* [PATCH 25/28] cindex: implement reindex
@ 2023-03-21 23:07 4% ` Eric Wong
0 siblings, 0 replies; 2+ results
From: Eric Wong @ 2023-03-21 23:07 UTC (permalink / raw)
To: meta
This allows changing --indexlevel at the moment and will allow
us to fix some yet-to-be-discovered bugs or backwards-compatible
improvements in the future.
---
lib/PublicInbox/CodeSearchIdx.pm | 33 ++++++++++++++++++++++----------
t/cindex.t | 4 ++++
2 files changed, 27 insertions(+), 10 deletions(-)
diff --git a/lib/PublicInbox/CodeSearchIdx.pm b/lib/PublicInbox/CodeSearchIdx.pm
index 095c153e..5e6c0d22 100644
--- a/lib/PublicInbox/CodeSearchIdx.pm
+++ b/lib/PublicInbox/CodeSearchIdx.pm
@@ -43,7 +43,8 @@ our (
@RDONLY_SHARDS, # Xapian::Database
@IDX_SHARDS, # clones of self
$MAX_SIZE,
- $TMP_GIT, # PublicInbox::Git object for --reindex and --prune
+ $TMP_GIT, # PublicInbox::Git object for --prune
+ $REINDEX, # PublicInbox::SharedKV
);
# stop walking history if we see >$SEEN_MAX existing commits, this assumes
@@ -89,12 +90,13 @@ sub new {
# TODO: may be used for reshard/compact
sub count_shards { scalar($_[0]->xdb_shards_flat) }
-sub add_commit ($$) {
+sub update_commit ($$) {
my ($self, $cmt) = @_; # fields from @FMT
my $x = 'Q'.$cmt->{H};
- for (docids_by_postlist($self, $x)) {
- $self->{xdb}->delete_document($_)
- }
+ my ($docid, @extra) = sort { $a <=> $b } docids_by_postlist($self, $x);
+ @extra and warn "W: $cmt->{H} indexed multiple times, pruning ",
+ join(', ', map { "#$_" } @extra), "\n";
+ $self->{xdb}->delete_document($_) for @extra;
my $doc = $PublicInbox::Search::X{Document}->new;
$doc->add_boolean_term($x);
$doc->add_boolean_term('G'.$_) for @{$self->{roots}};
@@ -119,7 +121,8 @@ sub add_commit ($$) {
$x = delete $cmt->{b};
$self->index_body_text($doc, \$x) if $x =~ /\S/s;
- $self->{xdb}->add_document($doc);
+ defined($docid) ? $self->{xdb}->replace_document($docid, $doc) :
+ $self->{xdb}->add_document($doc);
}
sub progress {
@@ -235,7 +238,7 @@ sub shard_index { # via wq_io_do
cidx_ckpoint($self, "[$n] $nr");
$TXN_BYTES = $batch_bytes - $len;
}
- add_commit($self, $cmt);
+ update_commit($self, $cmt);
++$nr;
if ($TXN_BYTES <= 0) {
cidx_ckpoint($self, "[$n] $nr");
@@ -398,7 +401,7 @@ sub check_existing { # retry_reopen callback
my $docid = shift(@docids) // return get_roots($self, $git);
my $doc = $shard->{xdb}->get_document($docid) //
die "BUG: no #$docid ($git->{git_dir})";
- my $old_fp = $doc->get_data;
+ my $old_fp = $REINDEX ? "\0invalid" : $doc->get_data;
if ($old_fp eq $git->{-repo}->{fp}) { # no change
delete $git->{-repo};
return;
@@ -426,7 +429,10 @@ sub partition_refs ($$$) {
while (defined(my $cmt = <$rfh>)) {
chomp $cmt;
my $n = hex(substr($cmt, 0, 8)) % scalar(@RDONLY_SHARDS);
- if (seen($RDONLY_SHARDS[$n], 'Q'.$cmt)) {
+ if ($REINDEX && $REINDEX->set_maybe(pack('H*', $cmt), '')) {
+ say { $shard_in[$n] } $cmt or die "say: $!";
+ ++$nchange;
+ } elsif (seen($RDONLY_SHARDS[$n], 'Q'.$cmt)) {
last if ++$seen > $SEEN_MAX;
} else {
say { $shard_in[$n] } $cmt or die "say: $!";
@@ -687,7 +693,7 @@ sub parent_quit {
sub init_tmp_git_dir ($) {
my ($self) = @_;
- return unless ($self->{-opt}->{prune} || $self->{-opt}->{reindex});
+ return unless $self->{-opt}->{prune};
require File::Temp;
require PublicInbox::Import;
my $tmp = File::Temp->newdir('cidx-all-git-XXXX', TMPDIR => 1);
@@ -729,6 +735,13 @@ sub cidx_run { # main entry point
$cb->($m, @_);
};
load_existing($self);
+ local $REINDEX;
+ if ($self->{-opt}->{reindex}) {
+ require PublicInbox::SharedKV;
+ $REINDEX = PublicInbox::SharedKV->new;
+ delete $REINDEX->{lock_path};
+ $REINDEX->dbh;
+ }
my @nc = grep { File::Spec->canonpath($_) ne $_ } @{$self->{git_dirs}};
if (@nc) {
warn "E: BUG? paths in $self->{cidx_dir} not canonicalized:\n";
diff --git a/t/cindex.t b/t/cindex.t
index 5d269217..eb66b2e6 100644
--- a/t/cindex.t
+++ b/t/cindex.t
@@ -93,6 +93,10 @@ EOM
ok(run_script([qw(-cindex -qu -d), "$tmp/ext"]), '-cindex -u');
$mset = $csrch->reopen->mset('dfn:for-update');
is(scalar($mset->items), 1, 'got updated result');
+
+ ok(run_script([qw(-cindex -qu --reindex -d), "$tmp/ext"]), 'reindex');
+ $mset = $csrch->reopen->mset('dfn:for-update');
+ is(scalar($mset->items), 1, 'same result after reindex');
}
if ('--prune') {
^ permalink raw reply related [relevance 4%]
* [PATCH 00/28] cindex coderepo commit indexer
@ 2023-03-21 23:07 7% Eric Wong
0 siblings, 1 reply; 2+ results
From: Eric Wong @ 2023-03-21 23:07 UTC (permalink / raw)
To: meta
Not wired up to WWW nor lei, yet; but indexing + pruning of
commits works.
I'm not sure if indexing (root) tree OIDs or committer
names+emails is worth it, since I don't think those are very
important terms to search for.
I first wanted to shoehorn this into extindex, but I think it
works better as a separate Xapian schema.
It allows both internal indexes ($GIT_DIR/public-inbox-cindex)
for unforked repos, as well as extindex-style external index
to encompass several projects.
The indexer is structured a bit more nicely than existing
indexers since I'm relying on OnDestroy and `local', more.
I would like to trickle some of these improvements back to
the mail indexers at some point.
--prune and --reindex currently block incremental updates, which
isn't great since both take a while for giant Xapian DBs.
Pruning is pretty important since it's much common for coderepos
(e.g. `seen' branch of git.git)
`lei cq' will probably be a new command which behaves
similarly to `lei q -f text', but takes `git log' options
for output...
Eric Wong (28):
ipc: move nproc_shards from v2writable
search: relocate all_terms from lei_search
admin: hoist out resolve_git_dir
admin: ensure resolved GIT_DIR is absolute
test_common: create_inbox: use `$!' properly on mkdir failure
codesearch: initial cut w/ -cindex tool
cindex: parallelize prep phases
cindex: use read-only shards during prep phases
searchidxshard: improve comment wording
cindex: use DS and workqueues for parallelism
ds: @post_loop_do replaces SetPostLoopCallback
cindex: implement --exclude= like -clone
cindex: show shard number in progress message
cindex: drop `unchanged' progress message
cindex: handle graceful shutdown by default
sigfd: pass signal name rather than number to callback
cindex: implement --max-size=SIZE
cindex: check for checkpoint before giant messages
cindex: truncate or drop body for over-sized commits
cindex: attempt to give oldest commits lowest docids
cindex: improve granularity of quit checks
spawn: show failing directory for chdir failures
cindex: filter out non-existent git directories
cindex: add support for --prune
cindex: implement reindex
cindex: squelch incompatible options
cindex: respect existing permissions
cindex: ignore SIGPIPE
MANIFEST | 4 +
lib/PublicInbox/Admin.pm | 18 +-
lib/PublicInbox/CodeSearch.pm | 121 +++++
lib/PublicInbox/CodeSearchIdx.pm | 835 ++++++++++++++++++++++++++++++
lib/PublicInbox/Config.pm | 2 +-
lib/PublicInbox/DS.pm | 30 +-
lib/PublicInbox/Daemon.pm | 4 +-
lib/PublicInbox/ExtSearchIdx.pm | 2 +-
lib/PublicInbox/IPC.pm | 33 +-
lib/PublicInbox/LEI.pm | 4 +-
lib/PublicInbox/LeiSearch.pm | 14 -
lib/PublicInbox/MiscIdx.pm | 2 +-
lib/PublicInbox/Search.pm | 77 ++-
lib/PublicInbox/SearchIdx.pm | 88 ++--
lib/PublicInbox/SearchIdxShard.pm | 7 +-
lib/PublicInbox/Sigfd.pm | 10 +-
lib/PublicInbox/Spawn.pm | 6 +-
lib/PublicInbox/SpawnPP.pm | 2 +-
lib/PublicInbox/TestCommon.pm | 47 +-
lib/PublicInbox/V2Writable.pm | 26 +-
lib/PublicInbox/Watch.pm | 2 +-
script/public-inbox-cindex | 86 +++
script/public-inbox-convert | 2 +-
t/cindex.t | 134 +++++
t/dir_idle.t | 6 +-
t/ds-leak.t | 8 +-
t/imapd.t | 6 +-
t/nntpd.t | 2 +-
t/sigfd.t | 7 +-
t/watch_maildir.t | 8 +-
xt/mem-imapd-tls.t | 7 +-
xt/mem-nntpd-tls.t | 8 +-
xt/net_writer-imap.t | 4 +-
33 files changed, 1424 insertions(+), 188 deletions(-)
create mode 100644 lib/PublicInbox/CodeSearch.pm
create mode 100644 lib/PublicInbox/CodeSearchIdx.pm
create mode 100755 script/public-inbox-cindex
create mode 100644 t/cindex.t
^ permalink raw reply [relevance 7%]
Results 1-2 of 2 | reverse | options above
-- pct% links below jump to the message on this page, permalinks otherwise --
2023-03-21 23:07 7% [PATCH 00/28] cindex coderepo commit indexer Eric Wong
2023-03-21 23:07 ` [PATCH 01/28] ipc: move nproc_shards from v2writable Eric Wong
2023-03-21 23:07 4% ` [PATCH 25/28] cindex: implement reindex Eric Wong
Code repositories for project(s) associated with this public inbox
https://80x24.org/public-inbox.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).