bzip2.1 revision 78556
1.PU 2.TH bzip2 1 3.SH NAME 4bzip2, bunzip2 \- a block-sorting file compressor, v1.0 5.br 6bzcat \- decompresses files to stdout 7.br 8bzip2recover \- recovers data from damaged bzip2 files 9 10.SH SYNOPSIS 11.ll +8 12.B bzip2 13.RB [ " \-cdfkqstvzVL123456789 " ] 14[ 15.I "filenames \&..." 16] 17.ll -8 18.br 19.B bunzip2 20.RB [ " \-fkvsVL " ] 21[ 22.I "filenames \&..." 23] 24.br 25.B bzcat 26.RB [ " \-s " ] 27[ 28.I "filenames \&..." 29] 30.br 31.B bzip2recover 32.I "filename" 33 34.SH DESCRIPTION 35.I bzip2 36compresses files using the Burrows-Wheeler block sorting 37text compression algorithm, and Huffman coding. Compression is 38generally considerably better than that achieved by more conventional 39LZ77/LZ78-based compressors, and approaches the performance of the PPM 40family of statistical compressors. 41 42The command-line options are deliberately very similar to 43those of 44.I GNU gzip, 45but they are not identical. 46 47.I bzip2 48expects a list of file names to accompany the 49command-line flags. Each file is replaced by a compressed version of 50itself, with the name "original_name.bz2". 51Each compressed file 52has the same modification date, permissions, and, when possible, 53ownership as the corresponding original, so that these properties can 54be correctly restored at decompression time. File name handling is 55naive in the sense that there is no mechanism for preserving original 56file names, permissions, ownerships or dates in filesystems which lack 57these concepts, or have serious file name length restrictions, such as 58MS-DOS. 59 60.I bzip2 61and 62.I bunzip2 63will by default not overwrite existing 64files. If you want this to happen, specify the \-f flag. 65 66If no file names are specified, 67.I bzip2 68compresses from standard 69input to standard output. In this case, 70.I bzip2 71will decline to 72write compressed output to a terminal, as this would be entirely 73incomprehensible and therefore pointless. 74 75.I bunzip2 76(or 77.I bzip2 \-d) 78decompresses all 79specified files. Files which were not created by 80.I bzip2 81will be detected and ignored, and a warning issued. 82.I bzip2 83attempts to guess the filename for the decompressed file 84from that of the compressed file as follows: 85 86 filename.bz2 becomes filename 87 filename.bz becomes filename 88 filename.tbz2 becomes filename.tar 89 filename.tbz becomes filename.tar 90 anyothername becomes anyothername.out 91 92If the file does not end in one of the recognised endings, 93.I .bz2, 94.I .bz, 95.I .tbz2 96or 97.I .tbz, 98.I bzip2 99complains that it cannot 100guess the name of the original file, and uses the original name 101with 102.I .out 103appended. 104 105As with compression, supplying no 106filenames causes decompression from 107standard input to standard output. 108 109.I bunzip2 110will correctly decompress a file which is the 111concatenation of two or more compressed files. The result is the 112concatenation of the corresponding uncompressed files. Integrity 113testing (\-t) 114of concatenated 115compressed files is also supported. 116 117You can also compress or decompress files to the standard output by 118giving the \-c flag. Multiple files may be compressed and 119decompressed like this. The resulting outputs are fed sequentially to 120stdout. Compression of multiple files 121in this manner generates a stream 122containing multiple compressed file representations. Such a stream 123can be decompressed correctly only by 124.I bzip2 125version 0.9.0 or 126later. Earlier versions of 127.I bzip2 128will stop after decompressing 129the first file in the stream. 130 131.I bzcat 132(or 133.I bzip2 -dc) 134decompresses all specified files to 135the standard output. 136 137.I bzip2 138will read arguments from the environment variables 139.I BZIP2 140and 141.I BZIP, 142in that order, and will process them 143before any arguments read from the command line. This gives a 144convenient way to supply default arguments. 145 146Compression is always performed, even if the compressed 147file is slightly 148larger than the original. Files of less than about one hundred bytes 149tend to get larger, since the compression mechanism has a constant 150overhead in the region of 50 bytes. Random data (including the output 151of most file compressors) is coded at about 8.05 bits per byte, giving 152an expansion of around 0.5%. 153 154As a self-check for your protection, 155.I 156bzip2 157uses 32-bit CRCs to 158make sure that the decompressed version of a file is identical to the 159original. This guards against corruption of the compressed data, and 160against undetected bugs in 161.I bzip2 162(hopefully very unlikely). The 163chances of data corruption going undetected is microscopic, about one 164chance in four billion for each file processed. Be aware, though, that 165the check occurs upon decompression, so it can only tell you that 166something is wrong. It can't help you 167recover the original uncompressed 168data. You can use 169.I bzip2recover 170to try to recover data from 171damaged files. 172 173Return values: 0 for a normal exit, 1 for environmental problems (file 174not found, invalid flags, I/O errors, &c), 2 to indicate a corrupt 175compressed file, 3 for an internal consistency error (eg, bug) which 176caused 177.I bzip2 178to panic. 179 180.SH OPTIONS 181.TP 182.B \-c --stdout 183Compress or decompress to standard output. 184.TP 185.B \-d --decompress 186Force decompression. 187.I bzip2, 188.I bunzip2 189and 190.I bzcat 191are 192really the same program, and the decision about what actions to take is 193done on the basis of which name is used. This flag overrides that 194mechanism, and forces 195.I bzip2 196to decompress. 197.TP 198.B \-z --compress 199The complement to \-d: forces compression, regardless of the 200invokation name. 201.TP 202.B \-t --test 203Check integrity of the specified file(s), but don't decompress them. 204This really performs a trial decompression and throws away the result. 205.TP 206.B \-f --force 207Force overwrite of output files. Normally, 208.I bzip2 209will not overwrite 210existing output files. Also forces 211.I bzip2 212to break hard links 213to files, which it otherwise wouldn't do. 214.TP 215.B \-k --keep 216Keep (don't delete) input files during compression 217or decompression. 218.TP 219.B \-s --small 220Reduce memory usage, for compression, decompression and testing. Files 221are decompressed and tested using a modified algorithm which only 222requires 2.5 bytes per block byte. This means any file can be 223decompressed in 2300k of memory, albeit at about half the normal speed. 224 225During compression, \-s selects a block size of 200k, which limits 226memory use to around the same figure, at the expense of your compression 227ratio. In short, if your machine is low on memory (8 megabytes or 228less), use \-s for everything. See MEMORY MANAGEMENT below. 229.TP 230.B \-q --quiet 231Suppress non-essential warning messages. Messages pertaining to 232I/O errors and other critical events will not be suppressed. 233.TP 234.B \-v --verbose 235Verbose mode -- show the compression ratio for each file processed. 236Further \-v's increase the verbosity level, spewing out lots of 237information which is primarily of interest for diagnostic purposes. 238.TP 239.B \-L --license -V --version 240Display the software version, license terms and conditions. 241.TP 242.B \-1 to \-9 243Set the block size to 100 k, 200 k .. 900 k when compressing. Has no 244effect when decompressing. See MEMORY MANAGEMENT below. 245.TP 246.B \-- 247Treats all subsequent arguments as file names, even if they start 248with a dash. This is so you can handle files with names beginning 249with a dash, for example: bzip2 \-- \-myfilename. 250.TP 251.B \--repetitive-fast --repetitive-best 252These flags are redundant in versions 0.9.5 and above. They provided 253some coarse control over the behaviour of the sorting algorithm in 254earlier versions, which was sometimes useful. 0.9.5 and above have an 255improved algorithm which renders these flags irrelevant. 256 257.SH MEMORY MANAGEMENT 258.I bzip2 259compresses large files in blocks. The block size affects 260both the compression ratio achieved, and the amount of memory needed for 261compression and decompression. The flags \-1 through \-9 262specify the block size to be 100,000 bytes through 900,000 bytes (the 263default) respectively. At decompression time, the block size used for 264compression is read from the header of the compressed file, and 265.I bunzip2 266then allocates itself just enough memory to decompress 267the file. Since block sizes are stored in compressed files, it follows 268that the flags \-1 to \-9 are irrelevant to and so ignored 269during decompression. 270 271Compression and decompression requirements, 272in bytes, can be estimated as: 273 274 Compression: 400k + ( 8 x block size ) 275 276 Decompression: 100k + ( 4 x block size ), or 277 100k + ( 2.5 x block size ) 278 279Larger block sizes give rapidly diminishing marginal returns. Most of 280the compression comes from the first two or three hundred k of block 281size, a fact worth bearing in mind when using 282.I bzip2 283on small machines. 284It is also important to appreciate that the decompression memory 285requirement is set at compression time by the choice of block size. 286 287For files compressed with the default 900k block size, 288.I bunzip2 289will require about 3700 kbytes to decompress. To support decompression 290of any file on a 4 megabyte machine, 291.I bunzip2 292has an option to 293decompress using approximately half this amount of memory, about 2300 294kbytes. Decompression speed is also halved, so you should use this 295option only where necessary. The relevant flag is -s. 296 297In general, try and use the largest block size memory constraints allow, 298since that maximises the compression achieved. Compression and 299decompression speed are virtually unaffected by block size. 300 301Another significant point applies to files which fit in a single block 302-- that means most files you'd encounter using a large block size. The 303amount of real memory touched is proportional to the size of the file, 304since the file is smaller than a block. For example, compressing a file 30520,000 bytes long with the flag -9 will cause the compressor to 306allocate around 7600k of memory, but only touch 400k + 20000 * 8 = 560 307kbytes of it. Similarly, the decompressor will allocate 3700k but only 308touch 100k + 20000 * 4 = 180 kbytes. 309 310Here is a table which summarises the maximum memory usage for different 311block sizes. Also recorded is the total compressed size for 14 files of 312the Calgary Text Compression Corpus totalling 3,141,622 bytes. This 313column gives some feel for how compression varies with block size. 314These figures tend to understate the advantage of larger block sizes for 315larger files, since the Corpus is dominated by smaller files. 316 317 Compress Decompress Decompress Corpus 318 Flag usage usage -s usage Size 319 320 -1 1200k 500k 350k 914704 321 -2 2000k 900k 600k 877703 322 -3 2800k 1300k 850k 860338 323 -4 3600k 1700k 1100k 846899 324 -5 4400k 2100k 1350k 845160 325 -6 5200k 2500k 1600k 838626 326 -7 6100k 2900k 1850k 834096 327 -8 6800k 3300k 2100k 828642 328 -9 7600k 3700k 2350k 828642 329 330.SH RECOVERING DATA FROM DAMAGED FILES 331.I bzip2 332compresses files in blocks, usually 900kbytes long. Each 333block is handled independently. If a media or transmission error causes 334a multi-block .bz2 335file to become damaged, it may be possible to 336recover data from the undamaged blocks in the file. 337 338The compressed representation of each block is delimited by a 48-bit 339pattern, which makes it possible to find the block boundaries with 340reasonable certainty. Each block also carries its own 32-bit CRC, so 341damaged blocks can be distinguished from undamaged ones. 342 343.I bzip2recover 344is a simple program whose purpose is to search for 345blocks in .bz2 files, and write each block out into its own .bz2 346file. You can then use 347.I bzip2 348\-t 349to test the 350integrity of the resulting files, and decompress those which are 351undamaged. 352 353.I bzip2recover 354takes a single argument, the name of the damaged file, 355and writes a number of files "rec0001file.bz2", 356"rec0002file.bz2", etc, containing the extracted blocks. 357The output filenames are designed so that the use of 358wildcards in subsequent processing -- for example, 359"bzip2 -dc rec*file.bz2 > recovered_data" -- lists the files in 360the correct order. 361 362.I bzip2recover 363should be of most use dealing with large .bz2 364files, as these will contain many blocks. It is clearly 365futile to use it on damaged single-block files, since a 366damaged block cannot be recovered. If you wish to minimise 367any potential data loss through media or transmission errors, 368you might consider compressing with a smaller 369block size. 370 371.SH PERFORMANCE NOTES 372The sorting phase of compression gathers together similar strings in the 373file. Because of this, files containing very long runs of repeated 374symbols, like "aabaabaabaab ..." (repeated several hundred times) may 375compress more slowly than normal. Versions 0.9.5 and above fare much 376better than previous versions in this respect. The ratio between 377worst-case and average-case compression time is in the region of 10:1. 378For previous versions, this figure was more like 100:1. You can use the 379\-vvvv option to monitor progress in great detail, if you want. 380 381Decompression speed is unaffected by these phenomena. 382 383.I bzip2 384usually allocates several megabytes of memory to operate 385in, and then charges all over it in a fairly random fashion. This means 386that performance, both for compressing and decompressing, is largely 387determined by the speed at which your machine can service cache misses. 388Because of this, small changes to the code to reduce the miss rate have 389been observed to give disproportionately large performance improvements. 390I imagine 391.I bzip2 392will perform best on machines with very large caches. 393 394.SH CAVEATS 395I/O error messages are not as helpful as they could be. 396.I bzip2 397tries hard to detect I/O errors and exit cleanly, but the details of 398what the problem is sometimes seem rather misleading. 399 400This manual page pertains to version 1.0 of 401.I bzip2. 402Compressed 403data created by this version is entirely forwards and backwards 404compatible with the previous public releases, versions 0.1pl2, 0.9.0 405and 0.9.5, 406but with the following exception: 0.9.0 and above can correctly 407decompress multiple concatenated compressed files. 0.1pl2 cannot do 408this; it will stop after decompressing just the first file in the 409stream. 410 411.I bzip2recover 412uses 32-bit integers to represent bit positions in 413compressed files, so it cannot handle compressed files more than 512 414megabytes long. This could easily be fixed. 415 416.SH AUTHOR 417Julian Seward, jseward@acm.org. 418 419http://sourceware.cygnus.com/bzip2 420http://www.muraroa.demon.co.uk 421 422The ideas embodied in 423.I bzip2 424are due to (at least) the following 425people: Michael Burrows and David Wheeler (for the block sorting 426transformation), David Wheeler (again, for the Huffman coder), Peter 427Fenwick (for the structured coding model in the original 428.I bzip, 429and many refinements), and Alistair Moffat, Radford Neal and Ian Witten 430(for the arithmetic coder in the original 431.I bzip). 432I am much 433indebted for their help, support and advice. See the manual in the 434source distribution for pointers to sources of documentation. Christian 435von Roques encouraged me to look for faster sorting algorithms, so as to 436speed up compression. Bela Lubkin encouraged me to improve the 437worst-case compression performance. Many people sent patches, helped 438with portability problems, lent machines, gave advice and were generally 439helpful. 440