From: gtoal@tharr.UUCP (Graham Toal)
Newsgroups: alt.sources,alt.sources.d
Subject: bugfix for char '0' in dynhuff.{c,p}
Message-ID: <956@tharr.UUCP>
Date: 6 Sep 90 10:48:46 GMT

Archive-name: dynhuff.bug

A few days back I posted a huffman-coding program here; I've since
been playing with it to try modifying the algorithm for delta
compression; while I was doing that I discovered that it only worked
correctly for 1..255; 0 went wrong.  Here's a filthy fix -- I've
extended the alphabet by 1, and every time I store what should be a
zero, I store 256 instead.  It's a 1-1 mapping, so reverses OK.

The same hack works with both Pascal and C versions.

(Sorry about the diff - must get a real one sometime...)

Diff files 'huff.c' and 'huff2.c'
------------------------------------------------------- top level declarations
change c.huff, line 29 to 29
#define  n  256
to
#define  n  257        /* 0 doesn't work, so cheat :-) (change 256 to 257) */
------------------------------------------------------ EncodeAndTransmit(int j)
after c.huff line  108: 
add c.huff2 line  109:
   if (j == 0) j = 256;                           /* 0 -> 256 */
-------------------------------------------------------- ReceiveAndDecode(void)
change c.huff, line 161 to 161
   return(alpha[q]);
to c.huff2, line 163 to 164
   i = alpha[q]; if (i == 256) i = 0;             /* 256 -> 0 */
   return(i);
----------------------------------------------------------------- Update(int k)
after c.huff line  290:
   /* Set q to the node whose weight should increase */
add c.huff2 line  294:
   if (k == 0) k = 256;                           /* 0 -> 256 */
-------------------------------------------------------------------------------