Re: [HACKERS] CTE inlining - Mailing list pgsql-hackers

On Tue, May 9, 2017 at 12:15 PM, Peter Eisentraut <peter.eisentraut@2ndquadrant.com> wrote:

On 5/5/17 08:43, David Rowley wrote:
> How about we get the ball rolling on this in v10 and pull that part
> out of the docs. If anything that'll buy us a bit more wiggle room to
> change this in v11.
>
> I've attached a proposed patch.

If we just tell them that the thing they might have relied on might go
away, without a replacement to suggest, then we're just confusing and
scaring them, no?

​We'd end up suggesting our OFFSET 0 hack as true protection. If they know for a fact that their use of CTE for its barrier properties is not supported they are also more likely to document intentional usage with something like: "-- CHANGE THIS ONCE VERSION 11 IS RELEASED!!! --" which would make finding the call sites that need to add the new "MATERIALIZED" ​keyword much easier.