Christophe Pettus <xof@xxxxxxxxxxxx> writes: >> On Feb 22, 2020, at 14:02, Tom Lane <tgl@xxxxxxxxxxxxx> wrote: >> It's a really bad idea to mark a function that has side-effects >> (i.e., emitting a NOTICE) as immutable, especially if the occurrence >> of the side-effect at well-defined times is exactly what you're >> desirous of. > True, and it doesn't actually need to be immutable here; just cut and pasted from the wrong example. > (That being said, I'm not coming up with a specific bad thing that a RAISE NOTICE in an immutable function will cause. Is there one?) The problem that I'm worried about is premature evaluation of the "immutable" function, causing the NOTICE to come out once during query planning, independently of whether/how many times it should come out during execution. regards, tom lane
