Horton

Re: Early WIP/PoC for inlining CTEs

Robert Haas <robertmhaas@gmail.com>

View thread
From: Robert Haas <robertmhaas@gmail.com>
To: Andreas Karlsson <andreas@proxel.se>
Cc: Andres Freund <andres@anarazel.de>, Thomas Munro <thomas.munro@enterprisedb.com>, Andrew Gierth <andrew@tao11.riddles.org.uk>, Tom Lane <tgl@sss.pgh.pa.us>, David Fetter <david@fetter.org>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2019-01-18T20:34:46Z
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

On Thu, Jan 17, 2019 at 10:48 AM Andreas Karlsson <andreas@proxel.se> wrote:
> On 1/11/19 8:10 PM, Robert Haas wrote:
> > WITH cte_name [[NOT] MATERIALIZED] AS (query) main_query...
>
> Hm, when would one want "NOT MATERIALIZED"? I am not sure I see the
> usefulness of forcing inlining other than if we by default do not inline
> when a CTE is referenced multiple times.

When the planner materializes it, but the performance of the resulting
plan therefore sucks, I suppose.

I don't feel super-strongly about this, and Tom is right that there
may be cases where materialization is just not practical due to
implementation restrictions.  But it's not crazy to imagine that
inlining a multiply-referenced CTE might create opportunities for
optimization at each of those places, perhaps not the same ones in
each case, whereas materializing it results in doing extra work.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company