On Wed, Nov 30, 2016 at 8:44 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Merlin Moncure <mmoncure@gmail.com> writes:
>> On Wed, Nov 30, 2016 at 11:05 AM, George <pinkisntwell@gmail.com> wrote:
>>> So there is definitely something wrong here. This situation makes many
>>> row-level security use cases cumbersome since you need to have
>>> almost the same WHERE clause both in the row-level security policy and
>>> in every SELECT query in order for the index to be used.
>
>> can you give EXPLAIN ANALYZE for the 'good' query and the 'bad' query?
>
> Planning for queries affected by RLS is definitely an area where we need
> to improve (I'm working on a patch for that). Whether the OP's particular
> query is being hit by that is impossible to tell, though, since there
> isn't any actual RLS usage in the doubtless-oversimplified example.
The example is not over-simplified, I basically just took the clause
that the RLS would have to add and stuck it in the WHERE. Thus I
verified that even the normal, non-RLS planner is affected.
When I get to work tomorrow morning (Europe) I will post the EXPLAIN
ANALYZE output.