Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!ucla-cs!bath.cs.ucla.edu!rowley
From: rowley@bath.cs.ucla.edu (Michael T Rowley)
Newsgroups: comp.lang.eiffel
Subject: Renaming vs. Conformance
Message-ID: <32149@shemp.CS.UCLA.EDU>
Date: 21 Feb 90 21:08:52 GMT
Sender: news@CS.UCLA.EDU
Reply-To: rowley@cs.ucla.edu
Organization: UCLA Computer Science
Lines: 19

I have been looking into questions of multiple inheritance, especially
with regard to repeated inheritance. In eiffel a feature may be
renamed along one of the inheritance paths, giving the descendant two
features inherited from the same ancestor.  (See FRANCE_US_DIVER on
p.276 of OOSC).  If the feature is not renamed, the descendent gets
only one feature.  It seems that the use of renaming would break the
conformance rule (on p.262).

The problem arises with a variable declared as the ancestor type (A).
It should be possible to assign to this variable objects of any of
its descendent types.  But if a feature (f) was renamed on both paths to
the descendent (D), then what does A.f mean.

The conformance rule would claim that the assignment of A = D is
valid, and therefore it should be possible to use the features of A.
But, clearly this is ambiguous.  Should the conformance rule specify
that an object does not conform to an ancestral type from which it
inherits at least one of the ancestor's features under two different
names?