At the point where I need to handle data that is comprised of a mixture
of actual values and empty locations, the way I envision approaching it
in J would be to keep a marking array, a Boolean array with the same
shape. It might be kept sparse. Values in the main array that are marked
as empty would be handled differently. There might be a good deal of
overhead involved, but the algebra of it would be under the coder's full
control. There would be no risk that "magic" numbers would lead to data
corruption. There would, of course, be the real risks of data corruption
if the technique is not properly implemented.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
