Re: Using collections

[Roger M. Burkhart]

> I had a question about using collections for integers or other numbers.
> I'd like to use the collections instead of an ordinary array so I could use
> the probes and also be able to cleanly drop them, but they seem a bit
> cumbersome for simple types.  First, one of the integers could be zero, but
> nil (also 0) is used as the end of list marker in the loops; as well, the
> contains: predicate appears not to handle searching for zeros.

There are alternative ways to test for the end of a loop on collection
members, using an explicit test for index location:

  index = [aCollection begin: scratchZone];
  while ( 1 ) {
    member = [index next];
    if ( [index getLoc] == End ) break;
    ...
  }

So, if you don't mind casting to/from id type whenever you add/retrieve other
types from a collection, you should still be able to accomplish anything
you want to accomplish.  But you're right that the current default contains:
method doesn't consider the possible presence of a zero member right now;
that's a bug that needs to be fixed.

> Second,
> since there's only room for a pointer, this would seem to leave no
> efficient way to handle doubles.  Presumably the power of the
> createBegin-createEnd paradigm is that I could create an Array (say), set
> the type to doubles, and get back an array tailored for doubles.  I assume
> that this kind of optimization is desired, but not high priority--can you
> say where on the to-do list these features are?  Or maybe I've missed the
> better way to do this.

Yes, the createBegin/End sequence could be used to declare a different-sized
internal member, with a different implementation as a result.  This has been
considered and I'd like to see it some day, but lots of other things are
ahead of it, including core capabilities of collections on id types.

Roger