Le 12/05/2010 12:00, Nicolas Dumoulin a écrit :

After a quick debug, I can see that the problem is with this line: rexp.getAttribute(RInstructions.ATTRIBUTE_NAMES).asStrings()

The NPE occurs because rexp.getAttribute(RInstructions.ATTRIBUTE_NAMES) is null with - RInstructions.ATTRIBUTE_NAMES = "names" - and rexp is a REXPLanguage instance containing - a REXPSymbol "-" - a REXPDouble "1.0"

Thanks, I had a look to the code in R, and I got the attribute "names" containing a list of String : $names [1] "model" "M" "s" "omega" "X" "call" I will try to go further, but as nuiton-j2r do not modify the rexp return from R (and gotten from jri or rserve libraries), i guess this could be also a bug from there. Jean

Le 12/05/2010 12:00, Nicolas Dumoulin a écrit :

After a quick debug, I can see that the problem is with this line: rexp.getAttribute(RInstructions.ATTRIBUTE_NAMES).asStrings()

The NPE occurs because rexp.getAttribute(RInstructions.ATTRIBUTE_NAMES) is null with - RInstructions.ATTRIBUTE_NAMES = "names" - and rexp is a REXPLanguage instance containing - a REXPSymbol "-" - a REXPDouble "1.0"

Just a question, which version of j2r are you using ?

Le 12/05/2010 12:10, Nicolas Dumoulin a écrit :

Le mercredi 12 mai 2010 12:05:52 Jean Couteau, vous avez écrit :

...

Ok, it is trying to convert the result and got a NPE trying to do it. As you do not use the result from R in your code, you could replace the eval() method by the voidEval() one (that do not try to return the result from R and so do not make the conversion).

The give code is an example, but I need the object indeed ;-)

I guess that what you want is the experiment design so if you call your R object a, the a$X data.frame. What you could do is something like : engine.voidEval("a<-fast99..."); RDataframe myExperiment = (RDataframe)engine.eval("a$X");

Maybe, I can try to simplify the object in R script to help j2r magic convertor…

It is always better to get the basic types from R instead of complicated custom R objects (such as a fast99 object ;) ) In J2R, you can work with data.frames, lists, or the basic java types. Hope this helps, Jean

Le 12/05/2010 12:31, Nicolas Dumoulin a écrit :

Le mercredi 12 mai 2010 12:23:38 Jean Couteau, vous avez écrit :

...

I guess that what you want is the experiment design so if you call your R object a, the a$X data.frame.

Yes, it was I mean. But, as you may know, I need then the complete returned object for the analysis phase with the tell() method. So, I can store the result in a R variable, and fetch only $X, but as I use Rserve, I'm not what will happen if two users make a fast99 exploration on the same server…

You are safe this way, because on Rserve, you have a session per user, so you can assign variables without any problem and can use the tell method on the stored object, it will never be altered by other users. But you must always use the same REngine instance to stay on the smae user session.

...

Hope this helps, Jean

Thanks