Roger Hui <[EMAIL PROTECTED]> wrote:
> If _. is removed altogether and you get a NaN in 
> your data (from external sources), that makes the
> data very nasty to deal with, and imposes an overhead
> on other data that has nothing to do with NaN. e.g.
>
>    wd 'blah blah blah0'
> |NaN error
> |       wd'blah blah blah'
>    wd 'blah blah blah1'
>    NB. works, but J has to check data for NaN

> The indeterminate _. is a numeric atom. It is provided 
> to aid in dealing with NaN (not a number) in data from 
> external sources, and should be removed from such data 
> as soon as possible.  ...

I wasn't suggesting that _. be removed from the language;
it seems that in 6.02 this is already being done in part.
I was merely suggesting that the spelling '_.' be removed
from the number parser. so there was no 'easy' way for a
J prgram to create NaNs. This is currently already the
case for -NaN, and all the other flavors of NaN.

In this way, NaN values from external sources can still occur,
but the only way NaNs could be explicitly created by J software
would be using 3!:2 (for rare cases where NaN is REALLY necessary)
As I undertand it, J primitives in 6.02 can no longer
generate NaN results (other than by copying).

-- Mark D. Niemiec <[EMAIL PROTECTED]>

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

Reply via email to