sources.txt - Don Yang (uguu.org)

09/15/01


This file describes the differences between sources in each directory.

./

   Final version, each file represents a preprocessing stage.  These files are
   write only, meaning it was very easy to write them, but near impossible to
   figure out what they were doing.  You are better off looking at one of the
   subdirectories.

v1/

   Original encoder, for proof of correctness.  This encoder does not implement
   the patching algorithm.

   Because of the use of fileno, encode.c and decode.c are not ANSI compliant.
   They pass lclint 2.5q otherwise (use +posixlib option).

   Both encoder and decoder runs without arguments.

   If the first command line argument is present and it's not '-', it reads
   input from the named file instead of stdin.

   If second argument is present, it writes output to that file instead of
   stdout.  If stdout is not redirected to a file (i.e. it's a tty device),
   program will run in debug mode.

   Reading/writing files is the preferred mode of operation, especially under
   windows, since it translates every \n to \r\n.  I tried pretty hard to
   reopen stdin/stdout in binary mode with no success :P

   Design for this version is covered in design0.txt.

v2/

   Second version, with the patching algorithm.  Same input/output parameters
   as above.

   Design for this version is covered in design0.txt.

v3/

   This version uses the current design, in design.txt.

   Comes with a few optimizations, and no more patch bytes.  Tag that was used
   for patch bytes are now used for Huffman blocks.  stdin/stdout are now
   opened as binary under win32.

v4/

   This version adds a superior substring matching algorithm.  Emperical tests
   showed this version to be about 5 times faster than last version ^_^

v5/

   Last human readable version.  Huffman tree is limited to 7bits, and sources
   are merged to one file.  Debug encoder/decoders are also removed (the source
   is once again ansi compliant).  There are no more algorithm changes after
   this version.

   Command line:
      kirika0 x                   Decode stdin to stdout
      kirika0 x infile            Decode infile to stdout
      kirika0 x infile outfile    Decode infile to outfile
      kirika0 a                   Encode stdin to stdout
      kirika0 a infile            Encode infile to stdout
      kirika0 a infile outfile    Encode infile to outfile

   'x' was chosen as the decompress switch since unrar and unace both use it.
   It was set to 'd' earlier, which has the association (for me at least) to
   delete files, while 'x' sounded more like 'extract', so there.  Program only
   recognizes lowercase 'x', any other flag cause it to run in encode mode.

huffman/

   Huffman encoder/decoder/tree generator.

tests/

   Correctness test scripts (Unix only!)