Using SJava?
I'm just starting to look at java. The big win is people report 100% productivity (lines per hour) increase over C++. Eclipse is a great tool for learning java. I was using vector/list classes within minutes (you can easily find object in the base libraries by browsing or with autocompletion). I agree, performance is less of an issue these days. In the long run, if distributed vm's become available that implement transparent thread migration, then java progs will get really easy to scale-up to large problems (modulo the normal caveats about the problem domain). Someone should really port R onto the java vm. (Not optimal for R-like languages, but the ease of integrating with java, jpython, etc. might make it worth while.) Tim
On Fri, 2004-04-02 at 12:56, Roger Bivand wrote:
Is anyone on this list using SJava or any R/Java connection? Not only is the JTS Topology Suite that Tim just mentioned interesting, but so is the upcoming GeoTools2 at http://www.geotools.org/. Both are likely to be used quite a lot, and thus probably well-maintained. At the recent Association of American Geographers meeting, Frank Hardisty asked me about this, and while I was able to install SJava (from http://www.stats.ox.ac.uk/pub/RWin) on a Win XP laptop, we were not able to see how to get Java within R to see GeoTools in its classpath. Both of these Java resources are potentially useful, and R opinions two or three years ago, that Java is slow, may need revision given increased machine capacity. I feel we could benefit from mobilizing Java insight (but I feel personally Java-challenged!). I believe that Duncan Temple-Lang was playing with/working on object discovery - this could be very relevant in terms of matching R-internal object representations with those in existing software that we could - if SJava worked - link to. I also think that we could ask him for advice if we discussed it first here. What do others think?
Timothy H. Keitt Section of Integrative Biology University of Texas at Austin http://www.keittlab.org/