STATIST(1)STATIST(1)NAMEstatist - calculate Huffman distribution for freeze(1)SYNOPSISstatist [ -gx... ]
DESCRIPTION
The default table is tuned for both C texts and executable files (as in
LHARC). If you will freeze any other files (natural language texts,
databases, images, fonts, etc.) you can calculate the matching posi‐
tions distribution using the `statist' program, which calculates and
displays the mentioned distribution for the given file. It is useful
for large (100K or more) files.
Though the built-in position table is polyvalent, the tuning can
increase the compression rate up to one additional percent. (Observed
mainly on text files.)
USAGEstatist [-g...] < sample_file
or
gensample | statist [-g...]
where `gensample' is a program generating some sample stream of bytes
similar to files to be frozen.
The -g and -x switches have the same meaning as for freeze(1) and may
be repeated.
You can also see the intermediate values and watch their changes by
pressing INTR key when you wish.
Note: If you use gensample | statist , remember that INTR influence
BOTH processes !!
The results have the following format:
n1 n2 n3 n4 n5 n6 n7 n8 (uncertainty = x)
Average match length: xx.yy
Percentile 99.9: p999
Percentile 99.5: p995
Percentile 99.0: p990
Percentile 97.0: p970
Percentile 95.0: p950
Percentile 90.0: p900
Percentile 80.0: p800
Percentile 70.0: p700
Percentile 50.0: p500
Sigma: xx.yy
Here n1 - n8 are values of the calculated position table elements,
uncertainty is a number which denotes validity of given results (non-
zero values of uncertainty indicate that the results may be unusable).
Other values (average match length, percentiles and sigma) are FYI
only.
You may create the /etc/default/freeze file (if you don't like
/etc/default/ directory, choose another - in MS-DOS it is FREEZE.CNF in
the directory of FREEZE.EXE), which has the following format:
name = n1 n2 n3 n4 n5 n6 n7 n8
(name must start in column 1). For example:
---------- cut here -----------
# This is freeze's defaults file
russian=0 0 1 2 6 20 31 2 # The sample was mailx.lp (Russian)
english=0 0 1 2 7 16 36 0 # The sample was gcc.lp (English)
# End of file
---------- cut here -----------
If you find values, which are better THAN DEFAULT both for text (C pro‐
grams) and binary (executable) files, please send them to me.
Important note: statist.c is NOT a part of freeze package, it is an
aditional feature.
SEE ALSOfreeze(1), melt(1), fcat(1)DIAGNOSTICS
Huffman tree has more than 8 levels, reducing...
Self-explanatory, but sometimes reducing falls into infinite
loop.
xxxK
Progress indicator is written after each 4K of a file pro‐
cessed.
BUGS
Sometimes use of the results with uncertainty = 1 (on a file) gives
compression rate worse than default but use of the results with uncer‐
tainty = 13 (on other file) works quite good.
Found bugs descriptions, incompatibilities, etc. please send to
leo@s514.ipmce.su.
local STATIST(1)
