Re: Foreign Join pushdowns not working properly for outer joins

On Thu, Apr 13, 2017 at 7:35 AM, David Rowley
<david(dot)rowley(at)2ndquadrant(dot)com> wrote:
> On 13 April 2017 at 11:22, Peter Eisentraut
> <peter(dot)eisentraut(at)2ndquadrant(dot)com> wrote:
>> Is this patch considered ready for review as a backpatch candidate?
>
> Yes, however, the v5 patch is based on master. The v4 patch should
> apply to 9.6. Diffing the two patches I see another tiny change to a
> comment, of which I think needs re-worded anyway.
>
> + * This function should usually set FDW options in fpinfo after the join is
> + * deemed safe to push down to save some CPU cycles. But We need server
> + * specific options like extensions to decide push-down safety. For
> + * checking extension shippability, we need foreign server as well.
> + */
>
> This might be better written as:
>
> Ordinarily, we might be tempted into delaying the merging of the FDW
> options until we've deemed the foreign join to be ok. However, we must
> do this before performing this test so that we know which quals can be
> evaluated on the foreign server. This may depend on the
> shippable_extensions.
>

This looks better. Here are patches for master and 9.6.
Since join pushdown was supported in 9.6 the patch should be
backported to 9.6 as well. Attached is the patch (_96) for 9.6,
created by rebasing on 9.6 branch and removing conflict. _v6 is
applicable on master.

--
Best Wishes,
Ashutosh Bapat
EnterpriseDB Corporation
The Postgres Database Company