On Sat, 2023-07-29 at 12:44 -0400, Isaac Morland wrote: > Essentially, "just" observe efficiently (somehow) that no change is > needed, and skip changing it?
I gave this a try and it speeds things up some more. There might be a surprise factor with an optimization like that, though. If someone becomes accustomed to the function running fast, then changing the search_path in the caller could slow things down a lot and it would be hard for the user to understand what happened. Also, there are a few implementation complexities because I think we need to still trigger the same invalidation that happens in assign_search_path(). Regards, Jeff Davis