Re: plans for PostgreSQL 12

Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> writes:
> 2018-06-04 9:59 GMT+02:00 Vik Fearing <vik(dot)fearing(at)2ndquadrant(dot)com>:
>> On 04/06/18 09:37, Pavel Stehule wrote:
>>> Yes, it is incorrect mark. Unfortunately - this is often workaround for
>>> wrong estimations - so I afraid, in this case, your proposed fix breaks
>>> lot of applications.

>> I would say such applications are already broken.

> I cannot to agree, not in this moment:
> 1. there is not any workaround, how to force subselect evaluation in
> planning time - what can be correct for once only evaluated queries.

There's the sadly-underdocumented trick of writing the expression inside
a sub-select so that it becomes an InitPlan. We could doubtless improve
our support for that --- for instance, teach estimate_expression_value()
how to get an estimated value when the expression is stable --- but I'm
unconvinced that we need new infrastructure for this. Inventing a new
function volatility class would be an enormous mess from users'
standpoint, especially if the reason was only to distinguish cheating
uses from non-cheating uses of the existing class. I am not inclined
to promise that we'll never break cheating uses.

regards, tom lane