🚀 go-pugleaf

On further reflection I have realized that my missing value 
problem was not well specified. Assuming the use of the
IEEE 754-1985 standard, it is relatively easy to 
make a missing value in C and pass it back to S using the
NAOK=T option: any of NaN values in C (i.e., any number starting
with 0x7FF) will be converted by S to NA. Therefore, the
1/a - 1/a trick works.

However, I would also like to be able to pass missing values to 
S using the call_S routine. This doesn't have a NAOK option, so it is 
necessary to create a missing value within C with the one
true value that S uses internally for NA, i.e.,
7FF0000100000000 

So the question is now: what is an elegant way of assigning 
7FF0000100000000 to a double precision number within C? Ideally
I would like to define a macro MISSING, which could be used
like NA in S. Maybe S already has something like this internally.

I don't claim to be a C expert; the best I can do at present is:

  double arguments[1];
  char *memloc;
  int c;

  memloc = (char *) &arguments[0];
  for (c = 0; c < 8; c++)
     *(memloc + c) = 0;
  *memloc = 0x7f;
  *(memloc + 1) = 0xf0;
  *(memloc + 3) = 1;


Gary Houston
Department of Statistics
Wellington, NZ
D