[Prev][Next][Index][Thread]

Re: efficient dylan dispatch talk slides available


Well, the slides describe how to implement into Dylan a support for
dispatching on state. This reminds me of a statement from  one of the three UML
amigos.

"There are two ways to specify a subclass. The first is to describe the
incremental differences form the base class in terms of new attributes or
operations [...] I call this subclassing by augmentation.
The other way to form  a subclass is to specify some constraints that
instances of a base must satisfy to be considered members of the new class. I call
this subclassing by restriction [...]
You can think of restrictions as defining different states of a class. An
object can change states freely but cannot change class. It would be
convenient,  if object-oriented languages would support state using the class
descriptor mechanísm, including method dispatching dependent on state. [...], but I
suppose it is too late to hope for this kind of change."

 James Rumbaugh in  his article
 "Disinherited -- Examples of Misuse of Inheritance"
 JOOP (Feb 1993),
 also published in his book "OMT Insights"
 

There is hope: Dylan.

We now know that it is possible to support dispatching on state in Dylan,
but who likes to implement it? Functional Objects?

Peter

-- 
GMX - Die Kommunikationsplattform im Internet.
http://www.gmx.net


-- 
Posted from mx0.gmx.net [213.165.64.100] 
via Mailgate.ORG Server - http://www.Mailgate.ORG