Phil Endecott wrote:
Jonathan Franklin wrote:
On Tue, Nov 17, 2009 at 2:29 PM, Simonson, Lucanus J <lucanus.j.simonson@intel.com> wrote:
It is exactly because it is important that I don't want to rush it.
How long would you estimate it takes to get the interface right? It seems like that is really the piece that needs to be enterprise class at this point (or at least, when first introduced).
In the context of a Boost library we would surely expect the interface to a spatial index to be quite similar to the interfaces to std:: containers, so there are quite a few interface design decisions that are non-issues. In my experience, the only significant novelty in the interface is how to represent 2D ranges (i.e. to iterate over the points in a rectangle); a range2d type that takes a reference to the container and its bounds and provides a similar interface as the container itself has worked for me.
(Luke, I'd be interested to know if your indexes differ significantly from this.)
The data structure pre-dates my library and I did not enhance its interface except to make it support my rectangle concept. It provides output iterator for rectangles that match the query and may carry a payload of type T associated with the rectangle. I didn't like the interfaces and felt it would be too much work to enhance to interfaces to make them boost worthy only to have people complain that the performance is terrible and that the data structure is worthless for anything but CAD. Thanks, Luke