The argument about tolerance started when APL first allowed =. Fortran has
always (at least through F77) always said total equality.

For = there is no setup that won't create some crazy results somewhere
(except total equality). The decision with early APL allowed quite large
tolerances, sometimes quite convenient. I can understand why the decision
was made that says different integers mustn't be =. But the convenience of
fuzz in = to allow for numeric round-offs is well worth the price.

If you wish more than that you should expect to handle it yourself, we can
make J (and APL) hopelessly slow if we try to be too clever.

Ralph Selfridge
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to