From: Marcin 'Qrczak' Kowalczyk <qrczak@...>
Date: 2002-10-23T00:48:32+09:00
Subject: Re: adding overload to ruby

Fri, 27 Sep 2002 01:35:16 +0900, Christoph <chr_news@gmx.net> pisze:

> b)  CLOS way - Use the lexicographic or similar order
>      (for example a mixture of  path length + lexicographic order)
>      on the multi-dispatch argument types  to determine the
>      best (closest) match.  This is fairly unnatural imo.
> 
> c)  Determine if  a ``most specific '' specialization  on
>      argument type (+ self type) exists. In case of ambiguities
>      call  a method like ambiguous_method_call .
>      (default action is to  raise an Exception) . This is Cecil's
>      and Dylan's way I believe.
> 
> d)  (A combination of b & c).  Similar to c) but instead of
>      raising an exception in the presence of ambiguities choose
>      the  lexicographic largest (or largest in another
>      good total order)  among  the ``most   specific type
>       specializations.

What is the difference between d) and b)?

I have never used CLOS but I believe I once reinvented its ordering.
It never signals ambiguities and prefers matches with larger number of
dispatched types, then compares lexicographically (an implementation
can be much simpler than for the Cecil's way).

Isn't it the only reasonable choice among those which don't signal
ambiguities?

-- 
  __("<      Marcin Kowalczyk
  \__/     qrczak@knm.org.pl
   ^^    http://qrnik.knm.org.pl/~qrczak/