It is undefined. Operations on nan are not guaranteed to produce anything
sensible (except structural operations and 128!:5).
On Fri, 6 May 2022, Raul Miller wrote:
But what should the result of _. +. 1 be, if not NaN?
Thanks,
--
Raul
On Fri, May 6, 2022 at 2:46 PM Henry Rich <henryhr...@gmail.com> wrote:
USING a NaN is not an error. The result is undefined (sc whatever the
hardware produces). CREATING a NaN is an error.
_. = 4
0
_. + 4
_.
_ % _
|NaN error
| _ %_
1 +. _. is not a Boolean operation. It's calculating the LCM.
NANTEST calls the OS to see if any NaN-creating operations have been
performed since the most recent NAN0.
Henry Rich
On 5/6/2022 2:35 PM, Raul Miller wrote:
> Hmm... actually, there's a bunch of tests like this already there,
> with comments that NaN errors are treated as zeros in this context.
>
> Which suggests that I was wrong about what the fix should be.
>
> I'll shut up now...
>
--
This email has been checked for viruses by AVG.
https://www.avg.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
