| Subject: | [PATCH] Optimization for Class::DBI::Column |
I doing some optimization at $work i've been trying to speed up some
operations that make a lot of accessor hits, to the point that the top
exclusive time in nytprof is name_lc in Class::DBI::Column. The
attached patch makes Class:DBI::Column's cache the lowercased version of
their name rather than compute it on the fly. It also makes it so that
if the columns name is ever updated, that the name_lc is kept in sync.
there hasn't been much (any?) testing done on this, i've just run it in
my little test app thing and it was a significant improvement and didn't
cause any explosions. If there's actually a sliver of hope that this
will get in and released i'd be more than willing to write some tests
for it.
| Subject: | CDBI.patch |
--- lib/site_perl/Class/DBI/Column.pm 2010-07-20 16:07:57.472873202 -0400
+++ /home/aprime/Column.pm 2010-07-20 16:07:05.579941724 -0400
@@ -30,7 +30,7 @@
use Carp;
__PACKAGE__->mk_accessors(
- qw/name accessor mutator placeholder is_constrained/
+ qw/accessor mutator placeholder is_constrained/
);
use overload
@@ -52,6 +52,7 @@
return $class->SUPER::new(
{
name => $name,
+ name_lc => lc $name,
accessor => $name,
mutator => $name,
_groups => {},
@@ -61,7 +62,14 @@
);
}
-sub name_lc { lc shift->name }
+sub name {
+ return $_[0]->{'name'} if scalar(@_) == 1;
+ my $ret = $_[0]->{'name'} = scalar(@_) == 2 ? $_[1] : [@_[1..$#_]];
+ $_[0]->{'name_lc'} = lc $_[0]->name;
+ return $ret;
+}
+
+sub name_lc { shift->{name_lc} }
sub add_group {
my ($self, $group) = @_;