This is particular to postgres, although the
SQL behavior would have either dropped
the constraint or prevented the drop in the
first place.
There's been some talk of an ALTER FUNCTION
that would let you change the code behind
a function without a drop/create.
Generally you have to re-generate things that
reference functions that have been dropped
and re-created. This is a pain right now
for constraints, since it requires a dump
and restore of the table.
On Fri, 18 Aug 2000, Thomas Swan wrote:
>
> Using Postgresql 7.0.2 (Linux x86, 2.2.16)
>
> CERATE FUNCTION foo(text)
>
> CREATE TABLE bar(
> fud TEXT CHECK (foo(fud))
> );
>
> DROP FUNCTION foo(TEXT);
> CREATE FUNCTION foo( .....);
>
> INSERT INTO bar VALUES ('Hey'); results in the following error
>
> ERROR init_fcache: Cache lookup failed for procedure 128384
>
> Is this particular to postgres or is this a normal SQLxx standard behavior?