Henry Rich wrote:
> Right.  But he wanted a conjunction to replace !. , and that
> conjunction would have to look at its u, which could be anything,
> and come up with the tolerant version.

But what would the "epsilon" argument mean?

Conceptually, epsilon should always be so small it doesn't matter.
So it could be argued that any significant algorithmic dependence 
on the specific size of epsilon is a problem.

Of course, the above assumes we're working with infinite precision
numbers, and computers are very finite (which is the reason 
tolerant comparisons were introduced in the first place).

> This is even harder than rewriting dyad i. , because u could be
>    >
> or
>    [ > ]
> or
>    ] ] [ > ]
> or...

(>=<->)

But there's an easy way to deal with all these cases: throw a
nonce error for any non-primitive verb (or, more specifically,
for any unhandled verb).

-- 
Raul

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

Reply via email to