[GRASS5] lib/vector/rtree/split_l.[c|h]

Radim Blazek radim.blazek at gmail.com
Mon Aug 22 06:00:19 EDT 2005


On 8/21/05, Markus Neteler <neteler at itc.it> wrote:
> On Sat, Aug 20, 2005 at 11:25:31PM -0700, Brad Douglas wrote:
> > Is it okay if I remove split_l.[c|h] from lib/vector/rtree and it's
> > Makefile?  They are unused and essentially duplicated in split_q.[c|h].
> >
> > IMHO, it's clutter that serves as obfuscation.
> >
> 
> It should be ok (and it doesn't get lost in CVS).
> 
> Markus

I would prefer to keep it in CVS,  split_l and split_q implement 2 methods
for node splitting and I had not time to think/test both in GRASS. 
If we leave it there, hopefully someone can compare both with real data.
Look also in MAILS:

| regarding split_l.c vs. split_q.c: these involve a choice you can make between
| speed and box optimization. you should use the quadratic version ('q') if it's
| fast enough for your needs, and use the less expensive linear method
if not. note
| that in both cases various branches of the resulting rectangle trees
may overlap.
| there's another version of the technique called "R+ Trees" which guarantees no
| overlaps, but i don't have an implementation of that version. i
believe it's even
| slower than either r-tree splitting method, so you should probably
also take all
| of that into account.


Radim




More information about the grass-dev mailing list