On Wed, Jan 17, 2024 at 5:46 AM Tom Lane <t...@sss.pgh.pa.us> wrote: > > But perhaps someone has an argument for a different rule? > > Anyway, pending discussion of that point, I think the code is good > to go. I don't like the test cases much though: they expend many more > cycles than necessary. You could prove the same points just by > looking at the expansion of expressions, eg. >
your patch is far better! IMHO, worried about the support function, the transformed plan generates the wrong result, so we add the tests to make it bullet proof. Now I see your point. If the transformed plan is right, the whole added code should be fine. but keeping the textrange_supp related test should be a good idea. since we don't have SUBTYPE_OPCLASS related sql tests.