Geometry Boost List: [ggl] Spatial Indexes

Subject: [ggl] Spatial Indexes
From: Barend Gehrels (Barend.Gehrels)
Date: 2009-06-04 09:39:41

Hi Chris,

>
>
>         While browsing for a spatial indexing library, I came across
>         this boost SOC project
>         (http://svn.boost.org/svn/boost/sandbox/SOC/2008/spacial_indexing/)
>
>         Unfortunately, it seems abandonware. As it relies on GGL and
>         is geometry related, wouldn'it be a good idea to incorporate
>         it to GGL?
>
>     Good question. It is not abandoned but was indeed not yet taken
>     over into the previews. But it is recently incorporated in the
>     trunk, at least the rtree index, and will be part of next preview.
>     It works fine and speeds up results (though it should be
>     revisited). If you want to use it I would advise to change to the
>     trunk in an agreed moment (such that you can be confident that
>     everything runs fine). Also because of next point:
>
>
> Great news. Indeed, It is a major incentive to switch to GGL trunk, as 
> I'm busy incorporating the rtree from the aforementionned svn branch, 
> probably wasting some time ;-)
> Anyway, I'll stabilize what I have with the GGL and rtree sandbox as 
> to have a baseline, then switch to GGL trunk.
OK, then we've done the same thing indeed... Whatever you want is find, 
I can also send you what I have if you wish that.

>
> BTW, an issue that I have with the "stock" rtree is that I can only 
> remove a leaf by its bbox. This is an issue for me as if several 
> "Values" (i.e. objects) have the same bbox, the first one is 
> arbitrarily removed.
> I thus added a (not-yet-tested) "remove(bbox, value)" function, to be 
> sure to remove the intended value.
>
> Is it something you tackled?
No. It is a good point and should definitely be in it.

> A "remove(value)" would also be ok (certainly less efficient, but 
> easier for coding as I wouldn't have to save the "old" bbox before an 
> update)
I think it would be useful. I'm not yet so into this spatial indexing 
library. The author then was a student, now he's probably many other 
things to do.
There were some very useful comments about it on the boost list, we have 
to rework them.

>
>
>     The polygon clipping problems seem to be solved now. Can you send
>     me some pictures of the situations going wrong in the old version?
>     I can then check if they are solved. It is still in progress but
>     will be finished and incorporated in the trunk soon (probably this
>     weekend).
>
>
> Ok. I have to hack my code to make things visible, so it'll probably 
> be in a couple of days (considering the stabilization).
OK, thanks, if it will take you too much time, it is not necessary, we 
then can just try. I've done some tests and it appears to be all fine.

Regards, Barend

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/ggl/attachments/20090604/50806dad/attachment.html