On Fri, Nov 18, 2022 at 11:53 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
Alvaro Herrera <alvherre@alvh.no-ip.org> writes: > On 2022-Nov-18, Marko Tiikkaja wrote: >> This is a documented limitation: >>> Trying to update the same row twice in a single statement is not >>> supported.
> I wonder if we should try to detect the case, and raise an error instead > of it resulting in undefined behavior.
My recollection is that that is really fallout from an ancient and intentional executor behavior, that we have to ignore multiple updates in order to not get into infinite loops. See comment about the "Halloween problem" in nodeLockRows.c. (I'm pretty sure there were once more comments about that, somewhere closer to ExecUpdate/ExecDelete --- this all dates back to Berkeley.)
I'm not really sure I'd want to change the behavior to perform multiple updates even if we could. But in a green field development I would prefer the error. Right now I'd side with introducing an error as well.