LZW Implementation notes

Wed Apr 16 23:25:51 PDT 2003

First implementation in perl.  Very straightfoward coding it except for one
really annoying quirk with perl.  Since "0" is considered FALSE the following
has a false value when "0" is the value of the table's entry:

   $str = $table[$index]

The solution I used for this was to just check of $index was out of range
since all I wanted to know was weather $table was defined at index '$index':

   $index >= @table


Thu Apr 17 19:21:05 PDT 2003

Ported the rapid protoyped perl version of lzw encode to C.  I took advantage of
the search.h library for its hashing functions.  The documentation for it was
severly lacking in a few certain details namely that it doesn't copy the
data/key structures internally.  You have to do that and keep track of them
including the freeing of them should you want to move on within the same process
and use the hashing function again.  Since this is a run once situation it's
not a problem.  It has prepared me for the string situation I might find myself
in when porting lzw decode.  By having str[65535] and ch[2] and using the
string.h facility It shouldn't be a big deal adding strings to the lookup table.

The C implementation is lacking in documentation.  I should include the general
description of the LZW algorithm.  We'll see.


Sat Apr 19 23:06:52 PDT 2003

Need to free table entries if this is to be used as a library call.  It would
seem to clean up much of the compression code I should move a few things into
the hash abstraction (which is clearly lacking).  The hash table should be
aware of it's size and for fun the largest key size perhaps?  It should also
keep track of data and keys, since it lack an iterator, for later freeing.
*phew* A lot of work still to do.


Wed Apr 23 19:50:07 PDT 2003

Since it's going to be just as much work to patch hsearch as it is to just use
and tweak existing string hash code, I won't continue using search.h stuff.