Re: [boost] [tree] Reviving the tree library

Subject: Re: [boost] [tree] Reviving the tree library
From: Erik Erlandson (eerlands_at_[hidden])
Date: 2011-05-05 20:34:52

On Wed, 2011-05-04 at 17:17 -0500, Rene Rivera wrote:

> It looks like Bernhard stopped development 1.5 years ago. Which sure 
> sounds like it's orphaned. Given that I've copied it over to 
> sandbox/tree location (the last trunk state for it). I'm willing to help 
> out in moving the library forward, especially since it's based on some 
> of the core concepts I suggested. Who else is willing to help out in 
> improving and finishing this lib? Note, that helping out involves 
> thinking about adjusting the TR proposal so that it can be submitted for 
> the soon to start up again TR2.
> 
> 

A couple other observations:

Some common tree-related methods like parent(), depth(), ply() seem like
they would be good additions.   Maintaining them adds some overhead, but
I found I was able to keep things logarithmic in complexity.

I found that I wanted to support what I called 'storage models' for
child nodes.  I ended up with three of them:  

1) a vector<>-like model, where node[j] gives the jth-child
2) a map<>-like model, where node[key] gives the child indexed by (key)
2) a multiset<>-like model where children are automatically maintained
in a sorted order.

I'm not sure if that has a role in a Hierarchical Container concept
definition -- such things might just be instantiations of such a
concept.