clitest – Command Line Tester
clitest is a portable POSIX shell script that performs automatic testing in Unix command lines.
It's the same concept as in Python's doctest module: you document both the commands and their expected output, using the familiar interactive prompt format, and a specialized tool tests them.
In fact, the doctest official description can also be used for clitest:
The doctest module searches for pieces of text that look like interactive Python sessions, and then executes those sessions to verify that they work exactly as shown.
The clitest command searches for pieces of text that look like interactive Unix command lines, and then executes those command lines to verify that they work exactly as shown.
Download & install
The full program is just a single shell script file.
Save it, make it executable and move it to a
curl -sOL https://raw.githubusercontent.com/aureliojargas/clitest/master/clitest chmod +x clitest sudo mv clitest /usr/bin
Now check if everything is fine:
$ clitest -V clitest HEAD https://github.com/aureliojargas/clitest/tree/HEAD $
Save the commands and their expected output in a text file:
$ echo "Hello World" Hello World $ cd /tmp $ pwd /tmp $ cd "$OLDPWD" $
Use clitest to run these commands and check their output:
$ clitest examples/intro.txt #1 echo "Hello World" #2 cd /tmp #3 pwd #4 cd "$OLDPWD" OK: 4 of 4 tests passed $
There's no syntax to learn.
The test files are identical to the good old command line interface (CLI) you're so familiar:
$ echo "one:two:three:four:five:six" | cut -d : -f 1 one $ echo "one:two:three:four:five:six" | cut -d : -f 4 four $ echo "one:two:three:four:five:six" | cut -d : -f 1,4 one:four $ echo "one:two:three:four:five:six" | cut -d : -f 4,1 one:four $ echo "one:two:three:four:five:six" | cut -d : -f 1-4 one:two:three:four $ echo "one:two:three:four:five:six" | cut -d : -f 4- four:five:six $
Just paste your shell session inside a text file and you have a ready-to-use test suite.
$ clitest examples/cut.txt #1 echo "one:two:three:four:five:six" | cut -d : -f 1 #2 echo "one:two:three:four:five:six" | cut -d : -f 4 #3 echo "one:two:three:four:five:six" | cut -d : -f 1,4 #4 echo "one:two:three:four:five:six" | cut -d : -f 4,1 #5 echo "one:two:three:four:five:six" | cut -d : -f 1-4 #6 echo "one:two:three:four:five:six" | cut -d : -f 4- OK: 6 of 6 tests passed $
Clitest can also extract and run command lines from documentation,
such as Markdown files. This very
README.md file you are now reading
is testable with
clitest README.md. All the command lines inside it
will be run and checked.
No more malfunctioning shell commands in your READMEs, you can have testable documentation.
Given the following Markdown sample document:
The numeric ranges of the Unix command "cut" ============================================ Use single numbers to extract one specific field: $ echo "one:two:three:four:five:six" | cut -d : -f 1 one $ echo "one:two:three:four:five:six" | cut -d : -f 4 four Use commas to inform more than one field: $ echo "one:two:three:four:five:six" | cut -d : -f 1,4 one:four Note that inverting the order will *not* invert the output: $ echo "one:two:three:four:five:six" | cut -d : -f 4,1 one:four Use an hyphen to inform a range of fields, from one to four: $ echo "one:two:three:four:five:six" | cut -d : -f 1-4 one:two:three:four If you omit the second range number, it matches until the last: $ echo "one:two:three:four:five:six" | cut -d : -f 4- four:five:six cut is cool, isn't it?
It is a technical article, not a boring code-only test file. You can read its final (formatted) version here.
You can give this article to clitest, who will identify all the shell command lines inside it, run them and check if the results are the same.
$ clitest --prefix tab examples/cut.md #1 echo "one:two:three:four:five:six" | cut -d : -f 1 #2 echo "one:two:three:four:five:six" | cut -d : -f 4 #3 echo "one:two:three:four:five:six" | cut -d : -f 1,4 #4 echo "one:two:three:four:five:six" | cut -d : -f 4,1 #5 echo "one:two:three:four:five:six" | cut -d : -f 1-4 #6 echo "one:two:three:four:five:six" | cut -d : -f 4- OK: 6 of 6 tests passed $
Note the use of
--prefix tab option, to inform clitest that the code
blocks are prefixed by a tab in this Markdown file. For files with
4-spaces indented code blocks, use
--prefix 4. When using
non-indented fenced code blocks (```), such as this README.md,
no prefix option is needed.
Examples of testable documentation handled by clitest:
Alternative Syntax: Inline Output
Now a nice extension to the original idea. Using the special marker
#=> you can embed the expected command output at the end of the
$ echo "foo" #=> foo $ echo $((10 + 2)) #=> 12
This is the same as doing:
$ echo "foo" foo $ echo $((10 + 2)) 12 $
Inline outputs are very readable when testing series of commands that result in short texts.
$ echo "abcdef" | cut -c 1 #=> a $ echo "abcdef" | cut -c 4 #=> d $ echo "abcdef" | cut -c 1,4 #=> ad $ echo "abcdef" | cut -c 1-4 #=> abcd
Note: If needed, you can change this marker (i.e., to
###) at the top of the script or using the
When using the
#=> marker, you can take advantage of special options
to change the default output matching method.
$ head /etc/passwd #=> --lines 10 $ tac /etc/passwd | tac #=> --file /etc/passwd $ cat /etc/passwd #=> --egrep ^root: $ echo $((2 + 10)) #=> --regex ^\d+$ $ make test #=> --exit 0 $ pwd #=> --eval echo $PWD
#=> --linesthe test will pass if the command output has exactly
Nlines. Handy when the output text is variable (unpredictable), but the number of resulting lines is constant.
#=> --filethe test will pass if the command output matches the contents of an external file. Useful to organize long/complex outputs into files.
#=> --egrepthe test will pass if
egrepmatches at least one line of the command output.
#=> --regexthe test will pass if the command output is matched by a Perl regular expression. A multiline output is matched as a single string, with inner
\n's. Use the
(?ims)modifiers when needed.
#=> --exitthe test will pass if the exit code of the command is equal to the code specified. Useful when testing commands that generate variable output (or no output at all), and the exit code is the best indication of success. Both STDIN and STDOUT are ignored when using this option.
#=> --evalthe test will pass if both commands result in the same output. Useful to expand variables which store the full or partial output.
$ clitest --help Usage: clitest [options] <file ...> Options: -1, --first Stop execution upon first failed test -l, --list List all the tests (no execution) -L, --list-run List all the tests with OK/FAIL status -t, --test RANGE Run specific tests, by number (1,2,4-7) -s, --skip RANGE Skip specific tests, by number (1,2,4-7) --pre-flight COMMAND Execute command before running the first test --post-flight COMMAND Execute command after running the last test -q, --quiet Quiet operation, no output shown -V, --version Show program version and exit Customization options: -P, --progress TYPE Set progress indicator: test, number, dot, none --color WHEN Set when to use colors: auto, always, never --diff-options OPTIONS Set diff command options (default: '-u') --inline-prefix PREFIX Set inline output prefix (default: '#=> ') --prefix PREFIX Set command line prefix (default: '') --prompt STRING Set prompt string (default: '$ ') $
0- All tests passed, or normal operation (--help, --list, …)
1- One or more tests have failed
2- An error occurred (file not found, invalid range, …)
--first option (or the short version
-1) to abort the
execution when any test fails.
Useful for Continuous Integration (CI), or when running sequential tests where the next test depends on the correct result of the previous.
When automating the tests execution, use
--quiet to show no output
and just check the exit code to make sure all tests have passed. Using
--first to fail fast is also a good idea in this case.
if clitest --quiet --first tests.txt then # all tests passed else # one or more tests failed :( fi
Run specific tests
To rerun a specific problematic test, or to limit the execution to a
set of tests, use
--test. To ignore one or more tests, use
If needed, you can combine both options to inform a very specific test
clitest --test 1-10 tests.txt # Run the first 10 tests clitest --test 1,2,6-8 tests.txt # Run tests #1, #2, #6, #7 and #8 clitest --skip 11,15 tests.txt # Run all tests, except #11 and #15 clitest -t 1-10 -s 5 tests.txt # Run first 10 tests, but skip #5
You can run a preparing script or command before the first test with
--pre-flight, for setting env variables and create auxiliary files.
At the end of all tests, run a final cleanup script/command with
--post-flight to remove temporary files or other transient data.
clitest --pre-flight ./test-init.sh --post-flight 'rm *.tmp' tests.txt
Use the customization options to extract and test command lines from documents or wiki pages. For example, to test all the command line examples listed inside a Markdown file using the 4-spaces syntax for code blocks:
clitest --prefix 4 README.md
Or maybe you use a different prompt (
$PS1) in your documentation?
clitest --prefix 4 --prompt '[john@localhost ~]$ ' README.md
Use any text file format for the tests, it doesn't matter. The command lines just need to be grepable and have a fixed prefix (or none). Even Windows text files (CR+LF) will work fine.
The cmdline power is available in your test files: use variables, pipes, redirection, create files, folders, move around…
All the commands are tested in the same shell. Defined variables, aliases and functions will persist between tests.
Both STDIN and STDOUT are catch, you can also test error messages.
To test STDOUT/STDERR and the exit code at the same time, add a
;echo $?after the command.
Use an empty
$prompt to close the last command output.
In the output, every single char (blank or not) counts. Any difference will cause a test to fail. To ignore the difference in blanks, use
--diff-options '-u -w'.
<BLANKLINE>, in clitest blank lines in the command output aren't a problem. Just insert them normally.
To test outputs with no final
\n, such as
printf foo, use
#=> --regex ^foo$.
In multifile mode, the current folder (
$PWD) is reset when starting to test a new file. This avoids that a
cdcommand in a previous file will affect the next.
Multiline prompts (
$PS2) are not yet supported.
Ellipsis (as in doctest) are not supported. Use
Choose the execution shell
The clitest shebang is
#!/bin/sh. That's the default shell that will be used to run your test command lines. Depending on the system, that path points to a different shell, such as ash, dash, or bash (running in POSIX mode).
To force your test commands to always run on a specific shell, just call the desired shell before:
clitest tests.txt # Uses /bin/sh bash clitest tests.txt # Uses Bash ksh clitest tests.txt # Uses Korn Shell
This script was carefully coded to be portable between POSIX shells.
It was tested in:
- Bash 3.2
- dash 0.5.5.1
- ksh 93u 2011-02-08
- zsh 5.0.5
Portability issues are considered serious bugs, please report them!
Developers: Learn more about portability in POSIX shells:
- How to make bash scripts work in dash
- Ubuntu — Dash as /bin/sh
- Rich’s sh (POSIX shell) tricks
- Official POSIX specification: Shell & Utilities
A shell script to test shell commands.
No other language or environment involved.