[R-sig-Geo] A bit of new code

[Roger Bivand]

Tim,

Thanks for taking the initiative on this, re:

https://stat.ethz.ch/pipermail/r-sig-geo/2014-January/020241.html

Your choice of Rcpp is one that the extension of R to include reference 
classes recently, provoked by Rcpp, makes possible and attractive.

Would you prefer to use this list for postings with regard to rgdal2, or 
gdal-dev, the wiki on the github repository, or some alternative?

Would you consider perhaps aligning rgdal2 with GDAL 2.0, rather than the 
GDAL 1.* series? In GDAL 2, breaking changes will be introduced - as Even 
writes: "this will require projects using GDAL to adapt multiple times":

http://lists.osgeo.org/pipermail/gdal-dev/2014-February/038126.html

This would give rgdal2 the advantage of being ready when GDAL 2 is ready, 
and puts GDAL 1.* and rgdal in "maintenance mode" while GDAL 1.* is still 
with us?

I see many issues with clang and gcc 4.8.2 in the present code, but this 
isn't the right forum to take them up. Probably also building with clang 
is sensible future-proofing. On the R side, I've recently been running CMD 
check with pqR too (but some pqR efficiency changes are being applied to 
R-devel anyway, so other R-like engines are maybe not so important.

Finally, do you know whether Julia supports an Rcpp-like interface, in 
which case jgdal2 might also be a personality of what you are proposing?

I'll post this on gdal-dev too.

Roger

On Thu, 13 Feb 2014, Tim Keitt wrote:

> Hi All,
>
> I've made a start on a rewrite of rgdal. You can find it at
> https://github.com/thk686/rgdal2.
>
> There is still lots to do. I would be happy to have a GSOC student work on
> this project if anyone is interested (via R or OSGeo). In particular I
> would really like to have a test suite. Writing the examples is what is
> slowing me down at this point.
>
> I do not store the auto-generated files in git, so you need roxygen2 and
> Rcpp to build.
>
> Some of the goals of the project are to expose GDAL/OGR facilities through
> a native R interface emulating arrays, lists, etc; to make it easy to
> iterate over raster blocks and vector features; and to provide support for
> raster-vector operations.
>
> It is semi-useable at this point; I'd be happy to have folks test it out.
>
> THK
>
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no