From: Matthew Miller <namille2@...>
Date: 2002-10-17T07:45:00+09:00
Subject: Re: Porting a CRC function to ruby

Hi Bill,

In article <aoiqu8$ib8$1@grapevine.wam.umd.edu>, William Djaja Tjokroaminata 
wrote:
> Hi Matthew,
> 
> The problem is not signed vs unsigned.  The problem is, given an int, the
> Ruby Fixnum is obtained by left shifting the int by one (and setting the
> least significant bit to 1).  Therefore, before you do your CRC, you have
> to right shift the Fixnum.  (Just be careful that the right shifted
> version of the Fixnum may not be a valid Ruby object, so "don't interact
> with Ruby" using the right shifted version.)
> 
> The problem is probably more complicated if the data type used in the C
> function is unsigned int or unsigned long.  In this case probably you have
> to use Bignum (and know how Bignum is implemented internally in
> Ruby).  However, as Ruby performs conversion between Bignum and Fixnum
> automatically, you have to be careful too.
> 
> Finally, because when you are dealing with communication packets the Ruby
> Fixnum is not really "clean", probably it is better and easier to use Ruby
> BitVector instead of Fixnum.

Thanks for the insight. I tried my suggestion of masking off all but the 
lower two bytes and the CRC works with the packets I've tested it with. 
Still I wish that ruby had an unsigned integer class.

BTW, why does Fixnum left shift by one the integer it stores? I can't think 
of any reason why that would be done.

Thanks, Matthew

-- 
"Man k�nnte froh sein, wenn die Luft so rein w�re wie das Bier"
"One could be happy if the air were as pure as the beer"