The rule is, if one argument is extended and the other
is a 64-bit float, both arguments are converted to
64-bit floats.  For the expression  _ > 10^1000x  ,  
_ is a float but 10^1000x is not, so 10^1000x is 
converted to a float, in this case _ , and _ > _ is 0.  
So the result of _ > 10^1000x  is 0.  This is not 
correct.

I believe the solution to this problem is to signal 
error when an extended integer or rational with
magnitude larger than the largest finite 64-bit
float (see below) is encountered in the conversion.

   0j_16 ": 3!:2 (}: 2 (3!:3) 0.1), '7fe',13$'f'
1.7976931348623157e308



----- Original Message -----
From: Andrew Nikitin <[EMAIL PROTECTED]>
Date: Tuesday, June 13, 2006 4:49 pm
Subject: [Jbeta] infinity and extendeds

>   _ > 10^1000x
> 0
>   (x:_) > 10^1000x
> 1
>   9!:14''
> j601/beta/2006-05-20/22:30


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

Reply via email to