Thomas Lockhart <lockhart@alumni.caltech.edu> writes:
> I'm not sure why we throw an error if you drop a
> function which does not exist, since that makes it tough to blindly do
> the "drop/create" pair. Why don't we just signal a warning or notice
> instead?
It doesn't matter unless you are inside a transaction --- but I can
see the value of replacing a function definition inside a transaction.
Perhaps "no such <whatever>" should be downgraded from ERROR to NOTICE
for all DROP-type commands. Another TODO item...
regards, tom lane
