Re: Default Inline CTE makes JOIN slower - Mailing list pgsql-bugs

From Pavel Stehule
Subject Re: Default Inline CTE makes JOIN slower
Date
Msg-id CAFj8pRCyavgQu38EJB_LGjU+yvC=HR-54E0bYdFRxSXobJcRig@mail.gmail.com
Whole thread Raw
In response to Default Inline CTE makes JOIN slower  (hao li <shell0fly@gmail.com>)
List pgsql-bugs
Hi

po 29. 3. 2021 v 15:39 odesílatel hao li <shell0fly@gmail.com> napsal:
Hi,
I have this example query in my code

create or replace function slow_int() returns integer language sql stable 

as $$

select pg_sleep(1);

select 1;

$$;


explain analyze WITH s AS (

SELECT * FROM (values (1), (2), (3)) s(a)

),

slow_int AS (

SELECT slow_int()

)

SELECT * FROM s

JOIN slow_int ON true;


and I got this explain analyze output which execute slow_int() three times, you can see the execution time is 3s, not 1s.


Values Scan on "*VALUES*"  (cost=0.00..0.79 rows=3 width=8) (actual time=1002.468..3005.399 rows=3 loops=1)
Planning Time: 0.326 ms
Execution Time: 3005.512 ms


It happens because PG12 automatically inline CTE, and if I force it to be materialized, I can get previous execution time.


sure. It is expected behaviour - and reason why there is new clause MATERIALIZED

Regards

Pavel

pgsql-bugs by date:

Previous
From: hao li
Date:
Subject: Default Inline CTE makes JOIN slower
Next
From: PG Bug reporting form
Date:
Subject: BUG #16946: It may be that the content of the document is missing