When programming in the Bash scripting language, you sometimes need to create a temporary file. For instance, you might need to have an intermediary file you can commit to disk so you can process it with another command. It's easy to create a file such as
temp or anything ending in
.tmp. However, those names are just as likely to be generated by some other process, so you could accidentally overwrite an existing temporary file. And besides that, you shouldn't have to expend mental effort coming up with names that seem unique. The
mktemp command on Fedora-based systems and
tempfile on Debian-based systems are specially designed to alleviate that burden by making it easy to create, use, and remove unique files.
Create a temporary file
tempfile create a temporary file as their default action and print the name and location of the file as output:
$ tempfile /tmp/fileR5dt6r $ mktemp /tmp/tmp.ojEfvMaJEp
Unless you specify a different path, the system places temporary files in the
/tmp directory. For
mktemp, use the
-p option to specify a path:
$ mktemp -p ~/Demo /home/tux/Demo/tmp.i8NuhzbEJN
tempfile, use the
$ tempfile --directory ~/Demo/ /home/sek/Demo/fileIhg9aX
Find your temporary file
The problem with using an auto-generated temporary file is that you have no way of knowing what its name is going to be. That's why both commands return the generated file name as output. You can use an interactive shell such as Konsole, GNOME Terminal, or rxvt to use the filename displayed on your terminal to interact with the file.
However, if you're writing a script, there's no way for you to intervene by reading the name of the file and using it in the following commands.
The authors of
tempfile thought of that problem, and there's an easy fix. The terminal sends output to a stream called stdout. You can capture stdout by setting a variable to the results of a command launched in a subshell:
$ TMPFILE=$(mktemp -p ~/Demo) $ echo $TMPFILE /home/tux/Demo/tmp.PjP3g6lCq1
$TMPFILE when referring to the file, and it's the same as interacting directly with the file itself.
Create a temporary directory with mktemp
You can also use the
mktemp command to create a directory instead of a file:
$ mktemp --directory -p ~/Demo/ /home/tux/Demo/tmp.68ukbuluqI $ file /home/tux/Demo/tmp.68ukbuluqI /home/tux/Demo/tmp.68ukbuluqI: directory
Customize temporary names
Sometimes you might want an element of predictability in even your pseudo-randomly generated filenames. You can customize the names of your temporary files with both commands.
mktemp, you can add a suffix to your filename:
$ mktemp -p ~/Demo/ --suffix .mine /home/tux/Demo/tmp.dufLYfwJLO.mine
tempfile, you can set a prefix and a suffix:
$ tempfile --directory ~/Demo/ \ --prefix tt_ --suffix .mine /home/tux/Demo/tt_0dfu5q.mine
Tempfile as touch
You can also set a custom name with
$ tempfile --name not_random not_random
When you use the
--name option, it's absolute, ignoring all other forms of customization. In fact, it even ignores the
$ tempfile --directory ~/Demo \ --prefix this_is_ --suffix .all \ --name not_random_at not_random_at
In a way,
tempfile can be a substitute for
test because it refuses to create a file that already exists:
$ tempfile --name example.txt open: file exists
tempfile command isn't installed on all Linux distributions by default, so you must ensure that it exists before you use it as a hack around
test in a script.
Install mktemp and tempfile
GNU Core Utils includes the
mktemp command. Major distributions include Core Utils by default (it's the same package that contains
du, and other essential commands).
The Debian Utils package includes the
tempfile command and is installed by default on most Debian-based distributions and Slackware Linux.
Temporary files are convenient because there's no confusion about whether they're safe to delete. They're temporary, meant to be used as needed and discarded without a second thought. Use them when you need them, and clear them out when you're done.