Re: Planner estimates cost of 'like' a lot lower than '='??

Mats Lofkvist <mal(at)algonet(dot)se> writes:
> I just did a quick test before going to bed but 7.2devel seems
> do to what we want:

> testdb=> select count(*) from data where value like 'test_0';
> count
> -------
> 9
> (1 row)

> testdb=> explain select count(*) from data where value like 'test_0';
> NOTICE: QUERY PLAN:

> Aggregate (cost=5581.45..5581.45 rows=1 width=0)
> -> Seq Scan on data (cost=0.00..5580.69 rows=306 width=0)

> EXPLAIN
> testdb=> select count(*) from data where value >= 'test' and value < 'tesu';
> count
> -------
> 10000
> (1 row)

> testdb=> explain select count(*) from data where value >= 'test' and value < 'tesu';
> NOTICE: QUERY PLAN:

> Aggregate (cost=6007.24..6007.24 rows=1 width=0)
> -> Seq Scan on data (cost=0.00..5986.02 rows=8487 width=0)

> EXPLAIN

Those estimates do look a lot closer to reality, all right. And the
join plan is the right thing now. Excellent...

regards, tom lane