Horton

Re: Early WIP/PoC for inlining CTEs

Andrew Gierth <andrew@tao11.riddles.org.uk>

View thread
From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Andreas Karlsson <andreas@proxel.se>, David Fetter <david@fetter.org>, Thomas Munro <thomas.munro@enterprisedb.com>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2019-01-27T13:22:57Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Prevent inlining of multiply-referenced CTEs with outer recursive refs.

  2. Allow user control of CTE materialization, and change the default behavior.

  3. Split QTW_EXAMINE_RTES flag into QTW_EXAMINE_RTES_BEFORE/_AFTER.

  4. document when PREPARE uses generic plans

>>>>> "Tom" == Tom Lane <tgl@sss.pgh.pa.us> writes:

 Tom> I was interested to find, while writing the docs, that it's a real
 Tom> struggle to invent plausible reasons to write MATERIALIZED given
 Tom> the above specification. You pretty much have to have lied to the
 Tom> planner, eg by making a volatile function that's not marked
 Tom> volatile, before there's a real need for that. Am I missing
 Tom> something? If I'm not, then we're in a really good place
 Tom> backwards-compatibility-wise, because the new default behavior
 Tom> shouldn't break any cases where people weren't cheating.

The cases where the new default will break things are where people used
WITH to force a choice of join order or otherwise constrain the planner
in order to avoid a misplan.

I'm not sure we should nail down the rule that the absence of NOT
MATERIALIZED will mean a multiply-referenced CTE is evaluated once. One
would hope that in the future the planner might be taught to inline or
not in that case depending on cost. I think it makes more sense to say
that we never inline if MATERIALIZED is specified, that we always inline
if NOT MATERIALIZED is specified, and that if neither is specified the
planner will choose (but perhaps note that currently it always chooses
only based on refcount).

-- 
Andrew (irc:RhodiumToad)