user/dev discussion of public-inbox itself
 help / color / mirror / code / Atom feed
From: Eric Wong <e@80x24.org>
To: meta@public-inbox.org
Subject: [PATCH 07/36] lei_to_mail: start atomic and compressed mbox writing
Date: Thu, 31 Dec 2020 13:51:25 +0000	[thread overview]
Message-ID: <20201231135154.6070-8-e@80x24.org> (raw)
In-Reply-To: <20201231135154.6070-1-e@80x24.org>

We'll allow using multiple workers to write to a single
mbox (which could be compressed).  This is can be done
safely with O_APPEND + syswrite for uncompressed files,
and using a lock when piping to pigz/gzip/bzip2/xz.
---
 lib/PublicInbox/LeiToMail.pm   | 99 ++++++++++++++++++++++++++++++++--
 lib/PublicInbox/ProcessPipe.pm | 21 ++++++--
 t/lei_to_mail.t                | 47 ++++++++++++++++
 3 files changed, 158 insertions(+), 9 deletions(-)

diff --git a/lib/PublicInbox/LeiToMail.pm b/lib/PublicInbox/LeiToMail.pm
index b0d4b664..ebb50c50 100644
--- a/lib/PublicInbox/LeiToMail.pm
+++ b/lib/PublicInbox/LeiToMail.pm
@@ -6,6 +6,12 @@ package PublicInbox::LeiToMail;
 use strict;
 use v5.10.1;
 use PublicInbox::Eml;
+use PublicInbox::Lock;
+use PublicInbox::ProcessPipe;
+use PublicInbox::Spawn qw(which spawn);
+use Symbol qw(gensym);
+use File::Temp ();
+use IO::Handle; # ->autoflush
 
 my %kw2char = ( # Maildir characters
 	draft => 'D',
@@ -45,10 +51,14 @@ sub _mbox_hdr_buf ($$$) {
 	$buf;
 }
 
-sub write_in_full_atomic ($$) {
-	my ($fh, $buf) = @_;
-	defined(my $w = syswrite($fh, $$buf)) or die "write: $!";
-	$w == length($$buf) or die "short write: $w != ".length($$buf);
+sub write_in_full ($$$) {
+	my ($fh, $buf, $atomic) = @_;
+	if ($atomic) {
+		defined(my $w = syswrite($fh, $$buf)) or die "write: $!";
+		$w == length($$buf) or die "short write: $w != ".length($$buf);
+	} else {
+		print $fh $$buf or die "print: $!";
+	}
 }
 
 sub eml2mboxrd ($;$) {
@@ -106,4 +116,85 @@ sub eml2mboxcl2 {
 	$buf;
 }
 
+sub mkmaildir ($) {
+	my ($maildir) = @_;
+	for (qw(new tmp cur)) {
+		my $d = "$maildir/$_";
+		next if -d $d;
+		require File::Path;
+		if (!File::Path::mkpath($d) && !-d $d) {
+			die "failed to mkpath($d): $!\n";
+		}
+	}
+}
+
+sub git_to_mail { # git->cat_async callback
+	my ($bref, $oid, $type, $size, $arg) = @_;
+	if ($type ne 'blob') {
+		if ($type eq 'missing') {
+			warn "missing $oid\n";
+		} else {
+			warn "unexpected type=$type for $oid\n";
+		}
+	}
+	if ($size > 0) {
+		my ($write_cb, $kw) = @$arg;
+		$write_cb->($bref, $oid, $kw);
+	}
+}
+
+sub reap_compress { # dwaitpid callback
+	my ($lei, $pid) = @_;
+	my $cmd = delete $lei->{"pid.$pid"};
+	return if $? == 0;
+	$lei->fail("@$cmd failed", $? >> 8);
+}
+
+sub compress_dst {
+	my ($out, $sfx, $lei) = @_;
+	my $cmd = [];
+	if ($sfx eq 'gz') {
+		$cmd->[0] = which($lei->{env}->{GZIP} // 'pigz') //
+				which('gzip') //
+			die "pigz or gzip missing for $sfx\n";
+			# TODO: use IO::Compress::Gzip
+		push @$cmd, '-c'; # stdout
+		push @$cmd, '--rsyncable' if $lei->{opt}->{rsyncable};
+	} else {
+		die "TODO $sfx"
+	}
+	pipe(my ($r, $w)) or die "pipe: $!";
+	my $rdr = { 0 => $r, 1 => $out, 2 => $lei->{2} };
+	my $pid = spawn($cmd, $lei->{env}, $rdr);
+	$lei->{"pid.$pid"} = $cmd;
+	my $pp = gensym;
+	tie *$pp, 'PublicInbox::ProcessPipe', $pid, $w, \&reap_compress, $lei;
+	my $tmp = File::Temp->new("$sfx.lock-XXXXXX", TMPDIR => 1);
+	my $pipe_lk = ($lei->{opt}->{jobs} // 0) > 1 ? bless({
+		lock_path => $tmp->filename,
+		tmp => $tmp
+	}, 'PublicInbox::Lock') : undef;
+	($pp, $pipe_lk);
+}
+
+sub write_cb {
+	my ($cls, $dst, $lei) = @_;
+	if ($dst =~ s!\A(mbox(?:rd|cl|cl2|o))?:!!) {
+		my $m = "eml2$1";
+		my $eml2mbox = $cls->can($m) or die "$cls->$m missing";
+		my ($out, $pipe_lk);
+		open $out, '>>', $dst or die "open $dst: $!";
+		my $atomic = !!(($lei->{opt}->{jobs} // 0) > 1);
+		if ($dst =~ /\.(gz|bz2|xz)\z/) {
+			($out, $pipe_lk) = compress_dst($out, $1, $lei);
+		}
+		sub {
+			my ($buf, $oid, $kw) = @_;
+			$buf = $eml2mbox->(PublicInbox::Eml->new($buf), $kw);
+			my $lock = $pipe_lk->lock_for_scope if $pipe_lk;
+			write_in_full($out, $buf, $atomic);
+		}
+	}
+}
+
 1;
diff --git a/lib/PublicInbox/ProcessPipe.pm b/lib/PublicInbox/ProcessPipe.pm
index 2ce7eb8f..c9234f42 100644
--- a/lib/PublicInbox/ProcessPipe.pm
+++ b/lib/PublicInbox/ProcessPipe.pm
@@ -4,28 +4,39 @@
 # a tied handle for auto reaping of children tied to a pipe, see perltie(1)
 package PublicInbox::ProcessPipe;
 use strict;
-use warnings;
+use v5.10.1;
 
 sub TIEHANDLE {
-	my ($class, $pid, $fh) = @_;
-	bless { pid => $pid, fh => $fh }, $class;
+	my ($class, $pid, $fh, $cb, $arg) = @_;
+	bless { pid => $pid, fh => $fh, cb => $cb, arg => $arg }, $class;
 }
 
 sub READ { read($_[0]->{fh}, $_[1], $_[2], $_[3] || 0) }
 
 sub READLINE { readline($_[0]->{fh}) }
 
+sub WRITE {
+	use bytes qw(length);
+	syswrite($_[0]->{fh}, $_[1], $_[2] // length($_[1]), $_[3] // 0);
+}
+
+sub PRINT {
+	my $self = shift;
+	print { $self->{fh} } @_;
+}
+
 sub CLOSE {
 	my $fh = delete($_[0]->{fh});
 	my $ret = defined $fh ? close($fh) : '';
-	my $pid = delete $_[0]->{pid};
+	my ($pid, $cb, $arg) = delete @{$_[0]}{qw(pid cb arg)};
 	if (defined $pid) {
 		# PublicInbox::DS may not be loaded
-		eval { PublicInbox::DS::dwaitpid($pid, undef, undef) };
+		eval { PublicInbox::DS::dwaitpid($pid, $cb, $arg) };
 
 		if ($@) { # ok, not in the event loop, work synchronously
 			waitpid($pid, 0);
 			$ret = '' if $?;
+			$cb->($arg, $pid) if $cb;
 		}
 	}
 	$ret;
diff --git a/t/lei_to_mail.t b/t/lei_to_mail.t
index 089a422e..231cf543 100644
--- a/t/lei_to_mail.t
+++ b/t/lei_to_mail.t
@@ -62,4 +62,51 @@ for my $mbox (qw(mboxrd mboxo mboxcl mboxcl2)) {
 	}
 }
 
+my ($tmpdir, $for_destroy) = tmpdir();
+local $ENV{TMPDIR} = $tmpdir;
+open my $err, '>>', "$tmpdir/lei.err" or BAIL_OUT $!;
+my $lei = { 2 => $err };
+my $buf = <<'EOM';
+From: x@example.com
+Subject: x
+
+blah
+EOM
+my $fn = "$tmpdir/x.mbox";
+my $orig = do {
+	my $wcb = PublicInbox::LeiToMail->write_cb("mboxcl2:$fn", $lei);
+	is(ref $wcb, 'CODE', 'write_cb returned callback');
+	ok(-f $fn && !-s _, 'empty file created');
+	$wcb->(\(my $dup = $buf), 'deadbeef', [ qw(seen) ]);
+	undef $wcb;
+	open my $fh, '<', $fn or BAIL_OUT $!;
+	my $raw = do { local $/; <$fh> };
+	like($raw, qr/^blah\n/sm, 'wrote content');
+	unlink $fn or BAIL_OUT $!;
+
+	local $lei->{opt} = { jobs => 2 };
+	$wcb = PublicInbox::LeiToMail->write_cb("mboxcl2:$fn", $lei);
+	$wcb->(\($dup = $buf), 'deadbeef', [ qw(seen) ]);
+	undef $wcb;
+	open $fh, '<', $fn or BAIL_OUT $!;
+	is($raw, do { local $/; <$fh> }, 'jobs > 1');
+	$raw;
+};
+SKIP: {
+	use PublicInbox::Spawn qw(which);
+	my $gzip = which('gzip') or skip 'gzip not found', 1;
+	my $wcb = PublicInbox::LeiToMail->write_cb("mboxcl2:$fn.gz", $lei);
+	$wcb->(\(my $dup = $buf), 'deadbeef', [ qw(seen) ]);
+	undef $wcb;
+	my $uncompressed = xqx([$gzip, '-dc', "$fn.gz"]);
+	is($uncompressed, $orig, 'gzip works');
+
+	local $lei->{opt} = { jobs => 2 };
+	unlink "$fn.gz" or die "unlink $!";
+	$wcb = PublicInbox::LeiToMail->write_cb("mboxcl2:$fn.gz", $lei);
+	$wcb->(\(my $dupe = $buf), 'deadbeef', [ qw(seen) ]);
+	undef $wcb;
+	is(xqx([$gzip, '-dc', "$fn.gz"]), $orig);
+}
+
 done_testing;

  parent reply	other threads:[~2020-12-31 13:51 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-31 13:51 [PATCH 00/36] another round of lei stuff Eric Wong
2020-12-31 13:51 ` [PATCH 01/36] import: respect init.defaultBranch Eric Wong
2020-12-31 13:51 ` [PATCH 02/36] lei_store: use per-machine refname as git HEAD Eric Wong
2020-12-31 13:51 ` [PATCH 03/36] revert "lei_store: use per-machine refname as git HEAD" Eric Wong
2020-12-31 13:51 ` [PATCH 04/36] lei_to_mail: initial implementation for writing mbox formats Eric Wong
2020-12-31 13:51 ` [PATCH 05/36] sharedkv: fork()-friendly key-value store Eric Wong
2020-12-31 13:51 ` [PATCH 06/36] sharedkv: split out index_values Eric Wong
2020-12-31 13:51 ` Eric Wong [this message]
2020-12-31 13:51 ` [PATCH 08/36] mboxreader: new class for reading various mbox formats Eric Wong
2020-12-31 13:51 ` [PATCH 09/36] lei_to_mail: start --augment, dedupe, bz2 and xz Eric Wong
2020-12-31 13:51 ` [PATCH 10/36] lei: implement various deduplication strategies Eric Wong
2020-12-31 13:51 ` [PATCH 11/36] lei_to_mail: lazy-require LeiDedupe Eric Wong
2020-12-31 13:51 ` [PATCH 12/36] lei_to_mail: support for non-seekable outputs Eric Wong
2020-12-31 13:51 ` [PATCH 13/36] lei_to_mail: support Maildir, fix+test --augment Eric Wong
2020-12-31 13:51 ` [PATCH 14/36] ipc: generic IPC dispatch based on Storable Eric Wong
2020-12-31 13:51 ` [PATCH 15/36] ipc: support Sereal Eric Wong
2020-12-31 13:51 ` [PATCH 16/36] lei_store: add ->set_eml, ->add_eml can return smsg Eric Wong
2020-12-31 13:51 ` [PATCH 17/36] lei: rename "extinbox" => "external" Eric Wong
2020-12-31 13:51 ` [PATCH 18/36] mid: use defined-or with `push' for uniqueness check Eric Wong
2020-12-31 13:51 ` [PATCH 19/36] mid: hoist out mids_in sub Eric Wong
2020-12-31 13:51 ` [PATCH 20/36] lei_store: handle messages without Message-ID at all Eric Wong
2020-12-31 13:51 ` [PATCH 21/36] ipc: use shutdown(2), base atfork* callback Eric Wong
2020-12-31 13:51 ` [PATCH 22/36] lei_to_mail: unlink mboxes if not augmenting Eric Wong
2020-12-31 13:51 ` [PATCH 23/36] lei: add --mfolder as an --output alias Eric Wong
2020-12-31 13:51 ` [PATCH 24/36] spawn: move run_die here from PublicInbox::Import Eric Wong
2020-12-31 13:51 ` [PATCH 25/36] init: remove embedded UnlinkMe package Eric Wong
2020-12-31 13:51 ` [PATCH 26/36] t/run: avoid uninitialized var on incomplete test Eric Wong
2020-12-31 13:51 ` [PATCH 27/36] gcf2client: reap process on DESTROY Eric Wong
2020-12-31 13:51 ` [PATCH 28/36] lei_to_mail: open FIFOs O_WRONLY so we block Eric Wong
2020-12-31 13:51 ` [PATCH 29/36] searchidxshard: call DS->Reset at worker start Eric Wong
2020-12-31 13:51 ` [PATCH 30/36] t/ipc.t: test for references via `die' Eric Wong
2020-12-31 13:51 ` [PATCH 31/36] use PublicInbox::DS for dwaitpid Eric Wong
2020-12-31 13:51 ` [PATCH 32/36] syscall: SFD_NONBLOCK can be a constant, again Eric Wong
2020-12-31 13:51 ` [PATCH 33/36] lei: avoid Spawn package when starting daemon Eric Wong
2020-12-31 13:51 ` [PATCH 34/36] avoid calling waitpid from children in DESTROY Eric Wong
2020-12-31 13:51 ` [PATCH 35/36] ds: clobber $in_loop first at reset Eric Wong
2020-12-31 13:51 ` [PATCH 36/36] on_destroy: support PID owner guard 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: http://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=20201231135154.6070-8-e@80x24.org \
    --to=e@80x24.org \
    --cc=meta@public-inbox.org \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).