[aplusdev] RE: A+/R
Alexander Skomorokhov
askom at obninsk.com
Wed Dec 10 00:01:30 EST 2003
Hi Dan,
Sorry for a delayed reply. We just elected so awful parliament that I
haven't turned on
computer for a couple of days:-)
What is the model for the A+/R integration your team is
currently working on?
We do nothing about this in general. I mean as a general
approach to A+R integration (as we have done for Windows Dyalog APL+R
integration). We are waiting until somebody clever with C knowledge will
do this.
Meantime we do some specialized integration. First of all about tree and
rpart
libraries (Classification and Regression Trees). We write our data from A+
to a
Linux file and start from A+ a script. This script starts R in batch mode,
reads
data, build a tree and writes the result in a flat file easy to read in
A+. In A+
the function reads this file, process it and build a classifier on the
base of tree learnt in R.
This classifier may be used in A+ as a normal function. All the steps
above also
made as an A+ function call. So, it is not a "real" integration, but it
simplify using
R trees from A+.
Is it "A+ over R", i.e. A+ code can call R code, but not
viceversa?
Yes, we are A+ centric. What we dream about is to only call R functions
from A+ and
pass data in both directions, of course.
Or is it "A+ under R", i.e. R code can call A+ code,
but not viceversa?
Let R people to do this if they want to add the real thing like A+. (I
think they are completely
happy with R).
Or is it "peer level", i.e. A+ code and R code can call
one another as a true coroutine system?
See above.
What about threads? Do A+ threads map 1-1 to R threads,
or is the model more flexible?
Unfortunately we have no model at all. I want it to be even "not
flexible":-)
What about exception handling? What happens if A+ code
calls R code, and the R code triggers an exception? Is
the exception handled in R, in A+ or in both?
See above.
I see that more and more people from our list are interested in A+ and R
integration. And I am very glad to see it. A+ is the brilliant language
and product,
But together with _myriad_ of statistical and data analysis methods from R
it would be just the best tool for data mining.
Regards,
Sasha.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://berdom.net/pipermail/apluslist/attachments/20031210/7d5c0ec2/attachment.html
More information about the apluslist
mailing list