"Raul Miller" <[EMAIL PROTECTED]> wrote:
> Here's a J expression which determines if value is _.
>
> isIndeterminate=:3 :0"0
> rep=.2(3!:1)y
> typ=.256#.a.i.rep{~8+i.8
> if.8~:typ do.0 return.end.
> bit=.(+/\1 12 e.~i.64) </. ,(8$2)#:a.i._8{.rep
> (*./1{::bit) *. +./2{::bit
> )
If something so very fundamental requires 6 lines of code,
it suggests that there is something quite wrong with the model.
I think that _. = _. would be much more sensible (and,
more importantly, much more useful). It would allow
many useful verbs such as _.&e. i.&_. etc.
= already has to have special code in it to handle _ and __
since comparison tolerance does not work with these;
the impact of also handling _. specially should be minimal.
It is not as though J hasn't changed the IEEE math defaults
to produce more useful results in several cases (0&* 0&% 0%0 etc.)
-- Mark D. Niemiec <[EMAIL PROTECTED]>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
