On Thu, Dec 05, 2024 at 06:34:31PM -0500, Tom Lane wrote: > You have a mighty optimistic view of what will happen. I predict > that if we do step (1), exactly nothing will happen in applications, > and step (2) will remain just as painful for them. (Assuming that > we remember to do step (2), which is no sure thing given our track > record for following through "in a few years".)
FWIW, my first thought after reading this paragraph is that you sound too dramatic here, especially after looking at codesearch to note that the PHP core code stores attndims but does not actually use it. Now, I've looked at github and noted that there's a large number of repositories (in the hundreds) that rely on a zero or non-zero value for *attndims* to check if they're dealing with an array, saving in what looks like catalog lookups. The same lookups done for typndims offer an opposite conclusion: there's no real code in the open that uses this value for similar checks, most likely because domains are less used in the wild. So I get the drama for the removal of attndims and I'd agree to keep it around, but I also see a very good point in removing typndims. -- Michael
signature.asc
Description: PGP signature
