Re: Value vs. Const

Joe Pluta wrote:
> ...
> The reason I ask is that I'm trying to set up some programming
> standards.  Should I always pass parameters as CONST (except when they
> are truly going to be modified)?  Or are there cases (let's say the data
> is only 2 bytes long) where it might be more efficient to use VALUE,
> because only the actual data is being passed on the stack, not a
> pointer.
> 

It is probably more efficient to pass integer/unsigned/float numeric
types by value than by reference.  But I don't think the difference is
really significant enough to bother making a special case.

Using VALUE for "large" parameters should be avoided.  If it's handy to
change the parameter in the procedure, it's more efficient to copy a
CONST parameter to local storage than to pass a VALUE parameter.  (The
system isn't particularly optimized to pass large parameters by value,
and I think that's probably true of most if not all systems, so this
rule of thumb probably applies on all systems for all languages.) 
Similarly, returning large values should be avoided, but that's more
difficult in RPG - it's nice to be able to return large strings, for
instance.  So, while it might be "better" to use an output parameter for
a large return value, I think I'd tend to take the performance hit to
make it easier to call.

The reason I would do the extra work for the VALUE/CONST parameters but
not for the return values is that a procedure should be as easy as
possible to call, possibly at the cost of making it more difficult to
write.  VALUE/CONST doesn't make any difference coding a call to the
procedure, but doing a call with an output parameter (which has to be
the correct type) is much ickier than having a return value.