With 11 epochs on LKML, it's the 11 current epoch time goes from
around 60ms to around 10ms, so it's a significant improvement.
And improve test coverage while we're at it.
---
lib/PublicInbox/ManifestJsGz.pm | 14 +++++++-------
t/extindex-psgi.t | 20 ++++++++++++++++++--
2 files changed, 25 insertions(+), 9 deletions(-)
diff --git a/lib/PublicInbox/ManifestJsGz.pm b/lib/PublicInbox/ManifestJsGz.pm
index cde60245..d5048a96 100644
--- a/lib/PublicInbox/ManifestJsGz.pm
+++ b/lib/PublicInbox/ManifestJsGz.pm
@@ -53,6 +53,7 @@ sub slow_manifest_add ($$) {
manifest_add($ctx, $ibx);
}
};
+ warn "E: $@" if $@;
}
sub eidx_manifest_add ($$$) {
@@ -65,6 +66,10 @@ sub eidx_manifest_add ($$$) {
}
} else {
warn "E: `${\$ibx->eidx_key}' not indexed by $ALL->{topdir}\n";
+ # do not use slow path for global manifest since
+ # it can become catastrophically slow. per-inbox manifest
+ # is not too bad with dozens of epochs, so never fail that:
+ slow_manifest_add($ctx, $ibx) if $ibx == $ctx->{ibx};
}
}
@@ -85,12 +90,8 @@ sub response {
sub ibx_entry {
my ($ctx, $ibx) = @_;
my $ALL = $ctx->{www}->{pi_cfg}->ALL;
- if ($ALL) { # FIXME: test this in t/
- eidx_manifest_add($ctx, $ALL, $ibx);
- } else {
+ $ALL ? eidx_manifest_add($ctx, $ALL, $ibx) :
slow_manifest_add($ctx, $ibx);
- warn "E: $@" if $@;
- }
}
sub hide_key { 'manifest' } # for WwwListing->list_match_i
@@ -112,8 +113,7 @@ sub psgi_triple {
sub per_inbox {
my ($ctx) = @_;
- # only one inbox, slow is probably OK
- slow_manifest_add($ctx, $ctx->{ibx});
+ ibx_entry($ctx, $ctx->{ibx});
psgi_triple($ctx);
}
diff --git a/t/extindex-psgi.t b/t/extindex-psgi.t
index 4e26962e..98dc2e48 100644
--- a/t/extindex-psgi.t
+++ b/t/extindex-psgi.t
@@ -11,6 +11,7 @@ require_git(2.6);
require_mods(qw(json DBD::SQLite Search::Xapian
HTTP::Request::Common Plack::Test URI::Escape Plack::Builder));
use_ok($_) for (qw(HTTP::Request::Common Plack::Test));
+use IO::Uncompress::Gunzip qw(gunzip);
require PublicInbox::WWW;
my ($ro_home, $cfg_path) = setup_public_inboxes;
my ($tmpdir, $for_destroy) = tmpdir;
@@ -18,11 +19,11 @@ my $home = "$tmpdir/home";
mkdir $home or BAIL_OUT $!;
mkdir "$home/.public-inbox" or BAIL_OUT $!;
my $pi_config = "$home/.public-inbox/config";
-cp("$ro_home/.public-inbox/config", $pi_config) or BAIL_OUT;
+cp($cfg_path, $pi_config) or BAIL_OUT;
my $env = { HOME => $home };
run_script([qw(-extindex --all), "$tmpdir/eidx"], $env) or BAIL_OUT;
{
- open my $cfgfh, '>', $pi_config or BAIL_OUT;
+ open my $cfgfh, '>>', $pi_config or BAIL_OUT;
$cfgfh->autoflush(1);
print $cfgfh <<EOM or BAIL_OUT;
[extindex "all"]
@@ -30,6 +31,7 @@ run_script([qw(-extindex --all), "$tmpdir/eidx"], $env) or BAIL_OUT;
url = http://bogus.example.com/all
[publicinbox]
wwwlisting = all
+ grokManifest = all
EOM
}
my $www = PublicInbox::WWW->new(PublicInbox::Config->new($pi_config));
@@ -67,6 +69,20 @@ my $client = sub {
is($res->code, 404, 'no inboxes matched');
unlike($res->content, qr!no inboxes, yet!,
'we have inboxes, just no matches');
+
+ my $m = {};
+ for my $pfx (qw(/t1 /t2), '') {
+ $res = $cb->(GET($pfx.'/manifest.js.gz'));
+ gunzip(\($res->content) => \(my $js));
+ $m->{$pfx} = json_utf8->decode($js);
+ }
+ is_deeply([sort keys %{$m->{''}}],
+ [ sort(keys %{$m->{'/t1'}}, keys %{$m->{'/t2'}}) ],
+ 't1 + t2 = all');
+ is_deeply([ sort keys %{$m->{'/t2'}} ], [ '/t2/git/0.git' ],
+ 't2 manifest');
+ is_deeply([ sort keys %{$m->{'/t1'}} ], [ '/t1' ],
+ 't2 manifest');
};
test_psgi(sub { $www->call(@_) }, $client);
%$env = (%$env, TMPDIR => $tmpdir, PI_CONFIG => $pi_config);