MAN page from Other tzip-1.12-1.i386.rpm

TZIP

Section: User Commands (1)
Updated: local
Index

NAME

tzip, tunzip - compress or expand files

SYNOPSIS

tzip[ -Vbcdhpqtv123456789 ][name ...]
tunzip[ -Vcfhpqtv ][name ...]

DESCRIPTION

tzipattempts to compress files by passing data through a pipeline that lookslike this:

RLE -> BWT -> MTF -> ARITH

RLE- Run-Length Encoder

BWT- Burrows-Wheeler Transform

MTF- Move-to-Front coder

ARITH- DCC95-stylearithmetic coder

Due to the nature of theBWT,compression is improved when data is processed in larger blocks, subjectto the law of diminishing returns. Larger block sizes causetzipto use more time and space. The default block size(256K)provides a reasonable compromise between compression and time/spacerequirements. Here is a table that summarizes memory usage and compressionperformance for each block size.

                      Compress   Decompress  Corpus
  Switch  Block Size  Allocated  Allocated   Bits/Byte
  ----------------------------------------------------
    -1      64K        812K       684K        2.39
    -2      128K       1324K      1068K       2.28
    -3      256K       2348K      1836K       2.20
    -4      384K       3372K      2604K       2.15
    -5      512K       4396K      3372K       2.15
    -6      640K       5420K      4140K       2.13
    -7      768K       6444K      4908K       2.11
    -8      896K       7468K      5676K       2.11
    -9      1024K      8492K      6444K       2.11

Compression did not improve after-7because no file in theCalgary Corpusis larger than 768K.

tzipemploys the same algorithms asbzip,and therefore the two compressors will achieve very similar compressionratios. The author ofbziphas taken greater pains in optimization than I have, and thereforebzipis slightly faster and more appropriate for practical use. However,tzipdoes employ the most important optimizations, and it is fast enough forpractical use. I have placed a greater emphasis on readability of thesource code, in the hope thattzipwill be of greater use for the purpose of educating people about thesewonderful algorithms (especially theBWT).

OPTIONS

-V: Display version number and exit.
-b: Set block size in kilobytes (i.e. if you want a block size larger than1MB).
-c: Write to standard output. The input file (if any) will not be unlinked. When compressing, the standard output must not be a tty, and only one file will be processed.
-d: Decompress. tzipwill compress by default.
-f: Force over-writing of existing files compression of symbolic links.
-h: Display a help message and exit.
-p: Preserve (don't unlink) input files.
-q: Be quiet -- don't print warning messages.
-t: Test compressed file integrity.
-v: Be verbose. Per-file and cumulative compression statistics will be shown.

REFERENCES

I wish I could take a lot of credit fortzip,but unfortunately it contains no algorithms of my own design. The followingpeople deserve the credit:

Julian Seward, author of bzip
Some implementation details were adapted from bzip.

  Michael Burrows and D. J. Wheeler:
   "A block-sorting lossless data compression algorithm"
   10th May 1994.
   Digital SRC Research Report 124.

  Peter Fenwick:
   "Block Sorting Text Compression"
   Proceedings of the 19th Australasian Computer Science
     Conference, Melbourne, Australia.
     Jan 31 - Feb 2, 1996.

  Alistair Moffat, Radford Neal and Ian Witten:
   "Arithmetic Coding Revisited"
   Presented at the 1995 IEEE Data Compression Conference,
     Snowbird, Utah, USA in March 1995.

  David Balazic <david.balazicAATTuni-mb.si>:
   Submitted a patch for improved compression performance.
   Thanks David!

This document was created byman2html,using the manual pages.