kexis-announce Mailing List for Kexis

Just to let you know I started hacking away on kexis again. Current CVS has morphed a bit since the last release. New fetures include accepting and outputting data from/to stdin/stdout by default. This makes Kexis a true unix filter. You can still operate on batches of files like before. :P

Also fixed some misc bugs, rewrote the option handling to use getopt so it behaves.

I'm adding a a block checking algo next to verify that files dont get mangled. The options are 32 bit CRC and md5sum. I'm thinking I'm gonna go with md5sum because its more robust.

Wayde

Lots of things have changed. The most noteworthy is that Kexis can now
compress/decompress multiple files at a time, has a more robust WAV file parser, and is
MUCH faster. Again, you must decompress 0.1.5 .kxs files with 0.1.5 and
recompress them with 0.2.0

Change log:

                            * Fixed a case in the bit manipulation code where
zero length data is not manipulated then never written.
                            * Combined the progress bar code into one
function.
                            * Progress bar now gets called 4 times a second as
long as the blocks are processed 4 times a second. If its slower, the progress
bar gets
                            called when the block finishes.
                            * Commented out some debug stuff for speed.
                            * Completely rewrote the bit manipulation routines.
Kexis now stores its data as longs and not shorts in its native file format.
this is great
                            for speed, but unfortunatly breaks old files.
                            * Added the ability to compress or decompress
multiple files from the commandline.
                            * More speed tweaks. Kexis should now run 2x to 3x
faster than 0.1.5.
                            * More robust Wav file parsing. Kexis now tries to
understand non canonical files. It succeddeds as long as there is 1 and only 1
fmt chunk,
                            and data chunk. If the data chunk comes before the
fmt chunk Kexis will fail. If there are more than 1 data chunks, kexis will only
encode the
                            first data chunk.

This release fixes the bloat bug in 0.1.4 An integer divide creeped in where a
floating point divide was supposed to be, causing rounding errors. 0.1.4
Produced LOSSLESS data, but at a non optimum compression rate. Please decompress with
0.1.4 and recompress with 0.1.5. To keep this from hapening again, There will be
no new releases until kexis undergoes a battery of release tests.

Thanks for your patience. 

Wayde

New Compressor which is tighter. It uses more history to determine optimum K
size. Its setable. Also a new predictor that doesnt preform as well as i thought
it would. It calculated the optimum prediictor based on what the optimum
predictor was on the last value, then uses that. Its not as good as a typical straight
2nd Order predictor.

NOTE: there was a small bug in the file format of 0.1.3 and before. This is
fixed in 0.1.4. However you MUST decompress any kexis files with 0.1.3 and
re-encode them with 0.1.4. Not only will they be smaller, they wont be buggy anymore.
:P

Btw, Kexis is now approaching shorten in file size. In fact, it beats it with a
shortne block size of 256, and a order of 10. You can manually adjust shorten
to still slightly edge out kexis though.

Wayde

Fixed ithe compile problem people were having with low compiler optimizations.

Use and abuse it, send back more bug reports. :)

Wayde