Path: news.daimi.aau.dk!poe
From: "Soren Brandt" <sbrandt@staff.cs.su.oz.au>
Newsgroups: comp.lang.beta
Subject: Re: Re-Two question about assignment
Date: 15 Jun 1995 08:38:56 GMT
Organization: DAIMI, Computer Science Dept. at Aarhus University
Lines: 40
Approved: mailtonews@daimi.aau.dk
Distribution: world
Message-ID: <3rorj0$n1c@belfort.daimi.aau.dk>
NNTP-Posting-Host: daimi.daimi.aau.dk

It is *not* possible to test, (by name), whether an object has some
attribute.  Doing so would require extensive runtime type information
that is currently not available.

What you *can* do, is to test whether a given object is an instance of some
pattern. For example:

    Person: (# ... #);
    o: ^Object;
    do ...
       (if o##<=Person## then
	  (* o refers an object that is at least a person. *)
       if);

Here, o## means "a reference to the actual pattern of which o is an instance",
Person## means "a reference to the Person pattern", and finally the "<=" 
operator compares patterns. "<=" applied to patterns returns true iff the 
lefthand side is a subpattern of the righthand side.

Hope this helps.

Regards,
Soren

> 
> My second question 'Can I know at run-time if an
> object supports a specific (by-name) attribute ?'
> is for debugging purpose only, also because if 
> I use extensively that technique I loose benefit 
> of BETA language and go back to Smalltalk untyped 
> language. However (maybe) I'll redo the question 
> when I'll explore singular pattern object or/and 
> I'll build an example with distributed logic.
> 
> 	Thanks again, 
> 
> 		Dario Gerosa
>