At 18:00 +1200 2001.05.02, Richard A. O'Keefe wrote:
>
>The problem is that using
>     self species new: (size)
>to return a copy of part of a sequence doesn't provide information about
>the additional instance variables.

But it could, if you are clever.

We are used to thinking that species when sent to a collection 
instance has to return a class.  But that is not actually the case.

What happens to the result of "self species"?  The code sends it new 
or new: size.  So, as long as the object that you get back from self 
species understands new or new:, everything will be well.  We need 
only a factory, not a class.

In your case, if you would like some initialized instance variables, 
you can build an object in #species that has them.  And then when you 
send the new: method, you could adjust it's size, or you could make a 
copy with a different size.

This is where a type inference system would be really nice.  It would 
tell us at a glance exactly what all of the existing classes do with 
self species.  (You would find that you have to implement hash too. 
and who knows what else...)

	Andrew