Robin wrote:
> > One good example is where IBM doesn't match the standard, it doesn't
> > always
> > evaluate the operand of the LENGTH builtin (which could invoke a
> > function) if it can
> > compute the result at compile-time. So, even though the expression
> > might have a
> > side-effect, it is not invoked.
>
> If the string is fixed length, PL/I doesn't need to evaluate anything.
> If the argument to LENGTH is an expression, the expression might need to
> be evaluated.
> For example, LENGTH (A || B) would not require the evaluation of the
> expression
> when both A and B are of fixed length. The length would be available at
> compile time.
>
Consider this example:
ex: proc options(main);
dcl var fixed bin(31) init(0);
dcl l fixed bin(31);
sub: proc(val) returns( char(8) );
dcl val fixed bin(31);
var = val;
return ('Hi there');
end;
l = length(sub(10));
put skip list('l is ', l);
put skip list('var is', var);
end;
what do you expect the output of that to be? In my reading of the PL/I
standard,
the output should be:
l is 8
var is 10
as it clearly indicates that argument to builtin functions should be evaluated.
You're right in that if the determination can be made at compile-time, there
is no problem with the compiler simply producing a constant with the
appropriate
attributes.
- Dave Rivers -
p.s.I have taken the liberty of changing the subject...
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
