Thread: Proof of concept for MIN/MAX optimization

I haven't done the work yet to add a column to pg_aggregate, so this
just knows about optimizing a couple of hard-wired cases (int4 and
text).  Other than that it's reasonably complete, I think.  Comments?

            regards, tom lane

Tom Lane wrote:
> I haven't done the work yet to add a column to pg_aggregate, so this
> just knows about optimizing a couple of hard-wired cases (int4 and
> text).  Other than that it's reasonably complete, I think.  Comments?
>

Looks good :

regression=# explain select max(unique1) from tenk1;
                                                QUERY PLAN
---------------------------------------------------------------------------------------------------------
  Result  (cost=0.15..0.16 rows=1 width=0)
    InitPlan
      ->  Limit  (cost=0.00..0.15 rows=1 width=4)
            ->  Index Scan Backward using tenk1_unique1 on tenk1
(cost=0.00..1512.59 rows=10000 width=4)
(4 rows)


The handling of nulls is a little unexpected (still todo?) :

regression=# insert into tenk2 (unique1,unique2) values(null, 10000);
INSERT 0 1
regression=# select max(unique1) from tenk2;
  max
-----

(1 row)
regression=# set enable_indexscan=0;
SET
regression=# select max(unique1) from tenk2;
  max
------
  9999
(1 row)

cheers

Mark

Mark Kirkwood <markir@paradise.net.nz> writes:
> The handling of nulls is a little unexpected (still todo?) :

Yeah, that was broken in the first draft :-( ... I think it's OK in
the committed version though.

            regards, tom lane

Tom Lane wrote:
> Mark Kirkwood <markir@paradise.net.nz> writes:
>
>>The handling of nulls is a little unexpected (still todo?) :
>
>
> Yeah, that was broken in the first draft :-( ... I think it's OK in
> the committed version though.
>

(post cvs update) yeah - looks good!

regards

Mark