Adding arima.c, make build work

[tninja]

The arima.c get copied here, it include ARIMA_transPars function. C_ARIMA_transPars is defined in renjin.R. The build is working now, but for some reason the arima() report ERROR: Could not resolve native method ARIMA_transPars.

@akbertram, do you have any clue what I missed? Thanks in advance.

[bddbot]

Can one of the admins verify this patch?

[akbertram]

@bddbot add to whitelist

[akbertram]

Looks good!

I think you are probably missing an entry in the stat package's NAMESPACE file.

Can you also add a simple test case for the arima function? See test.stats.runmed.R for a simple example.

Thanks again!

[tninja]

Not quite easy..

After adding stuffs inside arima.c to NAMESPACE, now it report:

arima(lh, order = c(3,0,0), optim.method="Nelder-Mead") ERROR: Sorry! polyroot not yet implemented! java.lang.ClassNotFoundException: org.renjin.primitives.R$primitive$polyroot

Found a do_polyroot function inside complex.c of R source code: SEXP attribute_hidden do_polyroot(SEXP call, SEXP op, SEXP args, SEXP rho)

Not sure how to get this thing work, I don't even know where to put complex.c. Do you have any clue?

[akbertram]

Yes, this is a bit trickier. polyroot is an internal base function that basically has two parts:

1. R_cpolyroot and its subroutines constitute the actual C implementation of the algorithm and is largely independent of any R types or methods. It can be added to the math/nmath module or perhaps we can create a new math/complex module.

2. The do_polyroot() is the R wrapper for R_cpolyroot. Because it's in the base package, it's difficult to compile the C code directly, so I'd suggest just porting it to Java. In Renjin, internal functions are implemented as static methods and registered in the Primitives class. See the complex function as an example.