* [PATCH 07/14] lei_dedupe+shared_kv: ensure round-tripping serialization
2021-01-14 7:06 7% [PATCH 00/14] lei: another pile of changes Eric Wong
@ 2021-01-14 7:06 6% ` Eric Wong
0 siblings, 0 replies; 2+ results
From: Eric Wong @ 2021-01-14 7:06 UTC (permalink / raw)
To: meta
We'll be passing these objects via PublicInbox::IPC which uses
Storable (or Sereal), so ensure they're safe to use after
serialization.
---
lib/PublicInbox/LeiDedupe.pm | 29 ++++++++++++++++-------------
lib/PublicInbox/SharedKV.pm | 12 +++++++++---
t/lei_dedupe.t | 13 +++++++++++++
3 files changed, 38 insertions(+), 16 deletions(-)
diff --git a/lib/PublicInbox/LeiDedupe.pm b/lib/PublicInbox/LeiDedupe.pm
index 58eee533..81754361 100644
--- a/lib/PublicInbox/LeiDedupe.pm
+++ b/lib/PublicInbox/LeiDedupe.pm
@@ -43,9 +43,9 @@ sub smsg_hash ($) {
}
# the paranoid option
-sub dedupe_oid () {
- my $skv = PublicInbox::SharedKV->new;
- ($skv, sub { # may be called in a child process
+sub dedupe_oid ($) {
+ my ($skv) = @_;
+ (sub { # may be called in a child process
my ($eml, $oid) = @_;
$skv->set_maybe(_oidbin($oid) // _regen_oid($eml), '');
}, sub {
@@ -55,9 +55,9 @@ sub dedupe_oid () {
}
# dangerous if there's duplicate messages with different Message-IDs
-sub dedupe_mid () {
- my $skv = PublicInbox::SharedKV->new;
- ($skv, sub { # may be called in a child process
+sub dedupe_mid ($) {
+ my ($skv) = @_;
+ (sub { # may be called in a child process
my ($eml, $oid) = @_;
# TODO: lei will support non-public messages w/o Message-ID
my $mid = $eml->header_raw('Message-ID') // _oidbin($oid) //
@@ -73,9 +73,9 @@ sub dedupe_mid () {
}
# our default deduplication strategy (used by v2, also)
-sub dedupe_content () {
- my $skv = PublicInbox::SharedKV->new;
- ($skv, sub { # may be called in a child process
+sub dedupe_content ($) {
+ my ($skv) = @_;
+ (sub { # may be called in a child process
my ($eml) = @_; # oid = $_[1], ignored
$skv->set_maybe(content_hash($eml), '');
}, sub {
@@ -86,7 +86,7 @@ sub dedupe_content () {
# no deduplication at all
sub true { 1 }
-sub dedupe_none () { (undef, \&true, \&true) }
+sub dedupe_none ($) { (\&true, \&true) }
sub new {
my ($cls, $lei, $dst) = @_;
@@ -94,10 +94,12 @@ sub new {
# allow "none" to bypass Eml->new if writing to directory:
return if ($dd eq 'none' && substr($dst // '', -1) eq '/');
+ my $m = "dedupe_$dd";
+ $cls->can($m) or die "unsupported dedupe strategy: $dd\n";
+ my $skv = $dd eq 'none' ? undef : PublicInbox::SharedKV->new;
- my $dd_new = $cls->can("dedupe_$dd") //
- die "unsupported dedupe strategy: $dd\n";
- bless [ $dd_new->() ], $cls; # [ $skv, $cb ]
+ # [ $skv, $eml_cb, $smsg_cb, "dedupe_$dd" ]
+ bless [ $skv, undef, undef, $m ], $cls;
}
# returns true on unseen messages according to the deduplication strategy,
@@ -115,6 +117,7 @@ sub is_smsg_dup {
sub prepare_dedupe {
my ($self) = @_;
my $skv = $self->[0];
+ $self->[1] or @$self[1,2] = $self->can($self->[3])->($skv);
$skv ? $skv->dbh : undef;
}
diff --git a/lib/PublicInbox/SharedKV.pm b/lib/PublicInbox/SharedKV.pm
index d75d8998..072c94ca 100644
--- a/lib/PublicInbox/SharedKV.pm
+++ b/lib/PublicInbox/SharedKV.pm
@@ -8,9 +8,10 @@ package PublicInbox::SharedKV;
use strict;
use v5.10.1;
use parent qw(PublicInbox::Lock);
-use File::Temp 0.19 (); # 0.19 for ->newdir
+use File::Temp qw(tempdir);
use DBI ();
use PublicInbox::Spawn;
+use File::Path qw(rmtree);
sub dbh {
my ($self, $lock) = @_;
@@ -44,8 +45,8 @@ sub new {
my ($cls, $dir, $base, $opt) = @_;
my $self = bless { opt => $opt }, $cls;
unless (defined $dir) {
- $self->{tmp} = File::Temp->newdir('kv-XXXXXX', TMPDIR => 1);
- $dir = $self->{tmp}->dirname;
+ $self->{tmpdir} = $dir = tempdir('skv-XXXXXX', TMPDIR => 1);
+ $self->{tmpid} = "$$.$self";
}
-d $dir or mkdir($dir) or die "mkdir($dir): $!";
$base //= '';
@@ -145,4 +146,9 @@ SELECT COUNT(k) FROM kv
$sth->fetchrow_array;
}
+sub DESTROY {
+ my ($self) = @_;
+ rmtree($self->{tmpdir}) if ($self->{tmpid} // '') eq "$$.$self";
+}
+
1;
diff --git a/t/lei_dedupe.t b/t/lei_dedupe.t
index 6e971b9b..bcb06a0a 100644
--- a/t/lei_dedupe.t
+++ b/t/lei_dedupe.t
@@ -17,8 +17,18 @@ my $smsg = bless { ds => time }, 'PublicInbox::Smsg';
$smsg->populate($eml);
$smsg->{$_} //= '' for (qw(to cc references)) ;
+my $check_storable = sub {
+ my ($x) = @_;
+ SKIP: {
+ require_mods('Storable', 1);
+ my $dup = Storable::thaw(Storable::freeze($x));
+ is_deeply($dup, $x, "$x->[3] round-trips through storable");
+ }
+};
+
my $lei = { opt => { dedupe => 'none' } };
my $dd = PublicInbox::LeiDedupe->new($lei);
+$check_storable->($dd);
$dd->prepare_dedupe;
ok(!$dd->is_dup($eml), '1st is_dup w/o dedupe');
ok(!$dd->is_dup($eml), '2nd is_dup w/o dedupe');
@@ -29,6 +39,7 @@ ok(!$dd->is_smsg_dup($smsg), 'smsg dedupe none 2');
for my $strat (undef, 'content') {
$lei->{opt}->{dedupe} = $strat;
$dd = PublicInbox::LeiDedupe->new($lei);
+ $check_storable->($dd);
$dd->prepare_dedupe;
my $desc = $strat // 'default';
ok(!$dd->is_dup($eml), "1st is_dup with $desc dedupe");
@@ -43,6 +54,7 @@ like($@, qr/unsupported.*bogus/, 'died on bogus strategy');
$lei->{opt}->{dedupe} = 'mid';
$dd = PublicInbox::LeiDedupe->new($lei);
+$check_storable->($dd);
$dd->prepare_dedupe;
ok(!$dd->is_dup($eml), '1st is_dup with mid dedupe');
ok($dd->is_dup($eml), '2nd seen with mid dedupe');
@@ -52,6 +64,7 @@ ok($dd->is_smsg_dup($smsg), 'smsg mid dedupe reject');
$lei->{opt}->{dedupe} = 'oid';
$dd = PublicInbox::LeiDedupe->new($lei);
+$check_storable->($dd);
$dd->prepare_dedupe;
# --augment won't have OIDs:
^ permalink raw reply related [relevance 6%]
* [PATCH 00/14] lei: another pile of changes
@ 2021-01-14 7:06 7% Eric Wong
2021-01-14 7:06 6% ` [PATCH 07/14] lei_dedupe+shared_kv: ensure round-tripping serialization Eric Wong
0 siblings, 1 reply; 2+ results
From: Eric Wong @ 2021-01-14 7:06 UTC (permalink / raw)
To: meta
PATCH 2/14 took forever to figure out; turns out I was hunting
an old bug in Perl :x (and led to PATCH 3/14, too)
We could probably go farther on 5/14 and eliminate the
need for @TO_CLOSE_ATFORK_CHILD completely, but my brain
was fried from 2/14 :x.
The "ts:" => "rt:" change is technically user-visible,
but "ts:" was never publicly documented so I doubt it
affects anybody. "rt:" (received time) may be documented
in the future.
Eric Wong (14):
cmd_ipc: support + test EINTR + EAGAIN, no FDs
lei: test SIGPIPE, stop xsearch workers on client abort
daemon+watch: fix localization of %SIG for non-signalfd users
lei: do not unlink socket path at exit
lei: reduce live FD references in wq child
lei: rely on localized $current_lei for warnings
lei_dedupe+shared_kv: ensure round-tripping serialization
lei q: reinstate smsg dedupe
search: rename "ts:" prefix to "rt:"
lei_overview: rename "references" to "refs"
lei: q: lock stdout on overview output
leixsearch: remove some commented out code
lei: remove temporary var on open
lei: pass FD to CWD via cmsg, use fchdir on server
MANIFEST | 2 +
lib/PublicInbox/CmdIPC4.pm | 6 +-
lib/PublicInbox/Daemon.pm | 4 +-
lib/PublicInbox/IMAPsearchqp.pm | 6 +-
lib/PublicInbox/IPC.pm | 45 +++-----
lib/PublicInbox/LEI.pm | 182 +++++++++++++++++---------------
lib/PublicInbox/LeiDedupe.pm | 29 ++---
lib/PublicInbox/LeiOverview.pm | 43 +++++++-
lib/PublicInbox/LeiQuery.pm | 27 ++---
lib/PublicInbox/LeiXSearch.pm | 60 +++++++----
lib/PublicInbox/Lock.pm | 2 +-
lib/PublicInbox/Search.pm | 2 +-
lib/PublicInbox/SharedKV.pm | 12 ++-
lib/PublicInbox/Spawn.pm | 13 ++-
script/lei | 88 +++++++++------
script/public-inbox-watch | 2 +-
t/cmd_ipc.t | 32 ++++++
t/imap_searchqp.t | 6 +-
t/lei.t | 33 +-----
t/lei_dedupe.t | 13 +++
t/lei_overview.t | 33 ++++++
xt/lei-sigpipe.t | 32 ++++++
22 files changed, 417 insertions(+), 255 deletions(-)
create mode 100644 t/lei_overview.t
create mode 100644 xt/lei-sigpipe.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 --
2021-01-14 7:06 7% [PATCH 00/14] lei: another pile of changes Eric Wong
2021-01-14 7:06 6% ` [PATCH 07/14] lei_dedupe+shared_kv: ensure round-tripping serialization 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).