Re: PG 7.2b4 bug?

Don Baccus <dhogaza@pacifier.com> writes:
> Maybe the behavior's implementation defined ... if not, I'd presume SQL3 
>   states that a function in the above context is called either once per 
> row or once per query, not sometimes one or sometimes the other.

AFAICT, the relevant concept in SQL99 is whether a function is
"deterministic" or not:
        An SQL-invoked routine is either deterministic or possibly non-        deterministic. An SQL-invoked function
thatis deterministic always        returns the same return value for a given list of SQL argument        values. An
SQL-invokedprocedure that is deterministic always        returns the same values in its output and inout SQL parameters
      for a given list of SQL argument values. An SQL-invoked routine        is possibly non-deterministic if, during
invocationof that SQL-        invoked routine, an SQL-implementation might, at two different        times when the
stateof the SQL-data is the same, produce unequal        results due to General Rules that specify
implementation-dependent       behavior.
 

It looks to me like the spec does NOT attempt to nail down the behavior
of non-deterministic functions; in the places where they talk about
non-deterministic functions at all, it's mostly to forbid their use in
contexts where nondeterminism would affect the final result.  Otherwise
the results are implementation-defined.
        regards, tom lane