Fraser Jackson wrote: > It does not seem to me that this should double the amount > of code required. That probably depends on what you are comparing. The size of the J interpreter would not double. The code which is replaced would have to be replaced with code which functions similarly but works different -- it's hard to imagine why the new code should be expected to be radically smaller than the code it's replacing. If anything, I'd expect it to be larger -- you need a core operation which does the matrix multiply but skips calculations not needed for the symmetric case. And you also need code which propagates the symmetric values into their companion positions. But, being generous and imagining that the new code is no larger than the old code, you still need to maintain both copies -- unless you give up on the idea of supporting the multiplication of non-symmetric matrices) which is where the doubling comes from. It might also be worth thinking about what this kind of thing would mean when the left argument to . is not +/ or when the right argument is not * -- Raul
---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
