Horton

Re: postgres_fdw bug in 9.6

Etsuro Fujita <fujita.etsuro@lab.ntt.co.jp>

View thread View on postgr.es
From: Etsuro Fujita <fujita.etsuro@lab.ntt.co.jp>
To: Robert Haas <robertmhaas@gmail.com>, Ashutosh Bapat <ashutosh.bapat@enterprisedb.com>
Cc: Jeff Janes <jeff.janes@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2016-12-27T03:44:09Z
Lists: pgsql-hackers 
On 2016/12/23 1:04, Robert Haas wrote:
> On Wed, Dec 21, 2016 at 11:04 AM, Ashutosh Bapat
> <ashutosh.bapat@enterprisedb.com> wrote:
>> Some review comments
>>
>> 1. postgres_fdw doesn't push down semi and anti joins so you may want to
>> discount these two too.
>> +           jointype == JOIN_SEMI ||
>> +           jointype == JOIN_ANTI);

> But in the future, it might.

I plan to work on adding those cases to postgres_fdw.

> We shouldn't randomly leave foot-guns
> lying around if there's an easy alternative.

Some FDWs might have already supported pushing down semi/anti joins.  So 
I think it's better to handle those joins as well.

>> 3. Adding new members to JoinPathExtraData may save some work for postgres_fdw
>> and other FDWs which would use CreateLocalJoinPath(), but it will add few bytes
>> to the structure even when there is no "FULL foreign join which requires EPQ"
>> involved in the query. That may not be so much of memory overhead since the
>> structure is used locally to add_paths_to_joinrel(), but it might be something
>> to think about. Instead, what if we call select_mergejoin_clauses() within
>> CreateLocalJoinPath() to get that information?

> I think that's exactly backwards.  The few bytes of storage don't
> matter, but extra CPU cycles might.

I agree with Robert.

Best regards,
Etsuro Fujita

Commits

  1. Fix test case for 'outer pathkeys do not match mergeclauses' fix.

  2. postgres_fdw: Avoid 'outer pathkeys do not match mergeclauses' error.

  3. postgres_fdw: Consider foreign joining and foreign sorting together.

  4. Allow foreign and custom joins to handle EvalPlanQual rechecks.

  5. Allow FDWs to push down quals without breaking EvalPlanQual rechecks.