[Rd] Best Practise

Tom McCallum termcc at googlemail.com
Mon May 14 10:23:02 CEST 2007


Just a quick question on best practise.  I am converting quite a bit of  
legacy C code into R packages and have the following situation:

(1) Legacy object with a double* array in, all over code so don't want to  
change any more than I have to.

(2) Do something like:
	SEXP arrayToPassToR;
	PROTECT( arrayToPassToR = allocVector( REALSXP, n ) );
	for(i=0; i < n; i++) {
		REAL(arrayToPassToR)[i] = oldCarray[i];   <----  very slow way to copy  
data, can I use memcpy/pointer assignment here to remove the loop without  
running into garbage collector?
	UNPROTECT( arrayToPassToR );
	SEXP returnValueFromR;
(3) Have made it to call back to an R function which returns a new /  
different SEXP double array.
	returnValueFromR = Test_tryEval(...);
(4) Copy back to oldCArray
	for(i=0; i < n; i++) {
		oldCarray[i] = REAL(returnValueFromR)[i]; <--- can I use memcpy/pointer  
assignment here to remove loop?

I have done the long winded copy as I am a bit paranoid about the garbage  
collection.  My question is is it legitimate to do the following

	double* oldCArray = REAL(arrayToPassToR);
	UNPROTECT(1); // where the 1 would remove protection from arrayToPassToR  
and potential garbage collection
	-- assume arrayToPassToR was garbage collected by R --
	Rprintf("%f", oldCArray[0]);

because if arrayToPassToR is garbage collected then oldCArray will cause a  
SEGFAULT when it is accessed afterwards, won't it?

Many thanks


PS Note this is the new email address - delete term at blueyonder.co.uk as it  
won't work soon!

More information about the R-devel mailing list