Here is a silly (maybe only in my mind) way to accomplish my goal:
A <- "This is a test."
B <- "This is the answer."
swap <- function(item1,item2) {
  tmp   <- item2
  item2 <- item1
  item1 <- tmp
  return(list(item1,item2))
}
AB <- swap(A,B)
A <- AB[[1]]
B <- AB[[2]]
But I'm considering something similar to the C code following:
void swap(int *a, int *b)
{
    int iTemp ;
    iTemp = *a;
    *a = *b;
    *b = iTemp;
}
My motivations:
- My real data is quite large, e.g. 5k*5k matrix, so the assignment of the existing variable in the iteration twice, inside the function and outside the function, must be time squandering.
- The closest question on the SO is this one, but just like the OP in the question, my R session also has lots of objects: I'm working with Rmpi, and each slave will have a great number of variables.
- In my humble opinion, Ris written inC, soRmay have pointers likeCdoes, while I can't find much on the net surprisingly.
 
     
    