[GRASS5] Want GNU libavl ?

[David D Gray]

Andrea Aime wrote:

> Mmm, I'm planning a full rewrite of s.voronoi, s.delaunay 
> cause the original Fortune algorithm implementation is not
> commented and a bit difficult to read, moreover comp. geom.
> textbooks tend to give a sligtly different description of the
> alghorithm... I need a priority queue and a balanced tree and
> I was considering using the code available at www.purists.org
> were also an AVL tree is available... 
> Should we put togheter a little container library for GRASS
> that could be statically linked into the executables? As far as
> I remember a b-tree implementation is already in the GRASS cvs,
> maybe a more complete library would be useful and could be included
> in GRASS 5.1...



Actually there is a simple binary tree from way back. Maybe a btree has 
been added recently, I don't know, but we do need better container 
support. I first came across this when I started on the import filters a 
couple years back. In fact I first linked the v.in.shape binary against 
Berkeley DB2 btree, which gave a good result, but introducing it at that 
time was potentially a large job, and there were questions of platform 
support and its a big package to bundle if it's to become a dependency.

So I agree that a number of different containers could be included, and 
it would also be good to have some kind of uniformity in the interface 
to these, like a system of wrapper functions.


David