bashT

Basht is a minimalist Bash testing utility. You can write tests in pure Bash, just pass it one or more Bash source files that define tests. Tests are functions that start with T_:

T_additionUsingBC() {
  result="$(echo 2+2 | bc)"
  [[ "$result" -eq 4 ]]
}

T_additionUsingDC() {
  result="$(echo 2 2+p | dc)"
  [[ "$result" -eq 4 ]]
}

Tests that return non-zero fail. Only the return value will fail a test, as set -e is not used. You can fail a test at any time by explicitly calling return with non-zero.

Getting basht

Download and uncompress the latest binary tarball from releases.

Alternatively, if you happen to also be using Go, you can install with go get:

$ go get github.com/progrium/basht

Running tests

Any filenames passed to basht are loaded and any tests found will be run. Take advantage of globbing for multiple files or directories of tests.

$ basht tests/example.bash
=== RUN T_additionUsingBC
--- PASS T_additionUsingBC (0s)
=== RUN T_additionUsingDC
--- PASS T_additionUsingDC (0s)

Ran 2 tests.

PASS

If tests pass, basht will exit zero. If any tests failed, basht exists non-zero with the number of failed tests.

$ basht tests/fails.bash
=== RUN T_failEquals
--- FAIL T_failEquals (0s)
=== RUN T_failMessage
--- FAIL T_failMessage (0s)
    tests/fails.bash:19: This is a fail message.

=== RUN T_failReturn
--- FAIL T_failReturn (0s)
=== RUN T_failSuccess
--- FAIL T_failSuccess (0s)

Ran 4 tests. 4 failed.

FAIL

Macros

Basht provides no special assertions or helpers. However, there is one macro basht provides:

$T_fail

Calling $T_fail <message> marks a test as failed and includes a failure message. There is an example above of how this is shown in the output. It includes the filename and line number with the message.

Keep in mind it does not return, so if used before the end of a test, you must return after.

T_failMessage() {
	false || $T_fail "This is a fail message."
}

T_failMultiple() {
	if ! something; then
		$T_fail "Something failed."
		return
	fi
	if ! another; then
		$T_fail "Another failed."
		return
	fi
}

Why not Bats or shunit2?

Good question. I've used both and enjoyed using them until I got tired of what I didn't like. With Bats these issues bothered me:

Useless syntactic sugar making test files require Bats instead of just Bash to process, making Bats more complicated and harder to debug.
Using set -e means you have to use the run helper for nearly everythng in larger tests.
The other Bats-specific helpers that could be replaced with idiomatic Bash-isms.
The multi-file, multi-directory install.

On the plus side, it did take a much more lightweight approach to unit testing. But the issues made me turn to shunit2. It addressed all the above issues, but then I ran into:

For what it did and what I used, it was not worth the 1000 lines of overly portable, overly clever shell script.
With all it did, I had to modify it anyway, and ended up putting it in the source of every project that used it.

Basht might not be for everybody, but I wrote it and use it because:

It's written in and made for pure Bash.
It's distributed as a single file binary.
It's only about 50 lines and does everything I need.
It's inspired in design and appearance by Go's testing.

License

BSD