NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/object/3020

Internet Message Format | 1992-07-22 | 1.9 KB

Xref: sparky comp.object:3020 comp.lang.eiffel:988
Newsgroups: comp.object,comp.lang.eiffel
Path: sparky!uunet!walter!geoff
From: [e-mail] (Geoffrey Clemm)
Subject: Re: Class methods (was: Re: How to design a data structure library)
In-Reply-To: [e-mail]'s message of 24 Jul 92 10:58:49 GMT
Message-ID: <[e-mail]>
Sender: [e-mail]
Nntp-Posting-Host: wodehouse
Organization: Bellcore
References: <[e-mail]> <[e-mail]>
Date: 27 Jul 92 10:06:01
Lines: 43

In article <[e-mail]> [e-mail] (Markus Stumptner) writes:
> With explicit binary operators (as in C++ or Smalltalk),
> you would write (if "+" expresses union)

> c := a + b

> That way, "+" is a message declared for sets, and used on sets.  If
> you want a named message, you would write

> c := a.union(b)

> (Still looks odd, but lacks that redundant third parameter)

> > In Smalltalk lingo, "union" should be a "class method".  The essence of
> > a class method is that it can be invoked when the object of the given
> > class does not exist yet.

> Yes, but you already have two sets, a, and b, either of which you can
> send the message to.  Nobody decrees that the receiver of a message
> has to be changed by it, therefore returning c is a valid way to do
> it.

I believe you missed the central point of Richie's article, namely that
Eiffel 3.0 has a clean way of handling this situation, to quote:

In article <[e-mail]> [e-mail] (Richard Bielak) writes:

> Given this, the above call to "union" would be written like this
> (using Eiffel 3 syntax):

>     !!c.union (a,b);

> Which I think is better.

With which I completely agree.  This avoids the unpleasant assymetry
of c:=a.union(b).

Cheers,

Geoff

--
[e-mail]