In my experience, J works best when you can map your entire data set into memory. Its primitives will all implicitly iterate over in-memory data. Hypothetically speaking, this could be virtual memory, but I suspect that you would run into some currently undocumentded performance issues if you push J heavily into swap.
However, with not too much extra work, with many applications you can arrange your data into good sizd blocks (some relatively small fraction of your total available memory, to leave room for intermediate results), and deal with large data sets that way. Whether this suits you, or not, depends on you, and your application(s). -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
