Hi Doug:

>From dougo@ccs.neu.edu Thu Jun 12 15:25:52 1997
>From: Doug Orleans <dougo@ccs.neu.edu>
>To: Karl Lieberherr <lieber@ccs.neu.edu>
>Cc: dem@ccs.neu.edu
>Subject: Re: Beta release planning
>
>Karl Lieberherr writes:
> > >making stub g_* methods (that instanciate a generic visitor and pass it
> > >   to the universal traversal) 
> > 
> > Yes, this worked very well for Demeter/C++. Will need strongly
> > connected component algorithm to compute vertex basis?
>
>The SCC algorithm is already there (but commented out), but why do we
>need the vertex basis?  The universal traversal already goes from
>every class to every class.  

You are right: when you generate one complete traversal method for each class,
you get what you call the universal traversal. No need to use the
vertex basis.

>Also, I'd prefer not to call them g_*;
>we should use Java functions where appropriate, e.g. toString() for the

I agree!

>PrintVisitor traversal, equals() for the EqualVisitor, etc.  For the
>CopyVisitor we can't really use clone() because I think the
>contract of clone() is shallow copy (or "structure copy"); in demjava
>I call it deepCopy().
>
Thos are all good names which are easy and natural for Java programmers.

-- Karl