Hm, ok, setting up a new database with the Conway implementation I see the
time penalty is not "negligible". However it is quite tolerable:
test=> explain analyze select * from test order by a; QUERY PLAN
----------------------------------------------------------------------------------------------------------------Sort
(cost=69.83..72.33rows=1000 width=32) (actual time=772.175..900.155 rows=206300 loops=1) Sort Key: a -> Seq Scan on
test (cost=0.00..20.00 rows=1000 width=32) (actual time=0.009..203.362 rows=206300 loops=1)Total runtime: 1055.137 ms
(4 rows)
test=> explain analyze select * from test order by lower(a); QUERY
PLAN
----------------------------------------------------------------------------------------------------------------Sort
(cost=72.33..74.83rows=1000 width=32) (actual time=1190.248..1318.607 rows=206300 loops=1) Sort Key: lower(a) -> Seq
Scanon test (cost=0.00..22.50 rows=1000 width=32) (actual time=0.021..517.679 rows=206300 loops=1)Total runtime:
1565.564ms
(4 rows)
test=> explain analyze select * from test order by strxfrm(a,'en_US');
QUERYPLAN
-----------------------------------------------------------------------------------------------------------------Sort
(cost=72.33..74.83rows=1000 width=32) (actual time=2135.535..2263.429 rows=206300 loops=1) Sort Key: strxfrm(a,
'en_US'::text) -> Seq Scan on test (cost=0.00..22.50 rows=1000 width=32) (actual time=0.060..1384.428 rows=206300
loops=1)Totalruntime: 2516.981 ms
(4 rows)
test=> explain analyze select * from test order by strxfrm(a,'fr_CA');
QUERY PLAN
------------------------------------------------------------------------------------------------------------------Sort
(cost=72.33..74.83rows=1000 width=32) (actual time=2155.991..2284.054 rows=206300 loops=1) Sort Key: strxfrm(a,
'fr_CA'::text) -> Seq Scan on test (cost=0.00..22.50 rows=1000 width=32) (actual time=18.455..1403.390 rows=206300
loops=1)Totalruntime: 2538.886 ms
(4 rows)
--
greg
