Geometry Boost List: [ggl] Hello + some newbie questions

Subject: [ggl] Hello + some newbie questions
From: Barend Gehrels (Barend.Gehrels)
Date: 2009-04-27 04:28:38

Hi Nicolas,

Welcome! Thanks for your interest in GGL.

> 1) When I've read the list archive I've seen reference to some commits 
> and works on ggl files, like de9im_segment.cpp. But I don't find these 
> files on boost/sandbox or in http://geometrylibrary.geodan.nl/. Is 
> there another place for checking ggl code?
Yes, there still is. The library is currently in "preview" at boost and 
for the last preview we submitted it to the boost/sandbox. Because we're 
working on it on a daily base we work still in the original SVN (still 
hosted at Geodan). The boost/sandbox files are "major steps" with a 
coherent state.
People can get access to Geodan's ggl-svn. So if you want I can organize 
it for you.

(It is also possible that we move it, because boost also has a 
"sandbox-branches" folder.)

> 2) Working with airbone laser scanner data I'm interrested by 
> multi_point feature and it use with algorithms as convex_hull. What is 
> your plan concerning multi_point?
It will be implemented for multi-point, but currently it is not. Because 
linestrings (ranges) and polygons are, the multi-point is now trivial so 
it can be implemented soon.

> 3)I'm also interrested by your plan for the projection part of ggl.
Nice, you're probably the first to ask things about this :-).

> - Have you compare performance of ggl and proj4?
It is an automatic conversion of proj4 code. By the conversion of C to 
C++ alone there seems to be a performance increment of some percents. 
Have to remeasure and to compare different projections to give more 
exact results. The ggl has a "template" version, completely compiletime, 
and a "virtual" version, with a factory, for runtime projection 
selection. The template version is, as expected, somewhat faster than 
the virtual version.

> - Is there a "gridshift" implementation plan to ggl?
> - What about a pj_transform algorithm on ggl ?
Both yes, currently only the projection-part is ported from proj4 but 
the plan is to have also the transformations. We've to dive into them.

> - Have you planned to refactore more deeply projection with 
> meta-programming tools ( introduce concept for datum, projection 
> etc..) or it is just utilities for ggl?
This is also (partly) yes. I made some comments about harmonize the 
ellipsoid in the distance function(s) with the projections. This should 
be worked out further. You mention concepts for datum, that was not yet 
mentioned before but is a good idea, it is a logical step.

Regards, Barend