Joe Conway <mail@joeconway.com> writes:
> On 5/11/21 1:30 PM, Bruce Momjian wrote:
>> It just feels like this change makes the function's behavior less
>> consistent.
> See Tom's commit message here:
> https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=3d0f68dd30612
> In particular:
> "The variants of these functions that take
> numeric inputs (OIDs or column numbers) are
> supposed to return NULL rather than failing
> on bad input; this rule reduces problems with
> snapshot skew when queries apply the functions
> to all rows of a catalog."
Yeah, the null-return-for-bad-numeric-input behavior is important.
Perhaps a case could be made for returning null for bad text
input too, but I don't recall that anybody has asked for that.
A case could also be made that changing the behavior on the text
side would break applications that expect the current behavior.
So I'm disinclined to make a wholesale change there, without more
evidence that it's a good idea.
regards, tom lane