[PATCH 2/3] lei: avoid SQLite COUNT() for dedupe

[Eric Wong <e@80x24.org>]

SQLite COUNT() is a slow operation that does a full table scan
with no conditions.  There's no need for it, since lei dedupe
only needs to know if it's empty or not to decide between
new/ and cur/ for Maildir outputs.
---
 lib/PublicInbox/LeiDedupe.pm      | 5 ++---
 lib/PublicInbox/LeiSavedSearch.pm | 8 ++++----
 lib/PublicInbox/LeiToMail.pm      | 4 ++--
 lib/PublicInbox/LeiXSearch.pm     | 2 +-
 lib/PublicInbox/SharedKV.pm       | 7 +++++++
 5 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/lib/PublicInbox/LeiDedupe.pm b/lib/PublicInbox/LeiDedupe.pm
index ed52e417..32f99cd0 100644
--- a/lib/PublicInbox/LeiDedupe.pm
+++ b/lib/PublicInbox/LeiDedupe.pm
@@ -127,10 +127,9 @@ sub pause_dedupe {
 	delete($skv->{dbh}) if $skv;
 }
 
-sub dedupe_nr {
+sub has_entries {
 	my $skv = $_[0]->[0] or return undef;
-	my @n = $skv->count;
-	$n[0];
+	$skv->has_entries;
 }
 
 1;
diff --git a/lib/PublicInbox/LeiSavedSearch.pm b/lib/PublicInbox/LeiSavedSearch.pm
index 929380ed..cfbf68c3 100644
--- a/lib/PublicInbox/LeiSavedSearch.pm
+++ b/lib/PublicInbox/LeiSavedSearch.pm
@@ -315,11 +315,11 @@ E: rename($dir_old, $dir_new) error: $!
 EOM
 }
 
-# cf. LeiDedupe->dedupe_nr
-sub dedupe_nr {
+# cf. LeiDedupe->has_entries
+sub has_entries {
 	my $oidx = $_[0]->{oidx} // die 'BUG: no {oidx}';
-	my @n = $oidx->{dbh}->selectrow_array('SELECT COUNT(*) FROM over');
-	$n[0];
+	my @n = $oidx->{dbh}->selectrow_array('SELECT num FROM over LIMIT 1');
+	scalar(@n) ? 1 : undef;
 }
 
 no warnings 'once';
diff --git a/lib/PublicInbox/LeiToMail.pm b/lib/PublicInbox/LeiToMail.pm
index b9405c0c..d782d4c7 100644
--- a/lib/PublicInbox/LeiToMail.pm
+++ b/lib/PublicInbox/LeiToMail.pm
@@ -198,7 +198,7 @@ sub _mbox_write_cb ($$) {
 	my $dedupe = $lei->{dedupe};
 	$dedupe->prepare_dedupe;
 	my $lse = $lei->{lse}; # may be undef
-	my $set_recent = $dedupe->dedupe_nr;
+	my $set_recent = $dedupe->has_entries;
 	sub { # for git_to_mail
 		my ($buf, $smsg, $eml) = @_;
 		$eml //= PublicInbox::Eml->new($buf);
@@ -293,7 +293,7 @@ sub _maildir_write_cb ($$) {
 	# Favor cur/ and only write to new/ when augmenting.  This
 	# saves MUAs from having to do a mass rename when the initial
 	# search result set is huge.
-	my $dir = $dedupe && $dedupe->dedupe_nr ? 'new/' : 'cur/';
+	my $dir = $dedupe && $dedupe->has_entries ? 'new/' : 'cur/';
 	sub { # for git_to_mail
 		my ($bref, $smsg, $eml) = @_;
 		$dst // return $lei->fail; # dst may be undef-ed in last run
diff --git a/lib/PublicInbox/LeiXSearch.pm b/lib/PublicInbox/LeiXSearch.pm
index cac7fb7d..3414e87d 100644
--- a/lib/PublicInbox/LeiXSearch.pm
+++ b/lib/PublicInbox/LeiXSearch.pm
@@ -504,7 +504,7 @@ sub do_query {
 		my $F_SETPIPE_SZ = $^O eq 'linux' ? 1031 : undef;
 		if ($l2m->{-wq_nr_workers} > 1 &&
 				$l2m->{base_type} =~ /\A(?:maildir|mbox)\z/) {
-			# setup two barriers to coordinate dedupe_nr
+			# setup two barriers to coordinate ->has_entries
 			# between l2m workers
 			pipe(my ($a_r, $a_w)) or die "pipe: $!";
 			fcntl($a_r, $F_SETPIPE_SZ, 4096) if $F_SETPIPE_SZ;
diff --git a/lib/PublicInbox/SharedKV.pm b/lib/PublicInbox/SharedKV.pm
index 8347b195..3487e820 100644
--- a/lib/PublicInbox/SharedKV.pm
+++ b/lib/PublicInbox/SharedKV.pm
@@ -154,6 +154,13 @@ SELECT COUNT(k) FROM kv
 	$sth->fetchrow_array;
 }
 
+# faster than ->count due to how SQLite works
+sub has_entries {
+	my ($self) = @_;
+	my @n = $self->{dbh}->selectrow_array('SELECT k FROM kv LIMIT 1');
+	scalar(@n) ? 1 : undef;
+}
+
 sub dbh_release {
 	my ($self, $lock) = @_;
 	my $dbh = delete $self->{dbh} or return;