Horton

Re: Bogus EXPLAIN results with column aliases for mismatched partitions

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Etsuro Fujita <etsuro.fujita@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2019-12-03T17:12:09Z
Lists: pgsql-hackers 
Etsuro Fujita <etsuro.fujita@gmail.com> writes:
> On Tue, Dec 3, 2019 at 6:45 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Concretely, I'm thinking of the attached (on top of the other patch,
>> which I just pushed).  This agrees exactly with what ExplainTargetRel
>> does for regular scans.

> Thanks for the patch!  (The patch didn't apply to HEAD cleanly,
> though.)  I like consistency, so +1 for the change.

Yeah, 55a1954da probably changed the expected output from what that
has.  I'll clean it up and push.

>> One could make an argument that we should schema-qualify, regardless
>> of the "verbose" flag, if the output format isn't EXPLAIN_FORMAT_TEXT.
>> That would reduce the amount of variability that plan analysis tools
>> have to cope with.  However, ExplainTargetRel itself doesn't provide
>> the schema in non-verbose mode.  Maybe it should, ie we should change
>> it like ...

> Seems like another issue to me.

Agreed.  When/if we change that, we could make this code follow along.

			regards, tom lane

Commits

  1. Further sync postgres_fdw's "Relations" output with the rest of EXPLAIN.

  2. Fix EXPLAIN's column alias output for mismatched child tables.

  3. Add a reverse-translation column number array to struct AppendRelInfo.

  4. Make postgres_fdw's "Relations" output agree with the rest of EXPLAIN.

  5. Allow access to child table statistics if user can read parent table.