user/dev discussion of public-inbox itself
 help / color / mirror / code / Atom feed
From: Eric Wong <e@yhbt.net>
To: meta@public-inbox.org
Subject: [PATCH 64/82] searchidx: v1 (re)-index uses git asynchronously
Date: Wed, 10 Jun 2020 07:05:01 +0000	[thread overview]
Message-ID: <20200610070519.18252-65-e@yhbt.net> (raw)
In-Reply-To: <20200610070519.18252-1-e@yhbt.net>

We can cleanup some of our v1 code slightly and let git do
I/O+decoding in parallel.  This gives a slight 2-4%
re-indexing performance boost even on an SSD.
---
 lib/PublicInbox/SearchIdx.pm | 76 +++++++++++++++++-------------------
 1 file changed, 35 insertions(+), 41 deletions(-)

diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm
index 3df7970ebf9..a790ac4076a 100644
--- a/lib/PublicInbox/SearchIdx.pm
+++ b/lib/PublicInbox/SearchIdx.pm
@@ -378,7 +378,7 @@ sub _msgmap_init ($) {
 
 sub add_message {
 	# mime = PublicInbox::Eml or Email::MIME object
-	my ($self, $mime, $smsg) = @_;
+	my ($self, $mime, $smsg, $sync) = @_;
 	my $hdr = $mime->header_obj;
 	my $mids = mids_for_index($hdr);
 	$smsg //= bless { blob => '' }, 'PublicInbox::Smsg'; # test-only compat
@@ -389,7 +389,7 @@ sub add_message {
 	};
 
 	# v1 and tests only:
-	$smsg->populate($hdr, $self);
+	$smsg->populate($hdr, $sync);
 	$smsg->{bytes} //= length($mime->as_string);
 
 	eval {
@@ -549,24 +549,24 @@ sub unindex_mm {
 	$self->{mm}->mid_delete(mid_mime($mime));
 }
 
-sub index_both {
-	my ($self, $mime, $smsg) = @_;
-	my $num = index_mm($self, $mime);
+sub index_both { # git->cat_async callback
+	my ($bref, $oid, $type, $size, $sync) = @_;
+	my ($nr, $max) = @$sync{qw(nr max)};
+	++$$nr;
+	$$max -= $size;
+	my $smsg = bless { bytes => $size, blob => $oid }, 'PublicInbox::Smsg';
+	my $self = $sync->{sidx};
+	my $eml = PublicInbox::Eml->new($bref);
+	my $num = index_mm($self, $eml);
 	$smsg->{num} = $num;
-	add_message($self, $mime, $smsg);
+	add_message($self, $eml, $smsg, $sync);
 }
 
-sub unindex_both {
-	my ($self, $mime) = @_;
-	unindex_blob($self, $mime);
-	unindex_mm($self, $mime);
-}
-
-sub do_cat_mail {
-	my ($git, $blob, $sizeref) = @_;
-	my $str = $git->cat_file($blob, $sizeref) or
-		die "BUG: $blob not found in $git->{git_dir}";
-	PublicInbox::Eml->new($str);
+sub unindex_both { # git->cat_async callback
+	my ($bref, $oid, $type, $size, $self) = @_;
+	my $eml = PublicInbox::Eml->new($bref);
+	unindex_blob($self, $eml);
+	unindex_mm($self, $eml);
 }
 
 # called by public-inbox-index
@@ -576,15 +576,6 @@ sub index_sync {
 	$self->{-inbox}->with_umask(sub { $self->_index_sync($opts) })
 }
 
-sub batch_adjust ($$$$$) {
-	my ($max, $bytes, $batch_cb, $latest, $nr) = @_;
-	$$max -= $bytes;
-	if ($$max <= 0) {
-		$$max = $BATCH_BYTES;
-		$batch_cb->($nr, $latest);
-	}
-}
-
 sub too_big ($$$) {
 	my ($self, $git, $oid) = @_;
 	my $max_size = $self->{index_max_size} or return;
@@ -597,24 +588,28 @@ sub too_big ($$$) {
 
 # only for v1
 sub read_log {
-	my ($self, $log, $add_cb, $del_cb, $batch_cb) = @_;
+	my ($self, $log, $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;
 	my $nr = 0;
+	my $sync = { sidx => $self, nr => \$nr, max => \$max };
 	while (defined($line = <$log>)) {
 		if ($line =~ /$addmsg/o) {
 			my $blob = $1;
 			if (delete $D{$blob}) {
+				# make sure pending index writes are done
+				# before writing to ->mm
+				$git->cat_async_wait;
+
 				if (defined $self->{regen_down}) {
 					my $num = $self->{regen_down}--;
 					$self->{mm}->num_highwater($num);
@@ -622,12 +617,12 @@ sub read_log {
 				next;
 			}
 			next if too_big($self, $git, $blob);
-			my $mime = do_cat_mail($git, $blob, \$bytes);
-			my $smsg = bless {}, 'PublicInbox::Smsg';
-			batch_adjust(\$max, $bytes, $batch_cb, $latest, ++$nr);
-			$smsg->{blob} = $blob;
-			$smsg->{bytes} = $bytes;
-			$add_cb->($self, $mime, $smsg);
+			$git->cat_async($blob, \&index_both, { %$sync });
+			if ($max <= 0) {
+				$git->cat_async_wait;
+				$max = $BATCH_BYTES;
+				$batch_cb->($nr, $latest);
+			}
 		} elsif ($line =~ /$delmsg/o) {
 			my $blob = $1;
 			$D{$blob} = 1 unless too_big($self, $git, $blob);
@@ -635,18 +630,17 @@ sub read_log {
 			$latest = $1;
 			$newest ||= $latest;
 		} elsif ($line =~ /^author .*? ([0-9]+) [\-\+][0-9]+$/) {
-			$self->{autime} = $1;
+			$sync->{autime} = $1;
 		} elsif ($line =~ /^committer .*? ([0-9]+) [\-\+][0-9]+$/) {
-			$self->{cotime} = $1;
+			$sync->{cotime} = $1;
 		}
 	}
 	close($log) or die "git log failed: \$?=$?";
 	# get the leftovers
 	foreach my $blob (keys %D) {
-		my $mime = do_cat_mail($git, $blob, \$bytes);
-		$del_cb->($self, $mime);
+		$git->cat_async($blob, \&unindex_both, $self);
 	}
-	delete @$self{qw(autime cotime)};
+	$git->cat_async_wait;
 	$batch_cb->($nr, $latest, $newest);
 }
 
@@ -774,7 +768,7 @@ sub _index_sync {
 	} while (_last_x_commit($self, $mm) ne $last_commit);
 
 	my $dbh = $mm->{dbh} if $mm;
-	my $cb = sub {
+	my $batch_cb = sub {
 		my ($nr, $commit, $newest) = @_;
 		if ($dbh) {
 			if ($newest) {
@@ -803,7 +797,7 @@ sub _index_sync {
 	};
 
 	$dbh->begin_work;
-	read_log($self, $xlog, *index_both, *unindex_both, $cb);
+	read_log($self, $xlog, $batch_cb);
 }
 
 sub DESTROY {

  parent reply	other threads:[~2020-06-10  7:07 UTC|newest]

Thread overview: 84+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-06-10  7:03 [PATCH 00/82] public-inbox-imapd: read-only IMAP server Eric Wong
2020-06-10  7:03 ` [PATCH 01/82] doc: add some IMAP standards Eric Wong
2020-06-10  7:03 ` [PATCH 02/82] nntpd: restrict allowed newsgroup names Eric Wong
2020-06-10  7:04 ` [PATCH 03/82] preliminary imap server implementation Eric Wong
2020-06-10  7:04 ` [PATCH 04/82] inboxidle: new class to detect inbox changes Eric Wong
2020-06-10  7:04 ` [PATCH 05/82] imap: support IDLE Eric Wong
2020-06-10  7:04 ` [PATCH 06/82] msgmap: split ->max into its own method Eric Wong
2020-06-10  7:04 ` [PATCH 07/82] imap: delay InboxIdle start, support refresh Eric Wong
2020-06-10  7:04 ` [PATCH 08/82] imap: implement STATUS command Eric Wong
2020-06-10  7:04 ` [PATCH 09/82] imap: use Text::ParseWords::parse_line to handle quoted words Eric Wong
2020-06-10  7:04 ` [PATCH 10/82] imap: support LIST command Eric Wong
2020-06-10  7:04 ` [PATCH 11/82] t/imapd: support FakeInotify and KQNotify Eric Wong
2020-06-10  7:04 ` [PATCH 12/82] imap: support fetch for BODYSTRUCTURE and BODY Eric Wong
2020-06-10  7:04 ` [PATCH 13/82] eml: each_part: single part $idx is 1 Eric Wong
2020-06-10  7:04 ` [PATCH 14/82] imap: allow fetch of partial of BODY[...] and headers Eric Wong
2020-06-10  7:04 ` [PATCH 15/82] imap: always include `resp-text' in responses Eric Wong
2020-06-10  7:04 ` [PATCH 16/82] imap: split out unit tests and benchmarks Eric Wong
2020-06-10  7:04 ` [PATCH 17/82] imap: fix multi-message partial header fetches Eric Wong
2020-06-10  7:04 ` [PATCH 18/82] imap: simplify partial fetch structure Eric Wong
2020-06-10  7:04 ` [PATCH 19/82] imap: support sequence number FETCH Eric Wong
2020-06-10  7:04 ` [PATCH 20/82] imap: do not include ".PEEK" in responses Eric Wong
2020-06-10  7:04 ` [PATCH 21/82] imap: support the CLOSE command Eric Wong
2020-06-10  7:04 ` [PATCH 22/82] imap: speed up HEADER.FIELDS[.NOT] range fetches Eric Wong
2020-06-10  7:04 ` [PATCH 23/82] git: async: flatten the inflight array Eric Wong
2020-06-10  7:04 ` [PATCH 24/82] git: do our own read buffering for cat-file Eric Wong
2020-06-10  7:04 ` [PATCH 25/82] imap: use git-cat-file asynchronously Eric Wong
2020-06-10  7:04 ` [PATCH 26/82] git: idle rbuf for async Eric Wong
2020-06-10  7:04 ` [PATCH 27/82] imap: support LSUB command Eric Wong
2020-06-10  7:04 ` [PATCH 28/82] imap: FETCH: support comma-delimited ranges Eric Wong
2020-06-10  7:04 ` [PATCH 29/82] add imapd compression test Eric Wong
2020-06-10  7:04 ` [PATCH 30/82] testcommon: tcp_(server|connect): BAIL_OUT on failure Eric Wong
2020-06-10  7:04 ` [PATCH 31/82] *deflate: drop invalid comment about rbuf Eric Wong
2020-06-10  7:04 ` [PATCH 32/82] imap: fix pipelining with async git Eric Wong
2020-06-10  7:04 ` [PATCH 33/82] git: cat_async: provide requested OID + "missing" on missing blobs Eric Wong
2020-06-10  7:04 ` [PATCH 34/82] git: move async_cat reference to PublicInbox::Git Eric Wong
2020-06-10  7:04 ` [PATCH 35/82] git: async: automatic retry on alternates change Eric Wong
2020-06-10  7:04 ` [PATCH 36/82] imapclient: wrapper for Mail::IMAPClient Eric Wong
2020-06-10  7:04 ` [PATCH 37/82] xt: add imapd-validate and imapd-mbsync-oimap Eric Wong
2020-06-10  7:04 ` [PATCH 38/82] imap: support out-of-bounds ranges Eric Wong
2020-06-10  7:04 ` [PATCH 39/82] xt/perf-imap-list: time refresh_inboxlist Eric Wong
2020-06-10  7:04 ` [PATCH 40/82] imap: case-insensitive mailbox name comparisons Eric Wong
2020-06-10  7:04 ` [PATCH 41/82] imap: break giant inboxes into sub-inboxes of 50K messages Eric Wong
2020-06-10  7:04 ` [PATCH 42/82] imap: start doing iterative config reloading Eric Wong
2020-06-10  7:04 ` [PATCH 43/82] imap: require ".$UID_MIN-$UID_END" suffix Eric Wong
2020-06-10  7:04 ` [PATCH 44/82] imapd: ensure LIST is sorted alphabetically, for now Eric Wong
2020-06-10  7:04 ` [PATCH 45/82] imap: omit $UID_END from mailbox name, use index Eric Wong
2020-06-10  7:04 ` [PATCH 46/82] t/config.t: always compare against git bool behavior Eric Wong
2020-06-10  7:04 ` [PATCH 47/82] xt/*: show some tunable parameters Eric Wong
2020-06-10  7:04 ` [PATCH 48/82] imap: STATUS and LIST are case-insensitive, too Eric Wong
2020-06-10  7:04 ` [PATCH 49/82] imap: EXAMINE/STATUS: return correct counts Eric Wong
2020-06-10  7:04 ` [PATCH 50/82] imap: avoid uninitialized warnings on incomplete commands Eric Wong
2020-06-10  7:04 ` [PATCH 51/82] imap: start parsing out queries for SQLite and Xapian Eric Wong
2020-06-10  7:04 ` [PATCH 52/82] imap: SEARCH: clamp results to the 50K UID range Eric Wong
2020-06-10  7:04 ` [PATCH 53/82] imap: allow UID range search on timestamps Eric Wong
2020-06-10  7:04 ` [PATCH 54/82] over: get_art: use dbh->prepare_cached Eric Wong
2020-06-10  7:04 ` [PATCH 55/82] search: index byte size of a message for IMAP search Eric Wong
2020-06-10  7:04 ` [PATCH 56/82] search: index UID for IMAP search, too Eric Wong
2020-06-10  7:04 ` [PATCH 57/82] imap: remove dummies from sequence number FETCH Eric Wong
2020-06-10  7:04 ` [PATCH 58/82] imap: compile UID FETCH to opcodes Eric Wong
2020-06-10  7:04 ` [PATCH 59/82] imap: UID FETCH: optimize for smsg-only case Eric Wong
2020-06-10  7:04 ` [PATCH 60/82] imap: UID FETCH: optimize (UID FLAGS) harder Eric Wong
2020-06-10  7:04 ` [PATCH 61/82] imap: IDLE: avoid extraneous wakeups, keep-alive Eric Wong
2020-06-10  7:04 ` [PATCH 62/82] imap: 30 minute auto-logout timer Eric Wong
2020-06-10  7:05 ` [PATCH 63/82] imap: split ->logged_in attribute into a separate class Eric Wong
2020-06-10  7:05 ` Eric Wong [this message]
2020-06-10  7:05 ` [PATCH 65/82] index: account for CRLF conversion when storing bytes Eric Wong
2020-06-10  7:05 ` [PATCH 66/82] imap: rely on smsg->{bytes} for RFC822.SIZE Eric Wong
2020-06-10  7:05 ` [PATCH 67/82] imap: UID FETCH requires at least one data item Eric Wong
2020-06-10  7:05 ` [PATCH 68/82] imap: LIST shows "INBOX" in all caps Eric Wong
2020-06-10  7:05 ` [PATCH 69/82] imap: support 8000 octet lines Eric Wong
2020-06-10  7:05 ` [PATCH 70/82] imap: reinstate some message sequence number support Eric Wong
2020-06-10  7:05 ` [PATCH 71/82] imap: cleanup ->{uid_base} usage Eric Wong
2020-06-10  7:05 ` [PATCH 72/82] imap: FETCH: more granular CRLF conversion Eric Wong
2020-06-10  7:05 ` [PATCH 73/82] imap: further speed up HEADER.FIELDS FETCH requests Eric Wong
2020-06-10  7:05 ` [PATCH 74/82] imap: FETCH: try to make fake MSNs sequentially Eric Wong
2020-06-10  7:05 ` [PATCH 75/82] imap: STATUS/EXAMINE: rely on SQLite overview Eric Wong
2020-06-10  7:05 ` [PATCH 76/82] imap: UID SEARCH: support multiple ranges Eric Wong
2020-06-10  7:05 ` [PATCH 77/82] imap: wire up Xapian, MSN SEARCH and multi sequence-sets Eric Wong
2020-06-10  7:05 ` [PATCH 78/82] imap: misc cleanups and notes Eric Wong
2020-06-10  7:05 ` [PATCH 79/82] imapd: don't bother sorting LIST output Eric Wong
2020-06-10  7:05 ` [PATCH 80/82] imap: remove non-UID SEARCH for now Eric Wong
2020-06-10  7:05 ` [PATCH 81/82] over: uid_range: remove LIMIT Eric Wong
2020-06-10  7:05 ` [PATCH 82/82] imap: FETCH: proper MSN => UID mapping for requests Eric Wong
2020-06-12 23:49 ` [PATCH 83/82] imap: introduce memory-efficient uo2m mapping Eric Wong

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://public-inbox.org/README

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200610070519.18252-65-e@yhbt.net \
    --to=e@yhbt.net \
    --cc=meta@public-inbox.org \
    --subject='Re: [PATCH 64/82] searchidx: v1 (re)-index uses git asynchronously' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Code repositories for project(s) associated with this 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).