zpaq - PAQ open standard maximum compressor
create : zpaq [opnsitqv]c<config>[,N...] archive.zpaq file [file ...]
append : zpaq [opnsitqv]a<config>[,N...] archive.zpaq file [file ...]
list : zpaq l archive.zpaq
extract: zpaq [opntq]x[N] archive.zpaq
debug : zpaq [pthv]rF[,N...] [args...]
PAQ is a series of open source data compression archivers that have
evolved through collaborative development to top rankings on several
benchmarks measuring compression ratio although at the expense of speed
and memory usage.
Zpaq ia a proposed standard format for highly compressed data that
allows new compression algorithms to be developed without breaking
compatibility with older programs. Zpaq is based on PAQ-like context
mixing algorithms which are top ranked on many benchmarks. The format
supports archivers, single file compressors, and memory to memory
ZPAQ is a configurable file compressor and archiver. Its goal is a high
compression ratio in an open format without loss of compatibility
between versions as advanced compression techniques are discovered.
Compression uses by default built-in configuration files. Three
examples are supplied:
min.cfg - Fast, minimal compression (LZP + order 3). Requires 4 MB memory.
mid.cfg - Average compression and speed. Requires 111 MB.
max.cfg - Slow but good compression. Requires 278 MB.
The config file is not needed to extract.
NOTE: in extract mode, if the FILES are listed the files are renamed
during written out.
a Append to archive.
c Create archive.
i Don't store file sizes as comments (saves a few bytes). Normally
the input file size is stored as a decimal string, taking a few
bytes. The comment field has no effect on the program except that
it is displayed by the l and x commands.
l List contents of archive.
n In create mode: Don't store filenames (names will be needed to
decompress). In extract mode: decompress all to one file. The
effect is to require that filenames be given during decompression.
During extract ignore all stored filenames and append all output to
one file, the first file in [files...].
o Optimize (run faster). You need a C++ compiler installed to use
this option. If not, drop the "o". You can still use zpaq but it
will take about twice as long to run.
If successful, compression is typically 50% to 100% faster. Zpaq
will look for a program named "zpaq_X" in the temporary directory,
where X is derived from the SHA1 checksum of the block header
produced by config file CONFIG with arguments N. If the program
exists, then Zpaq will call it with the same arguments to perform
the compression. If it does not exist then Zpaq will create a
source code file "zpaq_X.cpp" in the temporary directory, compile
it, and link it to "zpaq.cpp" or "zpaq.o" depending on the
The temporary directory is specified by the environment variable
TEMP if it exists, or else the current directory.
The program "zpaq_X" will compress its input in the same format as
described by CONFIG, but faster. If CONFIG specifies a
preprocessor, then "zpaq_X" will expect to find it too. It will
also decompress archive blocks in the same configuration but fail
if it attempts to decompress blocks in any other configuration.
Program "zpaq_X" will accept the c, a and x commands with all of
the same modifiers, but will ignore the v and o modifiers and
ignore any CONFIG file and arguments passed to it. It will not
accept the l or r commands. Extraction requires a block number
("x1", "x2", etc). A different optimized program is used to extract
Zpaq will call the external program "zpaqmake" to compile
"zpaq_X.cpp", passing it "zpaq_X" as an argument. Normally this
will be a script that calls a C++ compiler to produce "zpaq_X.o",
links to "zpaq.o" and outputs "zpaq_X". The script could link to
"zpaq.cpp" instead of "zpaq.o".
p In create mode: Store filename paths in archive. The default is to
store the name without the path. For example:
zpaq pc<CONFIG> books.zpaq dir/file
will store the name as "dir/file". If the p option is also given
during extraction, then ZPAQ will attempt to extract "file" to the
subdirectory instead of the current directory. This will fail if
directory does not exist. ZPAQ does not create directories as
In extract mode: extract to stored paths instead of current
The default is to extract to the current directory regardless of
how the file names are stored. Stored paths must be relative to the
current directory, not start with a "/", "\", a drive letter like
"C:" or contain "../" or "..\". If extracting to a subdirectory, it
must already exist. It will not be created.
[files...] overrides and has no restrictions on file names. Each
segment extracts to a different file. If any segments do not have a
stored filename then they can only be extracted using the p or n
q Quiet mode. Don't display compression progress on the screen.
s Don't store SHA1 checksums (saves 20 bytes).The decompressor will
not check that the output is identical to the original input.
t In create mode: Append locator tag to non-ZPAQ data.
Append a locator tag to non-ZPAQ data. The tag is a string of 13
bytes that allows ZPAQ and UNZPAQ to find the start of a sequence
of ZPAQ blocks embedded in other data. Program "zpaqsfx" already
has this tag at the end. However, if a new stub is compiled from
the source then the t command should be used when appending the
In extract mode: don't post-process (for debugging). Expect
v Verbose mode. Show CONFIG file as it compiles. This is useful for
x Extract. Use "ox" to extract fast. You can extract more slowly with
plain "x" if you don't have C++ installed. Output files are renamed
in the same order they are stored and listed. If you don't rename
the output files, then the files will be extracted to the current
directory with the same names they had when stored.
,N Usd in create mode. Pass numeric arguments to CONFIG file. Appended
suffix like ",2" means use 4 times more memory. Each increment
doubles usage. You need the same memory to decompress.
N Used in extract mode. Extract only block N (1, 2, 3...), where 1 is
the first block. Otherwise all blocks are extracted. The l command
shows which files are in each block.
Debug and Development Options
To debug CONFIG file, use:
zpaq [pthv]r<CONFIG>[,N...] [args...]
the r runn the ZPAQL program in HCOMP section of configuration file F.
The program is run once for each byte of input from the file named in
the first argument and once at EOF with the input byte (or -1) in the A
register. Output is to the file named in the second argument. If run
with no arguments then take input from stdin and output to stdout.
Modifiers are listed below.
h When tracing, display register and memory contents in hexadecimal
instead of decimal.
p Run PCOMP (default is to run HCOMP).
t Trace (single step), args are numeric inputs otherwise args are
input, output (default stdin, stdout). The arguments should be
numbers rather than file names. The program is run once for each
argument with the value in the A register. As each instruction is
executed the register contents are shown. At HALT, memory contents
v Verbose compile. Display the CONFIG file as it is being compiled.
If an error occurs, it will be easier to locate. Modifier v is also
useful for displaying jump targets.
,N Pass numeric arguments to CONFIG file. Pass up to 9 numeric
arguments to CONFIG file (like the c and a commands).
-h Display short help.
To create an archive:
zpaq c<CONFIG> archive.zpaq files ...
If the archive exists then it is overwritten. File names are stored
without a path.
To (a)ppend to an existing archive. If the archive does not exist then
it is created as with the c command:
zpaq a<CONFIG> archive.zpaq files ...
To list the contents of an archive. Files are listed in the same order
they were added:
zpaq l archive.zpaq
To extract the contents of the archive. New files are created and named
according to the stored filenames. Does not clobber existing files.
Extracts to current directory:
zpaq x archive.zpaq
If the files to be extracted already exist, then zpaq will refuse to
clobber them and skip to the next file. If the files are compressed
with a path (folder or directory), then that directory must exist when
the file is extracted. zpaq will not create directories.
To extract files and renames in the order they were added to the
archive. Clobbers any already existing output files. The number of
files extracted is the smaller of the number of filenames on the
command line or the number of files in the archive.
zpaq x archive.zpaq file ...
To extract and rename:
zpaq x archive.zpaq files ...
unzpaq x archive.zpaq files ...
Files are extracted in the same order they are saved and renamed.
Unlike using stored names, if the file exists, then it is overwritten
(clobbered). Only files named on the command line are extracted. Any
additional files in the archive are ignored. For example:
zpaq x archive.zpaq foo bar
To extracts files like x, but without post-processing. This may be
useful for debugging or developing config files:
zpaq t archive.zpaq [files ...]
The distribution contain several default CONFIG files:
min.cfg - for fast but poor compression.
max.cfg - for slow but good compression.
mid.cfg - for moderate speed and compression (default).
Other config files are available as add-on options or you can write
them as explained later.
A numeric argument may be appended to CONGIF to increase memory usage
for better compression. Each increment doubles usage. There should be
no space before or after the comma. For example:
zpaq cmax.cfg archive files... = 246 MB
zpaq cmax.cfg,1 archive files... = 476 MB
zpaq cmax.cfg,2 archive files... = 938 MB
zpaq cmax.cfg,3 archive files... = 1861 MB
zpaq cmax.cfg,-1 archive files... = 130 MB (negative values allowed)
Modifiers may be in any order before the "c" or "a" command. The
modifiers, command, and configuration file must be written together
without any spaces. An example: to create archive with options i, p, s
and configuration file "max.cfg". Modifiers have the following meaning:
zpaq ipsc<CONFIG> archive.zpaq file1 file22
Temporary directory TEMPDIR is use during optimize command o.
Compression commands c and a need a configuration file. See examples in
See zpaq*.pdf (ZPAQ Level 1 and later) in section AVAILABILITY . It is
anticipated that future levels (ZPAQ-2, ZPAQ-3, etc.) will be backward
compatible, such that newer levels can read archives produced by older
bzip2(1) gzip(1) lzop(1) lzma(1) p7zip(1) rzip(1) unace(1) unrar(1)
Program was written by Matt Mahoney <firstname.lastname@example.org>
This manual page was put together by Jari Aalto <email@example.com>.
under license GNU GPL version 2 or (at your option) any later version.
For more information about license, visit