Re: List container strawman

[Ted Dennison<dennison@telepath.com>]

In article <3bec1cbe$0$15824$626a54ce@news.free.fr>, Jean-Marc Bourguet says...
>
>Ted Dennison<dennison@telepath.com> writes:
>
>Why do you need dynamic allocation.  The pointer can be in the

The implementation he presented (a working one, I believe) had it. Since I've
only thought about it before, and he's actually implemented it, I'm assuming
there's some implementation reason he needed to do it that way that isn't
apparent just looking at a spec. Perhaps there isn't. That's what I'd like to
hear.

>> o Add extra dynamic memory operations to any list modification (there were
>> some there already).  
>
>I don't think so.

Hmmm. Yeah, your'e right. Its actually another one of those "ususally small"
internal list traversals. What will happen is that you have to check all other
iterators. But that's probably only nessecary when an item is deleted, which I
covered below.

>> o Add a (usually small) internal list traversal to any element
>> deletion.
>
>Agree.
>
>> o Add a small validity check (or possibly two) to every iterator
>> operation.
>
>Agree.
>
>> o Possibly make iterator creation non-determistic (time-wise).  
>
>I don't think so. (Iterator deletion would be made non deterministic
>if they are assigned in a non nested way and the iterator list is a
>singly linked one).

I said "probably" this for both the last bulletts because he implemented
iterators using accesses to controlled types. If it can be done on the stack,
then both these "possibly"s go away (and I would thus vote for it).

Perhaps it would be best if I could play with an implementation and see if its
doable. That would mean I'd have to tear myself away from Civ3 for a few hours
though...

---
T.E.D.    homepage   - http://www.telepath.com/dennison/Ted/TED.html

No trees were killed in the sending of this message. 
However a large number of electrons were terribly inconvenienced.