mm;
- my $mm_only;
my $cb = sub {
- my ($commit, $more) = @_;
+ my ($commit, $newest) = @_;
if ($dbh) {
- $mm->last_commit($commit) if $commit;
+ if ($newest) {
+ my $cur = $mm->last_commit || '';
+ if (need_update($self, $cur, $newest)) {
+ $mm->last_commit($newest);
+ }
+ }
$dbh->commit;
}
- if (!$mm_only) {
- $xdb->set_metadata($mkey, $commit) if $mkey && $commit;
- $self->commit_txn_lazy;
+ if ($mkey && $newest) {
+ my $cur = $xdb->get_metadata($mkey);
+ if (need_update($self, $cur, $newest)) {
+ $xdb->set_metadata($mkey, $newest);
+ }
}
+ $self->commit_txn_lazy;
# let another process do some work... <
- if ($more) {
- if (!$mm_only) {
- $xdb = $self->begin_txn_lazy;
- }
+ if (!$newest) {
+ $xdb = $self->begin_txn_lazy;
$dbh->begin_work if $dbh;
}
};
- if ($mm) {
- $dbh->begin_work;
- my $lm = $mm->last_commit || '';
- if ($lm eq $lx) {
- # Common case is the indexes are synced,
- # we only need to run git-log once:
- rlog($self, $xlog, *index_both, *unindex_both, $cb);
- } else {
- # Uncommon case, msgmap and xapian are out-of-sync
- # do not care for performance (but git is fast :>)
- # This happens if we have to reindex Xapian since
- # msgmap is a frozen format and our Xapian format
- # is evolving.
- my $r = $lm eq '' ? $tip : "$lm..$tip";
-
- # first, ensure msgmap is up-to-date:
- my $mkey_prev = $mkey;
- $mkey = undef; # ignore xapian, for now
- my $mlog = _git_log($self, $r);
- $mm_only = 1;
- rlog($self, $mlog, *index_mm, *unindex_mm, $cb);
- $mm_only = $mlog = undef;
-
- # now deal with Xapian
- $mkey = $mkey_prev;
- $dbh = undef;
- rlog($self, $xlog, *index_mm2, *unindex_mm2, $cb);
- }
- } else {
- # user didn't install DBD::SQLite and DBI
- rlog($self, $xlog, *add_message, *unindex_blob, $cb);
- }
+ $dbh->begin_work;
+ read_log($self, $xlog, *index_both, *unindex_both, $cb);
}
sub DESTROY {
my ($self) = @_;
return if $self->{txn};
my $xdb = $self->{xdb} || $self->_xdb_acquire;
- $self->{over}->begin_lazy;
+ $self->{over}->begin_lazy if $self->{over};
$xdb->begin_transaction;
$self->{txn} = 1;
$xdb;
my ($self) = @_;
delete $self->{txn} or return;
$self->{xdb}->commit_transaction;
- $self->{over}->commit_lazy;
+ $self->{over}->commit_lazy if $self->{over};
}
sub worker_done {