On Fri, 11 May 2007 04:24:55 +0200, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Leif B. Kristensen" <leif@solumslekt.org> writes:
>> Would it be reasonable to suggest that later versions of PostgreSQL
>> could examine if a function changes data, and quietly marks a function
>> as 'stable' if it doesn't?
>
> My instinctive CS-major reply to that is "only if you've found a
> solution to the halting problem". However, it's possible that we could
> detect this case for a useful subset of real-world functions ... not
> sure offhand what could be covered.
>
> regards, tom lane
Why not simply have PG issue a warning if the user doesn't specify one of
("stable", "immutable", etc) on function creation ?
like :
WARNING: Function marked as Volatile by default
INFO: if the function does not modify the database, you might want to mark
it STABLE or IMMUTABLE to improve performance