When used as a filter (no file name), the commands cannot actually compress or decompress the original file and remove it because there is no file name. Filter commands simply compress or decompress the data in the input stream; the file is not changed.
Some helpful z-commands have been created to directly access compressed files and save typing gunzip in a pipe all the time:
$ gunzip <bad.txt.gz | less # hard way to paginate contents
$ zless bad.txt.gz # easy way
$ gunzip <bad.txt.gz | fgrep '.cn' # hard way to fgrep contents
$ zfgrep '.cn' bad.txt.gz # easy way
Since all of the z-commands are filters (they are small shell scripts), none of the z-commands affect the given file. The file is not decompressed and then removed. Only the file contents are decompressed and sent to standard output.
The commands bzip2 and bunzip2 are similar to gzip and gunzip but they use a different, often better, compression algorithm. The default file extension is .bz2 instead of .gz:
As with gzip, the original file is removed after being compressed, unless the command is used as a filter (without a file name). The modify time of the original file is preserved.
If you give bunzip2 a file name that does not end in .bz2, it decompresses the file into the same file name with .out appended:
$ bzip2 </etc/passwd >foo
$ file foo
foo: bzip2 compressed data, block size = 900k
$ bunzip2 foo
bunzip2: Can't guess original name for foo -- using foo.out
$ ls -l /etc/passwd foo.out
-rw-r--r-- 1 root root 168835 Mar 6 16:13 /etc/passwd
-rw-rw-r-- 1 idallen idallen 168835 Mar 8 04:05 foo.out
Some helpful bz-commands have been created to directly access compressed files and save typing bunzip2 in a pipe all the time: bzcat bzdiff bzfgrep bzgrep bzless:
$ bunzip2 <bad.txt.bz2 | less # hard way to paginate contents
$ bzless bad.txt.bz2 # easy way
These helpers have similar names and work the same way as the gzip helper z-commands. See the man pages for the other helpers.
Long before software package managers such as YUM, RPM, and APT, there were tar archives. Originally written as a magnetic Tape ARchiver, the command is common to every Unix/Linux system. A tar archive file is the Unix version of a zip file. It is one file that contains many other files inside it. You can download and extract a tar format archive file on most any Unix/Linux system back to 1969.
A tar archive, also called a “tarball”, is a single file that contains multiple uncompressed files and directories. Unix/Linux software source is often distributed as a “tarball”.
The syntax of the tar command is irregular – you don’t have to put dashes in front of the operation letters (but you can if you like):
Syntax: tar <operation> [options] -f <archive_file> [<pathnames>]
$ tar cf /tmp/my.tar . # create archive of current directory
$ tar -cf stuff.tar *.c # archive all the .c files
$ tar -xvf my.tar # extract everything into current dir
$ tar xvf my.tar mydir # only extract mydir from the archive
The name of the tar archive can be anything; the suffixes are there simply for human readers to better know what the files contain.
The archive name must always directly follow the -f option with no other option letters in between:
$ tar -tvf my.tar # correct use of -f
$ tar -vft my.tar # WRONG use of -f
$ tar -fvt my.tar # WRONG use of -f
You must always use one of three major operation letters:
-t: list the pathnames in the archive (a table of contents)
-x: extract (all or some) pathnames from the archive
-c: create a new tar archive (erases existing contents!)
You may optionally use some other relevant options:
-f: select the archive pathname (almost always used; must be last option)
-p: preserve permissions when extracting
-v: verbose (more messages about what is happening, or more detail)
-z: the entire archive is gzip compressed (or uncompressed if extracting)
-j: the entire archive is bzip2 compressed (or uncompressed if extracting)
The -f archive pathname option is almost always used, unless you happen to own a tape drive! Always use -f and an archive file name. The archive file name must immediately follow the -f option with no other option letters in between, i.e. tar -tvf my.tar
The -v “verbose” option above lists all the file names as they are put into an archive file, or as they are extracted. This is useful for debugging, but isn’t usually used for a production system where you know exactly what is going into the archive; leave it out for normal use.
If an uncompressed tarball file is damaged, the damage may affect only some of the files in the tarball and the other files, even files stored after the damage point, may still be recoverable.
3.1 Compressed tarballs: tarball.tar.gz and tarball.tar.bz2Index
A compressed tarball is simply a single tarball file that has been compressed with either gzip or bzip2. The compression compresses the entire tarball, not the individual files inside the tarball.
A tarball file may be first created and then compressed as a whole using either the gzip or bzip2 file compression commands:
$ tar -cf tarball.tar *.c # create archive named tarball.tar
$ gzip tarball.tar # compress into tarball.tar.gz
$ tar -cf tarball.tar *.c # create archive named tarball.tar
$ bzip2 tarball.tar # compress into tarball.tar.bz2
Modern versions of tar have an option letter that does this compression for you (less typing). A compressed tar archive can be created and compressed in one step by an option to the tar command itself:
$ tar -czf tarball.tar.gz *.c # create and gzip compress into tarball.tar.gz
$ tar -cjf tarball.tar.bz2 *.c # create and bzip2 compress into tarball.tar.bz2
You generate a table of contents, or extract all the files, using the appropriate de-compression option depending on if and how the tarball file was compressed:
$ tar -tf tarball.tar # table of contents if uncompressed
$ tar -tzf tarball.tar.gz # table of contents if gzip compressed
$ tar -tjf tarball.tar.bz2 # table of contents if bzip2 compressed
$ tar -xf tarball.tar # extract contents (uncompressed)
$ tar -xzf tarball.tar.gz # extract contents (gzip compressed)
$ tar -xjf tarball.tar.bz2 # extract contents (bzip2 compressed)
The tar command doesn’t care what you name your archive file. The gzip compressed tarballs usually have names ending with *.tar.gz or *.tgz and bzip2 compressed tarballs usually have names ending with *.tar.bz2 or *.tb2.
Modern versions of the tar command automatically recognize existing compressed archives and thus don’t require the extra z or j option letters to read compressed archives. You still need the appropriate letter to create a new compressed archive file.
If a compressed tarball file is damaged, all the files following the damage point cannot be decompressed and are usually unrecoverable.
3.2 Using tar to archive or restore a directoryIndex
The tar command will automatically recursively archive entire directories into a tarball if you give it directories. Software is often distributed as a tarball file.
$ cd # go to my home directory
$ tar czf /tmp/homedir.tar.gz . # archive current directory into a file
Do not place the output tarball file in any of the directories being used as input to tar!
When you have a tarball, you can then extract it into the current directory:
$ mkdir /some/backupdir
$ cd /some/backupdir
$ tar xzpf /tmp/homedir.tar.gz # extract the whole archive into current directory
The p option preserves the modes (permissions) of the files as they are extracted.
This legacy use of tar to copy an entire directory has been replaced by cp -a or the rsync command.
You can do a directory copy with tar using a pipe instead of an output file by using the special file name - that stands for either standard output (when creating) or standard input (when extracting):
$ cd
$ tar cf - . | ( cd /some/backupdir && tar xpf - ) # local copy
$ tar cf - . | ( ssh otherhost 'cd /some/dir && tar xpf -' ) # remote host copy
The above uses of tar to copy a directory have been largely supplanted by the -a (archive) option to cp or by the rsync command.
A ZIP file is a single file containing individually compressed files. (This is not the same format as a compressed tarball, which is a single compressed file containing individual uncompressed files.)
Unix/Linux can also manipulate ZIP format file archives (often used on Microsoft systems) using zip and unzip:
$ touch file1 file2 file3
$ zip foo file1 file2 file3 # create foo.zip with three files
adding: file1 (stored 0%)
adding: file2 (stored 0%)
adding: file3 (stored 0%)
$ ls -l foo.zip
-rw-rw-r-- 1 idallen idallen 436 Mar 9 03:44 foo.zip
$ unzip -l foo.zip # list the contents (do not extract)
Archive: foo.zip
Length Date Time Name
--------- ---------- ----- ----
0 2016-03-09 03:44 file1
0 2016-03-09 03:44 file2
0 2016-03-09 03:44 file3
--------- -------
0 3 files
$ rm file?
$ unzip foo.zip # extract all the files
Archive: foo.zip
extracting: file1
extracting: file2
extracting: file3
Other options can preserve directory hierarchy and do other things. See the man page.
If a ZIP file is damaged, the damage usually affects only some of the files in the ZIP file and the other files, even files stored after the damage point, may still recoverable.
Q: Is a tarball an archive of separate, individually compressed files (which is also the structure of a zip file), or does tar archive together all the files first (uncompressed) and then compress the whole archive?
Q: Which is more space-efficient: a zip file or a compressed tar file, and why? (Hint: Consider archiving 1000 copies of the same file.)
Q: Which is less affected by file damage: a zip file or a compressed tar file, and why?
A 
