Re: [9.5] next question: rls and indexes

On 15 July 2015 at 09:34, Andreas Kretschmer <akretschmer@spamfence.net> wrote:
> test=> explain select * from rls_test ;
>                          QUERY PLAN
> -------------------------------------------------------------
>  Seq Scan on rls_test  (cost=0.00..630.00 rows=91 width=335)
>    Filter: (name = ("current_user"())::text)
> (2 rows)
>
> As you can see, the index isn't in use, but i think, in this case PG
> should use it.
>
> Bug or Feature?
>

Yes, that's a bug. It should indeed use the index for queries like
that. Debugging it reveals that the reason it didn't is a collation
mismatch which caused it to believe there were no suitable indexes to
use.

Fortunately this has been fixed (post 9.5 alpha 1) -- see
http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=808ea8fc7bb259ddd810353719cac66e85a608c8
-- and the same query on HEAD now does use the index.

Thanks for testing.

Regards,
Dean