On Thu, Oct 12, 2017 at 11:50 PM, Tom Lane wrote:
> Ants Aasma writes:
>> I stumbled upon a severe row count underestimation that confusingly
>> went away when two inner joins in the from clause were reordered.
>
> Hm, looks more like an overestimate in this example, but anyway ...
>
>> Does anyb
Ants Aasma writes:
> I stumbled upon a severe row count underestimation that confusingly
> went away when two inner joins in the from clause were reordered.
Hm, looks more like an overestimate in this example, but anyway ...
> Does anybody have any idea what is going on here?
set_joinrel_size_e
I stumbled upon a severe row count underestimation that confusingly
went away when two inner joins in the from clause were reordered. I
whittled it down to a reproducible test case.
Schema:
CREATE TABLE small (id serial primary key, ref_id int not null, subset
int not null);
CREATE TABLE big (id
