On 06/12/2017 01:41 PM, Chapman Flack wrote:
> I was recently guilty of writing a less-than-clear SQL example
> because I plain forgot whether InvalidOid was 0 (correct) or -1
> (my bad).
>
> Would there be any sense in letting oidin_subr accept the string
> InvalidOid for 0? I understand that changing oidout could break
> existing code outside of the tree. But what if oidout were to be
> conservative in what it does, and oidin liberal in what it accepts?
I am not sure I am a fan of this, but if we should have an alias for
InvalidOid how about reusing '-' which is used by the reg*in functions?
Or is that too non-obvious?
Andreas