Re: [ooc-compiler] libadt: Naming Issues

[Stewart G. <gre...@mu...>]

"Griebling, Michael" wrote:
> 
> I'm not sure whether this would solve any problems but
> languages such as Ada which support generic interfaces
> through which a data type could be supplied during
> instantiation.  This would at least take care of the
> need to subclass based on different types.  Someone
> was working on a method of doing something like this
> with one of the Oberon dialects but I can't recall
> which one it was.

Yes. I agree. Without generics (or parametric types), collections or
other abstract data types must be framed in terms of a universal type.
In Java (for example), its impossible to define a Set of X, or a Mapping
from A to B within the standard framework. The framework defines Set of
Object, and Mapping from Object to Object. Since everything sub-types
Object, any object can participate in these data structures, but there
is no static guarantee of the type of the members of these collections.
This means lots of (unnecessary) type guards in client code.

Roe and Szyperski proposed parametric polymorphism for Oberon:
  http://www.fit.qut.edu.au/~szypersk/pub/JMLC97.ps.gz
I'm not sure if anyone actually implemented it. Both Oberon Microsystems
and the QUT team have stated their intentions to support it in their
Component Pascal implementations. Actually, its probably not that hard
to support a limited generics facility using the technique outlined in
Roe and Szyperski's paper. It requires no change to the Oberon run-time
system, just a re-interpretation of the type rules for pointers.
Combined with interfaces and/or compound types, this could be very
powerful.

Cheers,
  Stewart