On 2016-09-01 15:46:45 -0400, Tom Lane wrote:
> Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
> > While investigating a CTE-related query, I've noticed that we don't
> > really remove all unreachable CTEs.
>
> We expend a grand total of three lines of code on making that happen.
> I'm pretty much -1 on adding a great deal more code or complexity
> to make it happen recursively;
Agreed. And the consequences are pretty much harmless.
> the case simply doesn't arise in reasonably well written queries.
Well, it might, when the CTE reference can be removed due to some other
part of the query (e.g. plan time evaluation of immutable function).
Andres