HUFFMAN CODING
----------------

Compilation Instructions:
-------------------------------------------------------------------------------
Both huff.c and dhuff.c #include tree.h.  huff.c and dhuff.c are both 
independent programs and each should be compiled together with tree.c, making 
sure that the math library (math.h) is properly included by the compiler.  
The source has compiled under Sun/Solaris using gcc and lcc and as a SIOW 
application under MacOS (the code needs to be modified due to the inability to 
pass command-line arguments to a MacOS app.) using the latest MPW-GM release
from Apple.

For gcc:
gcc -lm huff.c tree.c
gcc -lm dhuff.c tree.c
--------------------------------------------------------------------------------

huff.c has been tested on both ASCII text and binary files and appears to work.  
It works by constructing Huffman trees, which is actually NOT the most efficient 
way to perform Huffman coding, but it is the obvious way to go about it.  The 
code has acceptable running times on relatively large files - i.e., the Project 
Gutenberg edition of the King James Bible, which is about 4.1 MB in size - and 
decoding is about 5 times faster than encoding.

Decoding is accomplished by reconstructing the original Huffman tree from the 
original frequency data, which is embedded at the top of each compressed file.  
This information is stored in 5 byte chunks - one for each character appearing 
in the source file.  I picked this size because it easily stores all the 
frequency data for the Bible, which is a rather large text file.  If you have a
file with frequencies larger than 4 bytes, you're in trouble.  This could be 
fixed by scaling frequency counts to fit into one byte - just map the highest 
frequency to the number 255 and scale accordingly.