[R] Controlling R from Java

[Duncan Temple Lang]

> Date: Tue, 12 Dec 2000 15:38:16 -0500
> From: Marc Grushcow <Marc.Grushcow at ntt.ca>
> Reply-To: Marc.Grushcow at ntt.ca
> Organization: NTT SYSTEMS INC.
> Sender: owner-r-help at stat.math.ethz.ch
> Precedence: bulk
> 
> I'd like to run R (LINUX) under Java. I can create the process and connect to
> stdin and stdout. I can send commands in and get output back.  I don't know how
> to determine when the output is complete - there is no trailing "prompt".
> 
> For example, if I send "ls()\n" what I get back is:
> > ls()
> character(0)
> 
> Each of the above two lines is terminated with a 0x0A. I would have expected
> (liked)...
> ls()
> character(0)
> > 
> 
> Then I could parse for the "\n> " and know when the output is complete. The
> problem with checking for data availability on the input stream is that it is
> unreliable. R's output buffer might flush unpredictably and/or the task
> switching OS might cause R to stop after one buffer is full but before another
> is completed.  In any case, it is the usual problem of one program trying to
> parse another's output in a time sharing environment.


This is a good point and just one of the reasons that communicating
between two systems using strings and pipes, etc. can be a lot of work
and still quite unstable and non-portable.  It is also one of the
reasons that we have spent some time in the Omega project
(http://www.omegahat.org) coming up with a general way to allow
interpreters such as Java, Python and Perl to be embedded within R and
vice-versa. 

I hesitate in suggesting a totally different approach for your
application, especially since you may have invested some time already
and have other constraints. But here goes.....

You can use the Java package (http://www.omegahat.org/RSJava) to get
communication between an R and a Java interpreter.  You can run an R
session and execute Java code within it (directly from the R prompt),
or alternatively run a Java application and run a (non-interactive) R
session within it and evaluate R expressions.  I have just added
facilities to make the calling R from Java simpler and wrote some
notes which are available from

    http://www.omegahat.org/RSJava/RFromJava.pdf

Not only does this avoid the flushing problems you observere, but it
also avoids having to parse the output generated by R and preserves
much more information in the communication step.  Also, there is a
general mechanism to add ones own converters to translate Java objects
to R and vice-versa.

I hope this gives you another approach to consider for your
application.  If you have any questions, you can send me mail directly
or to the omega development mailing list (omega-devel at omegahat.org).

D.

> 
> I'm sure that this problem has been solved before.  Any suggestions would be
> appreciated.
> 
> Thanks
> Marc
> -- 
> =.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.
> mailto:Marc.Grushcow at NTT.ca		http://www.NTT.ca
> =.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.=.


-- 
_______________________________________________________________

Duncan Temple Lang                duncan at research.bell-labs.com
Bell Labs, Lucent Technologies    office: (908)582-3217
700 Mountain Avenue, Room 2C-259  fax:    (908)582-3340
Murray Hill, NJ  07974-2070       
         http://cm.bell-labs.com/stat/duncan
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._