6 bzip2, bunzip2 − a block‐sorting file compressor, v1.0.6
7 bzcat − decompresses files to stdout
8 bzip2recover − recovers data from damaged bzip2 files
11 S
\bSY
\bYN
\bNO
\bOP
\bPS
\bSI
\bIS
\bS
12 b
\bbz
\bzi
\bip
\bp2
\b2 [ −
\b−c
\bcd
\bdf
\bfk
\bkq
\bqs
\bst
\btv
\bvz
\bzV
\bVL
\bL1
\b12
\b23
\b34
\b45
\b56
\b67
\b78
\b89
\b9 ] [ _
\bf_
\bi_
\bl_
\be_
\bn_
\ba_
\bm_
\be_
\bs _
\b._
\b._
\b. ]
13 b
\bbu
\bun
\bnz
\bzi
\bip
\bp2
\b2 [ −
\b−f
\bfk
\bkv
\bvs
\bsV
\bVL
\bL ] [ _
\bf_
\bi_
\bl_
\be_
\bn_
\ba_
\bm_
\be_
\bs _
\b._
\b._
\b. ]
14 b
\bbz
\bzc
\bca
\bat
\bt [ −
\b−s
\bs ] [ _
\bf_
\bi_
\bl_
\be_
\bn_
\ba_
\bm_
\be_
\bs _
\b._
\b._
\b. ]
15 b
\bbz
\bzi
\bip
\bp2
\b2r
\bre
\bec
\bco
\bov
\bve
\ber
\br _
\bf_
\bi_
\bl_
\be_
\bn_
\ba_
\bm_
\be
18 D
\bDE
\bES
\bSC
\bCR
\bRI
\bIP
\bPT
\bTI
\bIO
\bON
\bN
19 _
\bb_
\bz_
\bi_
\bp_
\b2 compresses files using the Burrows‐Wheeler block
20 sorting text compression algorithm, and Huffman coding.
21 Compression is generally considerably better than that
22 achieved by more conventional LZ77/LZ78‐based compressors,
23 and approaches the performance of the PPM family of sta
26 The command‐line options are deliberately very similar to
27 those of _
\bG_
\bN_
\bU _
\bg_
\bz_
\bi_
\bp_
\b, but they are not identical.
29 _
\bb_
\bz_
\bi_
\bp_
\b2 expects a list of file names to accompany the com
30 mand‐line flags. Each file is replaced by a compressed
31 version of itself, with the name "original_name.bz2".
32 Each compressed file has the same modification date, per
33 missions, and, when possible, ownership as the correspond
34 ing original, so that these properties can be correctly
35 restored at decompression time. File name handling is
36 naive in the sense that there is no mechanism for preserv
37 ing original file names, permissions, ownerships or dates
38 in filesystems which lack these concepts, or have serious
39 file name length restrictions, such as MS‐DOS.
41 _
\bb_
\bz_
\bi_
\bp_
\b2 and _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 will by default not overwrite existing
42 files. If you want this to happen, specify the −f flag.
44 If no file names are specified, _
\bb_
\bz_
\bi_
\bp_
\b2 compresses from
45 standard input to standard output. In this case, _
\bb_
\bz_
\bi_
\bp_
\b2
46 will decline to write compressed output to a terminal, as
47 this would be entirely incomprehensible and therefore
50 _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 (or _
\bb_
\bz_
\bi_
\bp_
\b2 _
\b−_
\bd_
\b) decompresses all specified files.
51 Files which were not created by _
\bb_
\bz_
\bi_
\bp_
\b2 will be detected and
52 ignored, and a warning issued. _
\bb_
\bz_
\bi_
\bp_
\b2 attempts to guess
53 the filename for the decompressed file from that of the
54 compressed file as follows:
56 filename.bz2 becomes filename
57 filename.bz becomes filename
58 filename.tbz2 becomes filename.tar
59 filename.tbz becomes filename.tar
60 anyothername becomes anyothername.out
62 If the file does not end in one of the recognised endings,
63 _
\b._
\bb_
\bz_
\b2_
\b, _
\b._
\bb_
\bz_
\b, _
\b._
\bt_
\bb_
\bz_
\b2 or _
\b._
\bt_
\bb_
\bz_
\b, _
\bb_
\bz_
\bi_
\bp_
\b2 complains that it cannot
64 guess the name of the original file, and uses the original
65 name with _
\b._
\bo_
\bu_
\bt appended.
67 As with compression, supplying no filenames causes decom
68 pression from standard input to standard output.
70 _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 will correctly decompress a file which is the con
71 catenation of two or more compressed files. The result is
72 the concatenation of the corresponding uncompressed files.
73 Integrity testing (−t) of concatenated compressed files is
76 You can also compress or decompress files to the standard
77 output by giving the −c flag. Multiple files may be com
78 pressed and decompressed like this. The resulting outputs
79 are fed sequentially to stdout. Compression of multiple
80 files in this manner generates a stream containing multi
81 ple compressed file representations. Such a stream can be
82 decompressed correctly only by _
\bb_
\bz_
\bi_
\bp_
\b2 version 0.9.0 or
83 later. Earlier versions of _
\bb_
\bz_
\bi_
\bp_
\b2 will stop after decom
84 pressing the first file in the stream.
86 _
\bb_
\bz_
\bc_
\ba_
\bt (or _
\bb_
\bz_
\bi_
\bp_
\b2 _
\b‐_
\bd_
\bc_
\b) decompresses all specified files to
89 _
\bb_
\bz_
\bi_
\bp_
\b2 will read arguments from the environment variables
90 _
\bB_
\bZ_
\bI_
\bP_
\b2 and _
\bB_
\bZ_
\bI_
\bP_
\b, in that order, and will process them
91 before any arguments read from the command line. This
92 gives a convenient way to supply default arguments.
94 Compression is always performed, even if the compressed
95 file is slightly larger than the original. Files of less
96 than about one hundred bytes tend to get larger, since the
97 compression mechanism has a constant overhead in the
98 region of 50 bytes. Random data (including the output of
99 most file compressors) is coded at about 8.05 bits per
100 byte, giving an expansion of around 0.5%.
102 As a self‐check for your protection, _
\bb_
\bz_
\bi_
\bp_
\b2 uses 32‐bit
103 CRCs to make sure that the decompressed version of a file
104 is identical to the original. This guards against corrup
105 tion of the compressed data, and against undetected bugs
106 in _
\bb_
\bz_
\bi_
\bp_
\b2 (hopefully very unlikely). The chances of data
107 corruption going undetected is microscopic, about one
108 chance in four billion for each file processed. Be aware,
109 though, that the check occurs upon decompression, so it
110 can only tell you that something is wrong. It can’t help
111 you recover the original uncompressed data. You can use
112 _
\bb_
\bz_
\bi_
\bp_
\b2_
\br_
\be_
\bc_
\bo_
\bv_
\be_
\br to try to recover data from damaged files.
114 Return values: 0 for a normal exit, 1 for environmental
115 problems (file not found, invalid flags, I/O errors, &c),
116 2 to indicate a corrupt compressed file, 3 for an internal
117 consistency error (eg, bug) which caused _
\bb_
\bz_
\bi_
\bp_
\b2 to panic.
120 O
\bOP
\bPT
\bTI
\bIO
\bON
\bNS
\bS
121 −
\b−c
\bc ‐
\b‐‐
\b‐s
\bst
\btd
\bdo
\bou
\but
\bt
122 Compress or decompress to standard output.
124 −
\b−d
\bd ‐
\b‐‐
\b‐d
\bde
\bec
\bco
\bom
\bmp
\bpr
\bre
\bes
\bss
\bs
125 Force decompression. _
\bb_
\bz_
\bi_
\bp_
\b2_
\b, _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 and _
\bb_
\bz_
\bc_
\ba_
\bt are
126 really the same program, and the decision about
127 what actions to take is done on the basis of which
128 name is used. This flag overrides that mechanism,
129 and forces _
\bb_
\bz_
\bi_
\bp_
\b2 to decompress.
131 −
\b−z
\bz ‐
\b‐‐
\b‐c
\bco
\bom
\bmp
\bpr
\bre
\bes
\bss
\bs
132 The complement to −d: forces compression,
133 regardless of the invocation name.
135 −
\b−t
\bt ‐
\b‐‐
\b‐t
\bte
\bes
\bst
\bt
136 Check integrity of the specified file(s), but don’t
137 decompress them. This really performs a trial
138 decompression and throws away the result.
140 −
\b−f
\bf ‐
\b‐‐
\b‐f
\bfo
\bor
\brc
\bce
\be
141 Force overwrite of output files. Normally, _
\bb_
\bz_
\bi_
\bp_
\b2
142 will not overwrite existing output files. Also
143 forces _
\bb_
\bz_
\bi_
\bp_
\b2 to break hard links to files, which it
144 otherwise wouldn’t do.
146 bzip2 normally declines to decompress files which
147 don’t have the correct magic header bytes. If
148 forced (‐f), however, it will pass such files
149 through unmodified. This is how GNU gzip behaves.
151 −
\b−k
\bk ‐
\b‐‐
\b‐k
\bke
\bee
\bep
\bp
152 Keep (don’t delete) input files during compression
155 −
\b−s
\bs ‐
\b‐‐
\b‐s
\bsm
\bma
\bal
\bll
\bl
156 Reduce memory usage, for compression, decompression
157 and testing. Files are decompressed and tested
158 using a modified algorithm which only requires 2.5
159 bytes per block byte. This means any file can be
160 decompressed in 2300k of memory, albeit at about
161 half the normal speed.
163 During compression, −s selects a block size of
164 200k, which limits memory use to around the same
165 figure, at the expense of your compression ratio.
166 In short, if your machine is low on memory (8
167 megabytes or less), use −s for everything. See
168 MEMORY MANAGEMENT below.
170 −
\b−q
\bq ‐
\b‐‐
\b‐q
\bqu
\bui
\bie
\bet
\bt
171 Suppress non‐essential warning messages. Messages
172 pertaining to I/O errors and other critical events
173 will not be suppressed.
175 −
\b−v
\bv ‐
\b‐‐
\b‐v
\bve
\ber
\brb
\bbo
\bos
\bse
\be
176 Verbose mode ‐‐ show the compression ratio for each
177 file processed. Further −v’s increase the ver
178 bosity level, spewing out lots of information which
179 is primarily of interest for diagnostic purposes.
181 −
\b−L
\bL ‐
\b‐‐
\b‐l
\bli
\bic
\bce
\ben
\bns
\bse
\be ‐
\b‐V
\bV ‐
\b‐‐
\b‐v
\bve
\ber
\brs
\bsi
\bio
\bon
\bn
182 Display the software version, license terms and
185 −
\b−1
\b1 (
\b(o
\bor
\br −
\b−−
\b−f
\bfa
\bas
\bst
\bt)
\b) t
\bto
\bo −
\b−9
\b9 (
\b(o
\bor
\br −
\b−−
\b−b
\bbe
\bes
\bst
\bt)
\b)
186 Set the block size to 100 k, 200 k .. 900 k when
187 compressing. Has no effect when decompressing.
188 See MEMORY MANAGEMENT below. The −−fast and −−best
189 aliases are primarily for GNU gzip compatibility.
190 In particular, −−fast doesn’t make things signifi
191 cantly faster. And −−best merely selects the
194 −
\b−‐
\b‐ Treats all subsequent arguments as file names, even
195 if they start with a dash. This is so you can han
196 dle files with names beginning with a dash, for
197 example: bzip2 −‐ −myfilename.
199 −
\b−‐
\b‐r
\bre
\bep
\bpe
\bet
\bti
\bit
\bti
\biv
\bve
\be‐
\b‐f
\bfa
\bas
\bst
\bt ‐
\b‐‐
\b‐r
\bre
\bep
\bpe
\bet
\bti
\bit
\bti
\biv
\bve
\be‐
\b‐b
\bbe
\bes
\bst
\bt
200 These flags are redundant in versions 0.9.5 and
201 above. They provided some coarse control over the
202 behaviour of the sorting algorithm in earlier ver
203 sions, which was sometimes useful. 0.9.5 and above
204 have an improved algorithm which renders these
208 M
\bME
\bEM
\bMO
\bOR
\bRY
\bY M
\bMA
\bAN
\bNA
\bAG
\bGE
\bEM
\bME
\bEN
\bNT
\bT
209 _
\bb_
\bz_
\bi_
\bp_
\b2 compresses large files in blocks. The block size
210 affects both the compression ratio achieved, and the
211 amount of memory needed for compression and decompression.
212 The flags −1 through −9 specify the block size to be
213 100,000 bytes through 900,000 bytes (the default) respec
214 tively. At decompression time, the block size used for
215 compression is read from the header of the compressed
216 file, and _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 then allocates itself just enough memory
217 to decompress the file. Since block sizes are stored in
218 compressed files, it follows that the flags −1 to −9 are
219 irrelevant to and so ignored during decompression.
221 Compression and decompression requirements, in bytes, can
224 Compression: 400k + ( 8 x block size )
226 Decompression: 100k + ( 4 x block size ), or
227 100k + ( 2.5 x block size )
229 Larger block sizes give rapidly diminishing marginal
230 returns. Most of the compression comes from the first two
231 or three hundred k of block size, a fact worth bearing in
232 mind when using _
\bb_
\bz_
\bi_
\bp_
\b2 on small machines. It is also
233 important to appreciate that the decompression memory
234 requirement is set at compression time by the choice of
237 For files compressed with the default 900k block size,
238 _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 will require about 3700 kbytes to decompress. To
239 support decompression of any file on a 4 megabyte machine,
240 _
\bb_
\bu_
\bn_
\bz_
\bi_
\bp_
\b2 has an option to decompress using approximately
241 half this amount of memory, about 2300 kbytes. Decompres
242 sion speed is also halved, so you should use this option
243 only where necessary. The relevant flag is ‐s.
245 In general, try and use the largest block size memory con
246 straints allow, since that maximises the compression
247 achieved. Compression and decompression speed are virtu
248 ally unaffected by block size.
250 Another significant point applies to files which fit in a
251 single block ‐‐ that means most files you’d encounter
252 using a large block size. The amount of real memory
253 touched is proportional to the size of the file, since the
254 file is smaller than a block. For example, compressing a
255 file 20,000 bytes long with the flag ‐9 will cause the
256 compressor to allocate around 7600k of memory, but only
257 touch 400k + 20000 * 8 = 560 kbytes of it. Similarly, the
258 decompressor will allocate 3700k but only touch 100k +
259 20000 * 4 = 180 kbytes.
261 Here is a table which summarises the maximum memory usage
262 for different block sizes. Also recorded is the total
263 compressed size for 14 files of the Calgary Text Compres
264 sion Corpus totalling 3,141,622 bytes. This column gives
265 some feel for how compression varies with block size.
266 These figures tend to understate the advantage of larger
267 block sizes for larger files, since the Corpus is domi
268 nated by smaller files.
270 Compress Decompress Decompress Corpus
271 Flag usage usage ‐s usage Size
273 ‐1 1200k 500k 350k 914704
274 ‐2 2000k 900k 600k 877703
275 ‐3 2800k 1300k 850k 860338
276 ‐4 3600k 1700k 1100k 846899
277 ‐5 4400k 2100k 1350k 845160
278 ‐6 5200k 2500k 1600k 838626
279 ‐7 6100k 2900k 1850k 834096
280 ‐8 6800k 3300k 2100k 828642
281 ‐9 7600k 3700k 2350k 828642
284 R
\bRE
\bEC
\bCO
\bOV
\bVE
\bER
\bRI
\bIN
\bNG
\bG D
\bDA
\bAT
\bTA
\bA F
\bFR
\bRO
\bOM
\bM D
\bDA
\bAM
\bMA
\bAG
\bGE
\bED
\bD F
\bFI
\bIL
\bLE
\bES
\bS
285 _
\bb_
\bz_
\bi_
\bp_
\b2 compresses files in blocks, usually 900kbytes long.
286 Each block is handled independently. If a media or trans
287 mission error causes a multi‐block .bz2 file to become
288 damaged, it may be possible to recover data from the
289 undamaged blocks in the file.
291 The compressed representation of each block is delimited
292 by a 48‐bit pattern, which makes it possible to find the
293 block boundaries with reasonable certainty. Each block
294 also carries its own 32‐bit CRC, so damaged blocks can be
295 distinguished from undamaged ones.
297 _
\bb_
\bz_
\bi_
\bp_
\b2_
\br_
\be_
\bc_
\bo_
\bv_
\be_
\br is a simple program whose purpose is to
298 search for blocks in .bz2 files, and write each block out
299 into its own .bz2 file. You can then use _
\bb_
\bz_
\bi_
\bp_
\b2 −t to test
300 the integrity of the resulting files, and decompress those
303 _
\bb_
\bz_
\bi_
\bp_
\b2_
\br_
\be_
\bc_
\bo_
\bv_
\be_
\br takes a single argument, the name of the dam
304 aged file, and writes a number of files
305 "rec00001file.bz2", "rec00002file.bz2", etc, containing
306 the extracted blocks. The output filenames are
307 designed so that the use of wildcards in subsequent pro
308 cessing ‐‐ for example, "bzip2 ‐dc rec*file.bz2 > recov
309 ered_data" ‐‐ processes the files in the correct order.
311 _
\bb_
\bz_
\bi_
\bp_
\b2_
\br_
\be_
\bc_
\bo_
\bv_
\be_
\br should be of most use dealing with large .bz2
312 files, as these will contain many blocks. It is clearly
313 futile to use it on damaged single‐block files, since a
314 damaged block cannot be recovered. If you wish to min
315 imise any potential data loss through media or transmis
316 sion errors, you might consider compressing with a smaller
320 P
\bPE
\bER
\bRF
\bFO
\bOR
\bRM
\bMA
\bAN
\bNC
\bCE
\bE N
\bNO
\bOT
\bTE
\bES
\bS
321 The sorting phase of compression gathers together similar
322 strings in the file. Because of this, files containing
323 very long runs of repeated symbols, like "aabaabaabaab
324 ..." (repeated several hundred times) may compress more
325 slowly than normal. Versions 0.9.5 and above fare much
326 better than previous versions in this respect. The ratio
327 between worst‐case and average‐case compression time is in
328 the region of 10:1. For previous versions, this figure
329 was more like 100:1. You can use the −vvvv option to mon
330 itor progress in great detail, if you want.
332 Decompression speed is unaffected by these phenomena.
334 _
\bb_
\bz_
\bi_
\bp_
\b2 usually allocates several megabytes of memory to
335 operate in, and then charges all over it in a fairly ran
336 dom fashion. This means that performance, both for com
337 pressing and decompressing, is largely determined by the
338 speed at which your machine can service cache misses.
339 Because of this, small changes to the code to reduce the
340 miss rate have been observed to give disproportionately
341 large performance improvements. I imagine _
\bb_
\bz_
\bi_
\bp_
\b2 will per
342 form best on machines with very large caches.
345 C
\bCA
\bAV
\bVE
\bEA
\bAT
\bTS
\bS
346 I/O error messages are not as helpful as they could be.
347 _
\bb_
\bz_
\bi_
\bp_
\b2 tries hard to detect I/O errors and exit cleanly,
348 but the details of what the problem is sometimes seem
351 This manual page pertains to version 1.0.6 of _
\bb_
\bz_
\bi_
\bp_
\b2_
\b. Com
352 pressed data created by this version is entirely forwards
353 and backwards compatible with the previous public
354 releases, versions 0.1pl2, 0.9.0, 0.9.5, 1.0.0, 1.0.1,
355 1.0.2 and above, but with the following exception: 0.9.0
356 and above can correctly decompress multiple concatenated
357 compressed files. 0.1pl2 cannot do this; it will stop
358 after decompressing just the first file in the stream.
360 _
\bb_
\bz_
\bi_
\bp_
\b2_
\br_
\be_
\bc_
\bo_
\bv_
\be_
\br versions prior to 1.0.2 used 32‐bit integers
361 to represent bit positions in compressed files, so they
362 could not handle compressed files more than 512 megabytes
363 long. Versions 1.0.2 and above use 64‐bit ints on some
364 platforms which support them (GNU supported targets, and
365 Windows). To establish whether or not bzip2recover was
366 built with such a limitation, run it without arguments.
367 In any event you can build yourself an unlimited version
368 if you can recompile it with MaybeUInt64 set to be an
369 unsigned 64‐bit integer.
374 A
\bAU
\bUT
\bTH
\bHO
\bOR
\bR
375 Julian Seward, jsewardbzip.org.
379 The ideas embodied in _
\bb_
\bz_
\bi_
\bp_
\b2 are due to (at least) the fol
380 lowing people: Michael Burrows and David Wheeler (for the
381 block sorting transformation), David Wheeler (again, for
382 the Huffman coder), Peter Fenwick (for the structured cod
383 ing model in the original _
\bb_
\bz_
\bi_
\bp_
\b, and many refinements), and
384 Alistair Moffat, Radford Neal and Ian Witten (for the
385 arithmetic coder in the original _
\bb_
\bz_
\bi_
\bp_
\b)_
\b. I am much
386 indebted for their help, support and advice. See the man
387 ual in the source distribution for pointers to sources of
388 documentation. Christian von Roques encouraged me to look
389 for faster sorting algorithms, so as to speed up compres
390 sion. Bela Lubkin encouraged me to improve the worst‐case
391 compression performance. Donna Robinson XMLised the docu
392 mentation. The bz* scripts are derived from those of GNU
393 gzip. Many people sent patches, helped with portability
394 problems, lent machines, gave advice and were generally