Question about discrimiants

Question about discrimiants

[Erland Sommarskog]

The following declaration is incorrect (LRM 3.7.1(6)):

TYPE Nisse(Size : integer_type) IS
     RECORD
        A : Array(1..Size) of integer;
        B : Array(1..2*Size) of integer;
     END;
     
Does anyone know why? I can understand forbidding a dynamic expression,
but a static?
-- 
Erland Sommarskog       
ENEA Data, Stockholm    
sommar@enea.UUCP        
                        

Re: Question about discrimiants

[Dik T. Winter]

In article <2079@enea.UUCP> sommar@enea.UUCP(Erland Sommarskog) writes:
 > 
 > The following declaration is incorrect (LRM 3.7.1(6)):
 > 
 > TYPE Nisse(Size : integer_type) IS
 >      RECORD
 >         A : Array(1..Size) of integer;
 >         B : Array(1..2*Size) of integer;
 >      END;
 >      
 > Does anyone know why? I can understand forbidding a dynamic expression,
 > but a static?
(It would perhaps be better if programs posted to the net where at least
syntactically correct in order to reduce the number of messages pointing
to syntactic errors.  The following comments refer to this part after
making it syntactically correct.)

Yes, 3.7.1(6) explicitly states that discriminants may be used in subtype
definitions (as is done here), but in that case cannot be used in larger
expressions.
Anyhow, you appear to understand forbidding dynamic expressions, can you
elaborate?  At least, the context here is dynamic and non-static
(a discriminant is never static).

I think the designers wanted to avoid the need for a system to produce
run-time routines for many declarations (as in Algol-68), but for a full
explanation we have to wait for the Rationale of course.
-- 
dik t. winter, cwi, amsterdam, nederland
INTERNET   : dik@cwi.nl
BITNET/EARN: dik@mcvax

Re: Question about discrimiants

[stt]

I think it was strictly an implementability issue.
As a compiler-writer, I certainly sympathize with
the restriction.  As a user of Ada, I agree the restriction
seems arbitrary and not philosophically justifiable.

By the way, the expression is not "static" in the
normal Ada sense, since the value of a discriminant
is not a static expression.

Finally, a record type definition must end with "end record;"
rather than simply "end;"

Tucker Taft
c/o Intermetrics, Inc.
Cambridge, MA  02138