Jason Kim <git@jasonk.me> writes:
> In the middle of GIN index testing, there are some selects that are on
> a different table array_op_test that doesn't even have an index. They
> probably were supposed to be selects to table array_index_op_test like
> the other ones around the area.
I think it's probably intentional, else why have two tables at all?
I suppose the point of these test cases is to confirm that you get the
same results with or without use of an index.
Certainly, there's more than one way to do that. Perhaps we should
have only one table and perform the variant tests by manipulating
enable_indexscan et al. But I think what you did here is defeating
the intent.
regards, tom lane