Rodrigo Barboza <rodrigombufrj@gmail.com> writes:
> I created a type 'mytype' (an unsigned int) and created an operator class
> for index.
> Then I created a table with a column of my type and isnerted 1000 entries.
> But no matter how many entries I have in the table, it never uses the
> index. It always does a seq scan.
> Here is the explain analyze with 1000 entries:
> explain analyze select * from mytable where a > 120::mytype and a <
> 530::mytype;
> ---------------------------------------------------------------------------------------------------
> Seq Scan on mytable (cost=0.00..19.02 rows=400 width=4) (actual
> time=0.023..0.229 rows=409 loops=1)
> Filter: ((a > '120'::mytype) AND (a < '530'::mytpe))
> Total runtime: 0.297 ms
Using a seqscan to fetch 400 out of 1000 entries is the right thing.
(The crossover point where an index becomes unhelpful is a lot closer
to 1% of the table than it is to 40%.)
regards, tom lane