Re: Our trial to TPC-DS but optimizer made unreasonable plan

On 2015-08-19 15:14:03 -0700, Josh Berkus wrote:
> Asking users to refactor their applications to add OFFSET 0 is a bit
> painful, if we could take care of it via a backwards-compatibility GUC.
> We have many users who are specifically using the CTE optimization
> barrier to work around planner failures.

Agreed. I think we'll cause a lot of problems in migrations if we do
this unconditionally. I also think CTEs are a much cleaner optimization
barrier than OFFSET 0.

Some are probably going to hate me for this, but I think it'd be better
to change the grammar to something like
name opt_name_list AS '(' PreparableStmt ')' OPTIONS '(' cte_option_list ')'

and allow to specify 'inline' 'off'/'on'. The guc would simply change
the default value.

Greetings,

Andres Freund