On Mon, Nov 16, 2009 at 10:36 PM, Barend Gehrels <barend@geodan.nl> wrote:
As I've a GIS-background myself too I'm glad to read this. The reasons for moving it to extension were: - the kernel should be geometry and not GIS - there is one thing still to be incorporated in the design of the GIS-extension (this is: the earth model, it is referred to in distance and in projections, and it should be shared) - the library is large and projections should make much it larger, and we didn't write the maths ourselves
Moving it to extensions make sense. My point is that the extensions necessary to make the library useful for GIS should be part of the core library rather than a separate extension that the user has to find.
One question, do you mean with key-extension: - the geographic coordinate system - or also the projections?
The first is small, a few sources and distance calculations, and they were included in the distribution because they nicely show how to work with BGL (and GIS/SVG). The projections is more, ~90 complex math sources
I think the problem with the PROJ dependency could be handled similary to the MPI library, that provides bindings to the C library.
That is an option, but we converted PROJ to C++ style, that alone gives already some %'s performance gain. Besides that we want to be able to use GMP numbers (optionally), so we don't want to mingle them into existing (heavily macro's) C sources.
The first one is key, the second one you would need feedback from more reviewers. If there is no PROJ dependency then it's even easier to include it. regards jose