Horton

Re: Early WIP/PoC for inlining CTEs

Andreas Karlsson <andreas@proxel.se>

View thread
From: Andreas Karlsson <andreas@proxel.se>
To: Andrew Gierth <andrew@tao11.riddles.org.uk>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, David Fetter <david@fetter.org>, Thomas Munro <thomas.munro@enterprisedb.com>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2019-01-02T04:31:30Z
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 1/1/19 1:42 AM, Andrew Gierth wrote:
>>>>>> "Andreas" == Andreas Karlsson <andreas@proxel.se> writes:
> 
>   Andreas> I believe I have fixed these except for the comment on the
>   Andreas> conditions for when we inline.
> 
>   Andreas> Andrew Gierth: Why did you chose to not inline on FOR UPDATE
>   Andreas> but inline volatile functions? I feel that this might be
>   Andreas> inconsistent since in both cases the query in the CTE can
>   Andreas> change behavior if the planner pushes a WHERE clause into the
>   Andreas> subquery, but maybe I am missing something.
> 
> I chose not to inline FOR UPDATE because it was an obvious compatibility
> break, potentially changing the set of locked rows, and it was an easy
> condition to test.
> 
> I did not test for volatile functions simply because this was a very
> early stage of the project (which wasn't my project, I was just
> assisting someone else). I left the comment "this likely needs some
> additional checks" there for a reason.

Thanks, that makes sense! I will need to ponder some on if the behavior 
change when predicates are pushed into a subquery with volatile 
functions is ok. I am leaning towards no, because otherwise inlining 
CTEs would affect more than query performance.

Andreas