Re: [boost] Re: Re: Boost library inclusion: generic tree containers?

From: Andreas Pokorny (andreas.pokorny_at_[hidden])
Date: 2005-02-26 04:55:46

On Fri, Feb 25, 2005 at 01:11:29PM -0700, Justin Gottschlich <jgottschlich_at_[hidden]> wrote:
> Absolutely - I've already started reviewing Jeff's design and am having a 
> side discussion with him regarding his trees (so I can understand what the 
> heck is going on, heheh). If his solution works for the base generic tree 
> solution, then by all means, let's use that. =)
No it solves all tree wishes, all as in terms related to the word 'god' :).
There is no border, nothing is fixed, everything is a mutator, and
mutators are minimal implementations of features. 

> We just need to make sure that trees outside of purely algorithmic uses are 
> addressed. With the increasing use of XML, trees simply as storage devices 
> are become more and more popular. It is my believe (which could be 
> completely off) that our basic tree should be generalized in its structure 
> enough to support any kind of tree construction.
You will see that this difference is no problem in that library.
In my opinion it is not a good idea to provide a basic tree, if you do
that some wishes for tree types are impossbile to fulfill. If you
definde a basic tree then there is a fixed contract that you can either
reuse or not use. So take it for granted that there will be cases in
which that basic type cannot be reused. 

By defining the tree as a composition of feature, you will not fall in
that pit. Jeff Miriawasis tree library is not policy driven. Policy
driven base tree types will fall in that pit either. 

> I think we want our generalized tree to be basic enough so novice coders can 
> use it for its basic containment purposes, but powerful enough so advanced 
> coders can stress the heck out of it.
I doubt that this is possible. 

Regards 
Andreas Pokorny