🚀 go-pugleaf

john@linus.mitre.org writes:
>Joe Konstan writes:
>  First, it seems important to have declarations that declare an exact
>  type instead of just a general type.  Thus,
>    (DECLARE (EXACT-TYPE CLASS-1 OBJECT))
>  should indicate that object is treated as an instance of class-1,
>  and not as an instance of a subclass.
>and Graham Matthews (graham%maths.su.oz.au@uunet.uu.net) replies:
>  Why not take this the whole hog and treat your object heirarchy as
>  an ordered set (which it is in the single inheritance model - not
>  sure about multiple inheritance) and allow the programmer to write a
>  type spec like
>	FUNCTION F( <= TYPE_A, > TYPE_B )
>  etc
>A type hierarchy with single inheritance is NOT an ordered set, it's
>only partially ordered.  

I was thinking of one branch of the tree here. In general it is indeed
partially ordered.

>Still, such declarations might be useful, and
>you can already make them in Common Lisp:
>  (DEFUN FOO (X)
>    (DECLARE (TYPE (AND SYMBOL (NOT NULL)) X)
>    ...))
>This declares that the type of X is "exactly" SYMBOL.

This is not what I proposed. This allows on = on the order, and I
am suggesting to extend this to <, etc

>However, the
>declarer must know the exact nature of the hierarchy, whereas Joe
>Konstan's EXACT-TYPE declaration doesn't require that every
>subclass/subtype be enumerated in the declaration.

No this is not true. To use a type spec like, "t1<t2" one needs
only know that all types/classes less than t1 satisfy some condition.
In algebraic programming this is often the case, with group < ring <
field...

graham
--
Graham Matthews                 And it's true we are immune
Pure Math, Uni.Sydney, Oz       When fact is fiction and T.V. is reality
graham@maths.su.oz.au