"Miller, Scott L" wrote:
>  
>  > -----Original Message-----
>  > From: Nicholas Clark [mailto:nick@[...].org]
>  >
>  > Not sure if this is a sane idea. Can you
>  >
>  > define the width of a negative integer to be the smallest
>  > number of bits
>  > needed to represent it in 2s complement.
>  > So -9 (ie 1111111111110111) is 10111 (width 5)
>  >
>  >    -6 (ie 1111111111111010) is 1010 (width 4)
>  >
>  > -6 & -9 needs to promote -6 to width 5, [same as positive
>  > case] but do the
>  > promotion as sign extension (11010) then & the two together
>  >
>  > 11010 & 10111 == 10010
>  >
>  > which is -14. And gives the same answer as 16 bit and 32 bit
>  > fixed precision,
>  > as best I can tell.
>  >
>  > [not tested very much for other values or for | or ^ operators]
>  >
>  > Nicholas Clark
>  
>  How would you be able to tell if the number was intended to be negative?
>  11010 = -6 or 26, 10111 = -9 or 23...
>  
>  -Scott

All discussion of binary notation is pointless; BigInt's are stored as 
an array of Base-10**7 numbers.  We have to figure out how to handle 
the logical operators in a sane fashion with unlimited length integers
stored in a completely non-binary fashion.

Anyone with me that we should deprecate the logicals in M::BI completely 
and point people at Bit::Vector if they want to do unlimited length 
boolean math!

John

--
John Peacock
Director of Information Research and Technology
Rowman & Littlefield Publishing Group
4720 Boston Way
Lanham, MD 20706
301-459-3366 x.5010
fax 301-429-5747