bzip2(1)

bzip2(1)                                  bzip2(1)



NAME
       bzip2, bunzip2 - a block-sorting file compressor, v1.0.4
       bzcat - decompresses files to stdout
       bzip2recover - recovers data from damaged bzip2 files


SYNOPSIS
       bzip2 [ -cdfkqstvzVL123456789 ] [ filenames ...    ]
       bunzip2 [ -fkvsVL ] [ filenames ...  ]
       bzcat [ -s ] [ filenames ...  ]
       bzip2recover filename


DESCRIPTION
       bzip2  compresses  files     using    the Burrows-Wheeler block sorting text
       compression algorithm, and Huffman coding.   Compression     is  generally
       considerably   better   than   that   achieved    by  more  conventional
       LZ77/LZ78-based compressors, and approaches the performance of the  PPM
       family of statistical compressors.

       The  command-line options are deliberately very similar to those of GNU
       gzip, but they are not identical.

       bzip2 expects a list of file names to accompany the command-line flags.
       Each  file is replaced by a compressed version of itself, with the name
       "original_name.bz2".  Each compressed file has  the  same  modification
       date,  permissions,  and, when possible, ownership as the corresponding
       original, so that these properties can be correctly restored at    decom-
       pression     time.    File name handling is naive in the sense that there is
       no mechanism for preserving original file  names,  permissions,    owner-
       ships  or dates in filesystems which lack these concepts, or have seri-
       ous file name length restrictions, such as MS-DOS.

       bzip2 and bunzip2 will by default not overwrite existing files.    If you
       want this to happen, specify the -f flag.

       If no file names are specified, bzip2 compresses from standard input to
       standard output.     In this case, bzip2 will decline to write  compressed
       output  to  a  terminal, as this would be entirely incomprehensible and
       therefore pointless.

       bunzip2 (or bzip2 -d) decompresses all specified     files.      Files     which
       were  not  created by bzip2 will be detected and ignored, and a warning
       issued.    bzip2 attempts to guess the filename for the decompressed file
       from that of the compressed file as follows:

          filename.bz2    becomes    filename
          filename.bz     becomes    filename
          filename.tbz2   becomes    filename.tar
          filename.tbz    becomes    filename.tar
          anyothername    becomes    anyothername.out

       If  the    file does not end in one of the recognised endings, .bz2, .bz,
       .tbz2 or .tbz, bzip2 complains that it cannot guess  the     name  of  the
       original file, and uses the original name with .out appended.

       As  with     compression, supplying no filenames causes decompression from
       standard input to standard output.

       bunzip2 will correctly decompress a file which is the concatenation  of
       two  or    more compressed files.    The result is the concatenation of the
       corresponding  uncompressed   files.    Integrity   testing   (-t)   of
       concatenated compressed files is also supported.

       You  can     also  compress     or decompress files to the standard output by
       giving the -c flag.  Multiple files may be compressed and  decompressed
       like this.  The resulting outputs are fed sequentially to stdout.  Com-
       pression of multiple files in this manner generates a stream containing
       multiple     compressed file representations.  Such a stream can be decom-
       pressed correctly only by bzip2 version 0.9.0 or later.     Earlier  ver-
       sions  of  bzip2     will  stop  after decompressing the first file in the
       stream.

       bzcat (or bzip2 -dc) decompresses all specified files to     the  standard
       output.

       bzip2  will  read  arguments  from  the environment variables BZIP2 and
       BZIP, in that order, and will process them before  any  arguments  read
       from  the  command line.     This gives a convenient way to supply default
       arguments.

       Compression is  always  performed,  even     if  the  compressed  file  is
       slightly     larger     than the original.  Files of less than about one hun-
       dred bytes tend to get larger, since the compression  mechanism    has  a
       constant     overhead  in  the region of 50 bytes.    Random data (including
       the output of most file compressors) is coded at about  8.05  bits  per
       byte, giving an expansion of around 0.5%.

       As  a  self-check  for  your protection, bzip2 uses 32-bit CRCs to make
       sure that the decompressed version of a file is identical to the origi-
       nal.   This  guards  against  corruption     of  the  compressed data, and
       against undetected  bugs     in  bzip2  (hopefully    very  unlikely).   The
       chances    of  data corruption going undetected is microscopic, about one
       chance in four billion for each file processed.    Be aware, though, that
       the check occurs upon decompression, so it can only tell you that some-
       thing is wrong.    It can’t help you recover  the    original  uncompressed
       data.   You  can     use  bzip2recover to try to recover data from damaged
       files.

       Return values: 0 for a normal exit, 1 for environmental problems     (file
       not found, invalid flags, I/O errors, &c), 2 to indicate a corrupt com-
       pressed file, 3 for an  internal     consistency  error  (eg,  bug)     which
       caused bzip2 to panic.


OPTIONS
       -c --stdout
          Compress or decompress to standard output.

       -d --decompress
          Force  decompression.   bzip2,  bunzip2 and bzcat are really the
          same program, and the decision about what     actions  to  take  is
          done  on    the  basis of which name is used.  This flag overrides
          that mechanism, and forces bzip2 to decompress.

       -z --compress
          The complement to -d:  forces  compression,  regardless  of  the
          invocation name.

       -t --test
          Check  integrity    of the specified file(s), but don’t decompress
          them.  This really performs a  trial  decompression  and    throws
          away the result.

       -f --force
          Force overwrite of output files.    Normally, bzip2 will not over-
          write existing output files.  Also forces bzip2  to  break  hard
          links to files, which it otherwise wouldn’t do.

          bzip2 normally declines to decompress files which don’t have the
          correct magic header bytes.  If forced (-f),  however,  it  will
          pass  such  files     through  unmodified.    This  is  how GNU gzip
          behaves.

       -k --keep
          Keep (don’t delete) input files during compression or decompres-
          sion.

       -s --small
          Reduce memory usage, for compression, decompression and testing.
          Files are decompressed and tested     using    a  modified  algorithm
          which  only  requires  2.5 bytes per block byte.    This means any
          file can be decompressed in 2300k of  memory,  albeit  at     about
          half the normal speed.

          During  compression, -s selects a block size of 200k, which lim-
          its memory use to around the same figure, at the expense of your
          compression  ratio.   In short, if your machine is low on memory
          (8 megabytes or less), use -s for everything.  See  MEMORY  MAN-
          AGEMENT below.

       -q --quiet
          Suppress non-essential warning messages.    Messages pertaining to
          I/O errors and other critical events will not be suppressed.

       -v --verbose
          Verbose mode -- show the compression ratio for  each  file  pro-
          cessed.    Further -v’s increase the verbosity level, spewing out
          lots of information which is primarily of interest for  diagnos-
          tic purposes.

       -L --license -V --version
          Display the software version, license terms and conditions.

       -1 (or --fast) to -9 (or --best)
          Set  the    block size to 100 k, 200 k ..  900 k when compressing.
          Has no effect when decompressing.     See MEMORY MANAGEMENT    below.
          The --fast and --best aliases are primarily for GNU gzip compat-
          ibility.    In particular, --fast  doesn’t    make  things  signifi-
          cantly faster.  And --best merely selects the default behaviour.

       --     Treats all subsequent arguments as  file    names,    even  if  they
          start  with  a dash.  This is so you can handle files with names
          beginning with a dash, for example: bzip2 -- -myfilename.

       --repetitive-fast --repetitive-best
          These flags are redundant in versions  0.9.5  and     above.      They
          provided    some  coarse control over the behaviour of the sorting
          algorithm in  earlier  versions,    which  was  sometimes  useful.
          0.9.5  and  above have an improved algorithm which renders these
          flags irrelevant.


MEMORY MANAGEMENT
       bzip2 compresses large files in blocks.    The block  size     affects  both
       the  compression     ratio    achieved,  and the amount of memory needed for
       compression and decompression.  The flags -1  through  -9  specify  the
       block  size  to    be  100,000  bytes through 900,000 bytes (the default)
       respectively.  At decompression time, the block size used for  compres-
       sion  is     read from the header of the compressed file, and bunzip2 then
       allocates itself just enough memory  to    decompress  the     file.     Since
       block  sizes  are stored in compressed files, it follows that the flags
       -1 to -9 are irrelevant to and so ignored during decompression.

       Compression and decompression requirements, in bytes, can be  estimated
       as:

          Compression:   400k + ( 8 x block size )

          Decompression: 100k + ( 4 x block size ), or
                 100k + ( 2.5 x block size )

       Larger  block sizes give rapidly diminishing marginal returns.  Most of
       the compression comes from the first two or three hundred  k  of     block
       size,  a fact worth bearing in mind when using bzip2 on small machines.
       It is also  important  to  appreciate  that  the     decompression    memory
       requirement is set at compression time by the choice of block size.

       For  files  compressed  with  the default 900k block size, bunzip2 will
       require about 3700 kbytes to decompress.     To support  decompression  of
       any  file  on a 4 megabyte machine, bunzip2 has an option to decompress
       using approximately half this amount  of     memory,  about     2300  kbytes.
       Decompression  speed is also halved, so you should use this option only
       where necessary.     The relevant flag is -s.

       In general, try and use    the  largest  block  size  memory  constraints
       allow,  since that maximises the compression achieved.  Compression and
       decompression speed are virtually unaffected by block size.

       Another significant point applies to files which fit in a single     block
       -- that means most files you’d encounter using a large block size.  The
       amount of real memory touched is proportional to the size of the     file,
       since  the  file     is  smaller than a block.  For example, compressing a
       file 20,000 bytes long with the flag -9 will cause  the    compressor  to
       allocate     around 7600k of memory, but only touch 400k + 20000 * 8 = 560
       kbytes of it.  Similarly, the decompressor will allocate 3700k but only
       touch 100k + 20000 * 4 = 180 kbytes.

       Here is a table which summarises the maximum memory usage for different
       block sizes.  Also recorded is the total compressed size for  14     files
       of the Calgary Text Compression Corpus totalling 3,141,622 bytes.  This
       column gives some feel for how  compression  varies  with  block     size.
       These  figures  tend  to understate the advantage of larger block sizes
       for larger files, since the Corpus is dominated by smaller files.

          Compress   Decompress      Decompress   Corpus
       Flag        usage      usage       -s usage    Size

        -1        1200k    500k         350k      914704
        -2        2000k    900k         600k      877703
        -3        2800k      1300k         850k      860338
        -4        3600k      1700k        1100k      846899
        -5        4400k      2100k        1350k      845160
        -6        5200k      2500k        1600k      838626
        -7        6100k      2900k        1850k      834096
        -8        6800k      3300k        2100k      828642
        -9        7600k      3700k        2350k      828642


RECOVERING DATA FROM DAMAGED FILES
       bzip2 compresses files in blocks, usually 900kbytes long.   Each     block
       is  handled  independently.   If a media or transmission error causes a
       multi-block .bz2 file to become damaged, it may be possible to  recover
       data from the undamaged blocks in the file.

       The  compressed    representation    of each block is delimited by a 48-bit
       pattern, which makes it possible to find the block boundaries with rea-
       sonable certainty.  Each block also carries its own 32-bit CRC, so dam-
       aged blocks can be distinguished from undamaged ones.

       bzip2recover is a simple program whose purpose is to search for    blocks
       in  .bz2     files,     and write each block out into its own .bz2 file.  You
       can then use bzip2 -t to test the integrity of the resulting files, and
       decompress those which are undamaged.

       bzip2recover takes a single argument, the name of the damaged file, and
       writes a number of files "rec00001file.bz2",  "rec00002file.bz2",  etc,
       containing   the      extracted   blocks.    The   output   filenames   are
       designed     so  that the use of wildcards in subsequent processing -- for
       example,     "bzip2     -dc   rec*file.bz2 > recovered_data" -- processes the
       files in the correct order.

       bzip2recover should be of most use dealing with large .bz2  files,   as
       these will contain many blocks.    It is clearly futile to use it on dam-
       aged single-block  files,  since     a damaged  block  cannot   be    recov-
       ered.   If  you    wish to minimise any potential data loss through media
       or  transmission errors, you might consider compressing with a  smaller
       block size.


PERFORMANCE NOTES
       The  sorting  phase  of compression gathers together similar strings in
       the file.  Because of this, files containing very long runs of repeated
       symbols,     like "aabaabaabaab ..."  (repeated several hundred times) may
       compress more slowly than normal.  Versions 0.9.5 and above  fare  much
       better  than  previous  versions     in  this  respect.  The ratio between
       worst-case and average-case compression time is in the region of     10:1.
       For  previous  versions,     this figure was more like 100:1.  You can use
       the -vvvv option to monitor progress in great detail, if you want.

       Decompression speed is unaffected by these phenomena.

       bzip2 usually allocates several megabytes of memory to operate in,  and
       then  charges  all over it in a fairly random fashion.  This means that
       performance, both for compressing and decompressing, is largely    deter-
       mined  by  the  speed  at  which your machine can service cache misses.
       Because of this, small changes to the code to reduce the miss rate have
       been  observed  to  give     disproportionately large performance improve-
       ments.  I imagine bzip2 will perform best on machines with  very     large
       caches.


CAVEATS
       I/O  error  messages  are not as helpful as they could be.  bzip2 tries
       hard to detect I/O errors and exit cleanly, but the details of what the
       problem is sometimes seem rather misleading.

       This  manual  page pertains to version 1.0.4 of bzip2.  Compressed data
       created by this version is entirely forwards and     backwards  compatible
       with  the  previous  public  releases,  versions     0.1pl2, 0.9.0, 0.9.5,
       1.0.0, 1.0.1, 1.0.2 and 1.0.3, but with the following exception:     0.9.0
       and  above  can    correctly  decompress multiple concatenated compressed
       files.  0.1pl2 cannot do this; it will stop  after  decompressing  just
       the first file in the stream.

       bzip2recover  versions prior to 1.0.2 used 32-bit integers to represent
       bit positions in compressed files, so they could not handle  compressed
       files  more  than  512  megabytes  long.      Versions 1.0.2 and above use
       64-bit ints on some platforms which support them     (GNU  supported  tar-
       gets, and Windows).  To establish whether or not bzip2recover was built
       with such a limitation, run it without arguments.  In any event you can
       build  yourself    an unlimited version if you can recompile it with May-
       beUInt64 set to be an unsigned 64-bit integer.




AUTHOR
       Julian Seward, jsewardbzip.org.

       http://www.bzip.org

       The ideas embodied in bzip2 are due to (at least) the following people:
       Michael    Burrows     and  David Wheeler (for the block sorting transforma-
       tion), David Wheeler (again, for the Huffman coder), Peter Fenwick (for
       the  structured    coding    model  in  the original bzip, and many refine-
       ments), and Alistair Moffat, Radford  Neal  and    Ian  Witten  (for  the
       arithmetic  coder  in the original bzip).  I am much indebted for their
       help, support and advice.  See the manual in  the  source  distribution
       for pointers to sources of documentation.  Christian von Roques encour-
       aged me to look for faster sorting algorithms, so as to speed  up  com-
       pression.  Bela Lubkin encouraged me to improve the worst-case compres-
       sion performance.  Donna Robinson XMLised the documentation.   The  bz*
       scripts    are derived from those of GNU gzip.  Many people sent patches,
       helped with portability problems, lent machines, gave advice  and  were
       generally helpful.



                                      bzip2(1)