On Mon, 5 Apr 2004, Spencer Graves wrote:
> What tools and documentation exist for producing maps and working
> with polygon and polylist objects in R beyond the following:
>
> * Packages maps, mapproj, maptools, and mapdata.
>
> * Becker and Wilks (1993, 1995), listed as references in the
> documentation on the "map" function.
>
Packages maps, mapproj and mapdata relate to the porting to R of the code
to draw maps, and the geographical data base format described in the
Becker and Wilks references. To quite a large extent, the user is
restricted to using the polygons from the provided databases.
Alternative approaches try to allow for users importing foreign formats,
often from geographical information systems - for polygons, these are
vector formats. The thematic area is broad, and the code there is
(packages maptools and shapefiles on CRAN) can only be used with ESRI
(ArcGIS/ArcView) shapefiles, with the exception of RArcInfo (on CRAN),
that supports legacy ArcInfo topological vector formats. There are other
resources mentioned on the R spatial projects website (on Related Projects
on both the main R-project and CRAN navigation sidebars):
http://www.r-project.org/Rgeo
In addition, work is in progress to create S4 classes for spatial data -
alpha source packages are available on :
http://sourceforge.net/projects/r-spatial/
We are hoping to present something on the spatial "island" in Vienna
at
useR! 2004 in May, so good ideas and suggestions are very welcome.
The key (unanswered) question is whether (or how far) the treatment of
polygons should be simple or topology-based: shapefiles are simple, and
all the polygons are seen as independent rings, while "maps" uses an
underlying topology model, in which polygons are built on the fly from
lists of directed edges. Doing topology right is probably more a GIS
operation than a data analysis operation.
These things also get broader discussion on a more specialised mailing
list:
https://www.stat.math.ethz.ch/mailman/listinfo/r-sig-geo
Hope this helps - the more input and help with user requirements, typical
input and output formats, (and maybe code too?), the more likely we are to
move this forward.
Roger Bivand
PS. It would be really helpful if someone/anyone familiar with getting
SJava to work was willing to look at http://www.geotools.org and help
establish/investigate the magic words needed to use a solid and promising
library from within R.
> Thanks,
> Spencer Graves
>
--
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Breiviksveien 40, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 93 93
e-mail: Roger.Bivand at nhh.no