Re: [boost] Final report of GSOC project 'Spatial Indexes'

Subject: Re: [boost] Final report of GSOC project 'Spatial Indexes'
From: Michael Fawcett (michael.fawcett_at_[hidden])
Date: 2008-09-11 22:05:12

On Thu, Sep 11, 2008 at 8:43 PM, Christian Holmquist
<c.holmquist_at_[hidden]> wrote:
>
> I have one early question, have you considered pushing the quadtree and
> rtree classes into boost::intrusive, or designing them in the spirit of
> boost::intrusive?
> The audience who's in need of large spatial databases (the ones who doesn't
> need it are fine with linear scans) is likely to hunt for performance. IMO
> boost::intrusive has given a flexibility to c++ containers not available
> elsewhere, it'd be a clear winner if your quadtree and rtree can be
> implemented in generic terms.

I have a generic n-ary tree library that might be a good starting
point for integrating into Intrusive.  It already has its own
intrusive tree with a regular tree container built on top of that,
plus a post order iterator.  I've sent a preliminary version to Ion a
while ago, but it's since matured beyond that.  I'll also have to
check if I can make the full code available since it's work related,
but so far I've built a Bounding Interval Hierarchy off its
foundations.

Regardless of how this gets accomplished, I completely agree that the
base class for all of these spatial index classes belongs in
Intrusive, with Quad-trees, etc built on top in intrusive and
non-intrusive flavors.

--Michael Fawcett