Comments on: Waiting for 8.4 – pl/* srf functions in selects

By: depesz

depesz — Tue, 16 Dec 2008 09:22:46 +0000

@Sergey:
# select i, (test(1, i)).* from generate_series(1,3) i;
i | numerical | textual
—+———–+———
1 | 1 | i = 1
2 | 1 | i = 1
2 | 2 | i = 2
3 | 1 | i = 1
3 | 2 | i = 2
3 | 3 | i = 3
(6 rows)

By: Sergey Konoplev

Sergey Konoplev — Mon, 15 Dec 2008 23:30:27 +0000

Hubert, woud it be possible to do the same this way?
select i, (test(1, i)).* from generate_series(1,3) i;

By: depesz

depesz — Mon, 03 Nov 2008 15:01:21 +0000

@David:
heh, CTEs are so new that I simply forgot about them, but here is goes:

with source as (
select i, test(1, i) from generate_series(1,3) i
)
select i, (test).numerical, (test).textual from source;

but i’m not really sure if this will give better results than the subquery approach. explains look like this:

cte:
QUERY PLAN
—————————————————————————————————————————–
CTE Scan on source (cost=262.50..282.50 rows=1000 width=36) (actual time=0.217..0.426 rows=6 loops=1)
InitPlan
-> Function Scan on generate_series i (cost=0.00..262.50 rows=1000 width=4) (actual time=0.206..0.382 rows=6 loops=1)
Total runtime: 0.483 ms
(4 rows)

subquery:
QUERY PLAN
—————————————————————————————————————————
Subquery Scan x (cost=0.00..272.50 rows=1000 width=36) (actual time=0.200..0.388 rows=6 loops=1)
-> Function Scan on generate_series i (cost=0.00..262.50 rows=1000 width=4) (actual time=0.193..0.358 rows=6 loops=1)
Total runtime: 0.437 ms
(3 rows)

By: David Fetter

David Fetter — Mon, 03 Nov 2008 14:55:27 +0000

Hubert,

Since you’re talking about 8.4 anyhow, you might include CTEs in that last example, as this is one of the places they do wonders for

By: depesz

depesz — Mon, 03 Nov 2008 12:42:59 +0000

@Regina:
Just tested, with function defined as stable or immutable. I checked how many times function was called by adding “raise notice” in it. in all cases it got called twice (i.e. in case of subselect approach it was only once per row).

By: Regina

Regina — Mon, 03 Nov 2008 12:29:53 +0000

Hubert,

This is neat. I’ve been waiting for something like this so I don’t have to do an SQL wrapper hack.

One question. You sure the function currently gets called twice? I haven’t really tested so can’t be sure.

I was always under the impression that if you set the function to STABLE/IMMUTABLE, then it would only get called once per unique test(1, i). Otherwise why wouldn’t you be able to put 2 set returning functions in the SELECT and also get multiplicative?

Thanks,
Regina