SEDS for ISO 10303-11

[Ed Barkmeyer]

Oops!  I supported this too soon:

Jochen Haenisch wrote:

> Same for aggregates, but only if they are really assigned a different type
> of aggregate:
>   LOCAL
>     x : BAG OF INTEGER;
>     y: SET OF INTEGER := [1,2,3];
>   END_LOCAL;
>   x := y;
> In this case TYPEOF(x) should return ['SET', 'BAG'], not only ['BAG'].

NO!  In my view, TYPEOF(x) should return ['BAG'].
That is because there are two distinct "occurrences":  the variable x and the variable y.  The assignment x := y; takes the
value [1, 2, 3] from y, converts it to a value of type BAG OF INTEGER, and assigns the resulting value to x.  If subsequently,
we see:
  x = x + [1];
the result is [1,2,3,1], because that is what clause 12 says is the interpretation of x + [1].

If you have:
 FUNCTION f(x : BAG OF INTEGER);
 LOCAL
  t : SET OF STRING := TYPEOF(x);
 END_LOCAL;
  ...
 END_FUNCTION;

 FUNCTION q(...);
 LOCAL
   x : BAG OF INTEGER;
   y: SET OF INTEGER := [1,2,3];
 END_LOCAL;
   x := f(y);

then the TYPEOF(x) (in f) should return ['SET', 'BAG'], not only ['BAG'].
That is because there is only one occurrence -- the local variable y in q.  And that occurrence was of type SET OF INTEGER and
was the actual parameter passed down to TYPEOF.

Now, if in function f, we have:
 x := x + [1];
the result should be [1,2,3].  The actual parameter y is substituted for x, the expression x+[1] is then evaluated with that
substitution, being an operation on a SET type (not a BAG type), and the result is assigned to y, an instance of a SET type.

That is, I make a distinction between assignment and substitution, and I only want the "what am I really" behavior to occur for
formal parameters, generalized types and select types.  (The latter two explicitly say that the occurrence retains its type.)

> The assignment is the point! Only then the program can really trace the type
> of an instance.

And in so doing, you lose the ability to coerce the value of the instance to a different type.

Put another way, if an assignment retains the type of the right-hand operand, there is no point in declaring types for local
variables.  They are always of whatever data type the right-hand operand is interpreted to be.  That is a possible and perhaps
better model; but I don't think it is the one Express wants.  (And the one I want doesn't have local variables at all! ;-)

-Ed

-- 
Edward J. Barkmeyer                       Email: edbark at nist.gov
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Mail Stop 8260          Tel: +1 301-975-3528
Gaithersburg, MD 20899-8260               FAX: +1 301-975-4482

"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority."