On 27-11-2020 04:14, Greg Nancarrow wrote:
> On Thu, Nov 26, 2020 at 6:11 PM Luc Vlaming <luc@swarm64.com> wrote:
>>
>> If interesting I can make a draft of what this would look like if this
>> makes it easier to discuss?
>>
>
> Sure, that would help clarify it.
Okay. I will try to build an example but this will take a few weeks as
vacations and such are coming up too.
>
> I did debug this a bit, but it seems my gut feeling was wrong, even
> though it knows a type coercion is required and can be done, the
> parse/analyze code doesn't actually modify the nodes in place "for
> fear of changing the semantics", so when the types don't exactly match
> it's all left up to the planner, but for this parse tree it fails to
> produce a parallel plan.
>
Yes. However I think here also lies an opportunity, because to me it
seems much more appealing to have the planner being able to deal
correctly with all the situations rather than having things like
flatten_simple_union_all() that provide a solution for the ideal case.
> Regards,
> Greg Nancarrow
> Fujitsu Australia
>
Regards,
Luc
Swarm64