Thanks Laurent, > Multiprocessing is way to go (pickling was implemented to make it easier, > and there a google SoC project out bring storage-based arrays in). I think you're likely right...
> R GUI builders could clearly and immediately benefit from having the GIL > released, but in the meanwhile adopting a client-server strategy is the > recommended way to go. That has limitations but also advantages; building a > minimal server can be done in less than 20 lines of code (less than 10 lines > if no comments). > http://rpy.sourceforge.net/rpy2/doc-2.1/html/server.html#simple-socket-based-server-and-client > > Hoping this helps, Indeed it has: I have experimented with both Multiprocessing and the client-server implementation. The multiprocessing route seems to work quite well, however, I'm running into two problems, and I'm wondering if anyone else has managed to get this working: 1) As expected, if a plot is generated in a separate process, it is immediately destroyed once the process completes. I have tested this, and basically I get a plot for about half a second, then it disappears. Does anyone know of a way to get around this? 2) Assignments do not seem to span process' (again this is probably expected, and I'm just simply doing something wrong). If I do something like 'robjects.r("test <- c(1,2,3,4,5)")' in a subprocess, then 'test' is *not* available in the parent process (whereas it *is* obviously available in the subprocess). Does anyone know what I need to do to make this work properly? Here is a minimal example to illustrate issue 2 above (issue 1 can similarly be tested by changing 'cmd' to "plot(c(1,2,3,4,5))") : # -*- coding: utf-8 -*- import sys, os from multiprocessing import Process, Queue import rpy2.robjects as robs def r_command(q, cmd): q.put(robs.r(str(cmd))) print "Process %s output:" % os.getpid() print robs.r.ls() if __name__ == "__main__": q = Queue() cmd = "test <- c(1,2,3,4,5)" p = Process(target=r_command, args=(q,cmd,)) p.start() p.join() print "Process %s output:" % os.getpid() robs.r['print'](q.get()) print robs.r.ls() -- Carson J. Q. Farmer ISSP Doctoral Fellow National Centre for Geocomputation National University of Ireland, Maynooth, http://www.carsonfarmer.com/ ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ rpy-list mailing list rpy-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rpy-list
