On 23.02.24 16:19, Tom Lane wrote:
> Dean Rasheed <dean.a.rasheed@gmail.com> writes:
>> On Fri, 23 Feb 2024 at 14:35, Peter Eisentraut <peter@eisentraut.org> wrote:
>>> Various code comments say that the RangeTblEntry field inh may only be
>>> set for entries of kind RTE_RELATION.
>
>> Yes, it's explained a bit more clearly/accurately in expand_inherited_rtentry():
>
>> * "inh" is only allowed in two cases: RELATION and SUBQUERY RTEs.
>
> Yes. The latter has been accurate for a very long time, so I'm
> surprised that there are any places that think otherwise. We need
> to fix them --- where did you see this exactly?
In nodes/parsenodes.h, it says both
This *must* be false for RTEs other than RTE_RELATION entries.
and also puts it under
Fields valid in all RTEs:
which are both wrong on opposite ends of the spectrum.
I think it would make more sense to group inh under "Fields valid for a
plain relation RTE" and then explain the exception for subqueries, like
it is done for several other fields.
See attached patch for a proposal. (I also shuffled a few fields around
to make the order a bit more logical.)