Writing tests

Each Bats test file is evaluated n+1 times, where n is the number of test cases in the file. The first run counts the number of test cases, then iterates over the test cases and executes each one in its own process.

For more details about how Bats evaluates test files, see Bats Evaluation Process on the wiki.

For sample test files, see examples.

Tagging tests

Starting with version 1.8.0, Bats comes with a tagging system that allows users to categorize their tests and filter according to those categories.

Each test has a list of tags attached to it. Without specification, this list is empty. Tags can be defined in two ways. The first being # bats test_tags=:

# bats test_tags=tag:1, tag:2, tag:3
@test "first test" {
  # ...
}

@test "second test" {
  # ...
}

These tags (tag:1, tag:2, tag:3) will be attached to the test first test. The second test will have no tags attached. Values defined in the # bats test_tags= directive will be assigned to the next @test that is being encountered in the file and forgotten after that. Only the value of the last # bats test_tags= directive before a given test will be used.

Sometimes, we want to give all tests in a file a set of the same tags. This can be achieved via # bats file_tags=. They will be added to all tests in the file after that directive. An additional # bats file_tags= directive will override the previously defined values:

@test "Zeroth test" { 
  # will have no tags
}

# bats file_tags=a:b
# bats test_tags=c:d

@test "First test" { 
  # will be tagged a:b, c:d
}

# bats file_tags=

@test "Second test" {
  # will have no tags
}

Tags are case sensitive and must only consist of alphanumeric characters and _, -, or :. They must not contain whitespaces! The colon is intended as a separator for (recursive) namespacing.

Tag lists must be separated by commas and are allowed to contain whitespace. They must not contain empty tags like test_tags=,b (first tag is empty), test_tags=a,,c, test_tags=a,  ,c (second tag is only whitespace/empty), test_tags=a,b, (third tag is empty).

Every tag starting with bats: (case insensitive!) is reserved for Bats’ internal use.

Special tags

Focusing on tests with bats:focus tag

If a test with the tag bats:focus is encountered in a test suite, all other tests will be filtered out and only those tagged with this tag will be executed.

In focus mode, the exit code of successful runs will be overridden to 1 to prevent CI from silently running on a subset of tests due to an accidentally committed bats:focus tag.
Should you require the true exit code, e.g. for a git bisect operation, you can disable this behavior by setting BATS_NO_FAIL_FOCUS_RUN=1 when running bats, but make sure not to commit this to CI!

Filtering execution

Tags can be used for more finegrained filtering of which tests to run via --filter-tags. This accepts a comma separated list of tags. Only tests that match all of these tags will be executed. For example, bats --filter-tags a,b,c will pick up tests with tags a,b,c, but not tests that miss one or more of those tags.

Additionally, you can specify negative tags via bats --filter-tags a,!b,c, which now won’t match tests with tags a,b,c, due to the b, but will select a,c. To put it more formally, --filter-tags is a boolean conjunction.

To allow for more complex queries, you can specify multiple --filter-tags. A test will be executed, if it matches at least one of them. This means multiple --filter-tags form a boolean disjunction.

A query of --filter-tags a,!b --filter-tags b,c can be translated to: Execute only tests that (have tag a, but not tag b) or (have tag b and c).

An empty tag list matches tests without tags.

Comment syntax

External tools (like shellcheck, shfmt, and various IDE’s) may not support the standard .bats syntax. Because of this, we provide a valid bash alternative:

function invoking_foo_without_arguments_prints_usage { #@test
  run foo
  [ "$status" -eq 1 ]
  [ "${lines[0]}" = "usage: foo <filename>" ]
}

When using this syntax, the function name will be the title in the result output and the value checked when using --filter.

run: Test other commands

Many Bats tests need to run a command and then make assertions about its exit status and output. Bats includes a run helper that invokes its arguments as a command, saves the exit status and output into special global variables, and then returns with a 0 status code so you can continue to make assertions in your test case.

For example, let’s say you’re testing that the foo command, when passed a nonexistent filename, exits with a 1 status code and prints an error message.

@test "invoking foo with a nonexistent file prints an error" {
  run foo nonexistent_filename
  [ "$status" -eq 1 ]
  [ "$output" = "foo: no such file 'nonexistent_filename'" ]
  [ "$BATS_RUN_COMMAND" = "foo nonexistent_filename" ]

}

The $status variable contains the status code of the command, the $output variable contains the combined contents of the command’s standard output and standard error streams, and the $BATS_RUN_COMMAND string contains the command and command arguments passed to run for execution.

If invoked with one of the following as the first argument, run will perform an implicit check on the exit status of the invoked command:

    -N  expect exit status N (0-255), fail if otherwise
    ! expect nonzero exit status (1-255), fail if command succeeds

We can then write the above more elegantly as:

@test "invoking foo with a nonexistent file prints an error" {
  run -1 foo nonexistent_filename
  [ "$output" = "foo: no such file 'nonexistent_filename'" ]
}

A third special variable, the $lines array, is available for easily accessing individual lines of output. For example, if you want to test that invoking foo without any arguments prints usage information on the first line:

@test "invoking foo without arguments prints usage" {
  run -1 foo
  [ "${lines[0]}" = "usage: foo <filename>" ]
}

Note: The run helper executes its argument(s) in a subshell, so if writing tests against environmental side-effects like a variable’s value being changed, these changes will not persist after run completes.

By default run leaves out empty lines in ${lines[@]}. Use run --keep-empty-lines to retain them.

Additionally, you can use --separate-stderr to split stdout and stderr into $output/$stderr and ${lines[@]}/${stderr_lines[@]}.

All additional parameters to run should come before the command. If you want to run a command that starts with -, prefix it with -- to prevent run from parsing it as an option.

When not to use run

In case you only need to check the command succeeded, it is better to not use run, since the following code

run -0 command args ...

is equivalent to

command args ...

(because bats sets set -e for all tests).

Note: In contrast to the above, testing that a command failed is best done via

run ! command args ...

because

! command args ...

will only fail the test if it is the last command and thereby determines the test function’s exit code. This is due to Bash’s decision to (counterintuitively?) not trigger set -e on ! commands. (See also the associated gotcha)

run and pipes

Don’t fool yourself with pipes when using run. Bash parses the pipe outside of run, not internal to its command. Take this example:

run command args ... | jq -e '.limit == 42'

Here, jq receives no input (which is captured by run), executes no filters, and always succeeds, so the test does not work as expected.

To correctly handle commands with pipes see bats_pipe.

bats_pipe: Run commands with pipes

The bats_pipe helper command is meant to handle piping between commands. Its main purpose is to aide the run helper command (which cannot handle pipes, due to bash parsing priority). run command0 | command1 will parse | before run, which is commonly not intended by test authors.

Running run bats_pipe command0 \| command1 will have the piped commands run within the context of the run command, and thus have the output and status variables properly contained within the normal output and status variables.

Note that this requires the usage of \|, not |. This is to avoid bash parsing out | first, instead sending \| to the bats_pipe command for it to parse and set up intended piping. Running bats_pipe with no instances of \| will always fail; this is intended to catch typos (accidentally using |) by the test author.

The bats_pipe command will also properly propagate exit status from the piped commands. The default behavior mimics set -o pipefail, returning the status of the last (rightmost) command that exits with a non-zero status. This ensures that usage of pipes do not mask the exit statuses of earlier commands.

@test "invoking foo piped to bar" {
  run bats_pipe foo \| bar
  # asserting foo or bar would return 17 (from foo if bar returns 0).
  [ "$status" -eq 17 ]
  [ "$output" = "bar output." ]
}

Alternatively, if the test always cares about the status of a specific command, the -<N> option can be given (e.g. -0) to always return the status of the command of interest.

@test "invoking foo piped to bar always return foo status" {
  run bats_pipe -0 foo \| bar
  # status of bar is ignored, status is always from foo.
  [ "$status" -eq 2 ]
  [ "$output" = "bar output." ]
}

Similarly, --returned-status N (or --returned-status=N) can be used for similar functionality. This option supports negative values, which always return the of the command starting from the end and in reverse order.

@test "invoking foo piped to bar always return foo status" {
  run bats_pipe --returned-status -2 foo \| bar
  # status of bar is ignored, status is always from foo.
  [ "$status" -eq 2 ]
  [ "$output" = "bar output." ]
}

Piping of command output is especially helpful when the output needs to be modified in some way (e.g. the command outputs binary data into stdout, which cannot be stored as-is in an environment variable).

@test "invoking foo that returns binary data" {
  run bats_pipe foo \| hexdump -v -e "1/1 \"0x%02X \""
  [ "$status" -eq 17 ]
  [[ "$output" =~ 0xDE\ 0xAD ]]
}

Any number of pipes can be used in conjunction to chain output between some set of running commands.

load: Share common code

You may want to share common code across multiple test files. Bats includes a convenient load command for sourcing a Bash source files relative to the current test file and from absolute paths.

For example, if you have a Bats test in test/foo.bats, the command

load test_helper.bash

will source the script test/test_helper.bash in your test file (limitations apply, see below). This can be useful for sharing functions to set up your environment or load fixtures. load delegates to Bash’s source command after resolving paths.

If load encounters errors - e.g. because the targeted source file errored - it will print a message with the failing library and Bats exits.

To allow to use load in conditions bats_load_safe has been added. bats_load_safe prints a message and returns 1 if a source file cannot be loaded instead of exiting Bats. Aside from that bats_load_safe acts exactly like load.

As pointed out by @iatrou in https://www.tldp.org/LDP/abs/html/declareref.html, using the declare builtin restricts scope of a variable. Thus, since actual source-ing is performed in context of the load function, declared symbols will not be made available to callers of load.

load argument resolution

load supports the following arguments:

  • absolute paths

  • relative paths (to the current test file)

For backwards compatibility load first searches for a file ending in .bash (e.g. load test_helper searches for test_helper.bash before it looks for test_helper). This behaviour is deprecated and subject to change, please use exact filenames instead.

If argument is an absolute path load tries to determine the load path directly.

If argument is a relative path or a name load looks for a matching path in the directory of the current test.

bats_load_library: Load system wide libraries

Some libraries are installed on the system, e.g. by npm or brew. These should not be loaded, as their path depends on the installation method. Instead, one should use bats_load_library together with setting BATS_LIB_PATH, a PATH-like colon-delimited variable.

bats_load_library has two modes of resolving requests:

  1. by relative path from the BATS_LIB_PATH to a file in the library

  2. by library name, expecting libraries to have a load.bash entrypoint

For example if your BATS_LIB_PATH is set to ~/.bats/libs:/usr/lib/bats, then bats_load_library test_helper would look for existing files with the following paths:

  • ~/.bats/libs/test_helper

  • ~/.bats/libs/test_helper/load.bash

  • /usr/lib/bats/test_helper

  • /usr/lib/bats/test_helper/load.bash

The first existing file in this list will be sourced.

If you want to load only part of a library or the entry point is not named load.bash, you have to include it in the argument: bats_load_library library_name/file_to_load will try

  • ~/.bats/libs/library_name/file_to_load

  • ~/.bats/libs/library_name/file_to_load/load.bash

  • /usr/lib/bats/library_name/file_to_load

  • /usr/lib/bats/library_name/file_to_load/load.bash

Apart from the changed lookup rules, bats_load_library behaves like load.

Note: As seen above load.bash is the entry point for libraries and meant to load more files from its directory or other libraries.

Note: Obviously, the actual BATS_LIB_PATH is highly dependent on the environment. To maintain a uniform location across systems, (distribution) package maintainers are encouraged to use /usr/lib/bats/ as the install path for libraries where possible. However, if the package manager has another preferred location, like npm or brew, you should use this instead.

skip: Easily skip tests

Tests can be skipped by using the skip command at the point in a test you wish to skip.

@test "A test I don't want to execute for now" {
  skip
  run foo
  [ "$status" -eq 0 ]
}

Optionally, you may include a reason for skipping:

@test "A test I don't want to execute for now" {
  skip "This command will return zero soon, but not now"
  run foo
  [ "$status" -eq 0 ]
}

Or you can skip conditionally:

@test "A test which should run" {
  if [ foo != bar ]; then
    skip "foo isn't bar"
  fi

  run foo
  [ "$status" -eq 0 ]
}

Note: setup and teardown hooks still run for skipped tests.

setup and teardown: Pre- and post-test hooks

You can define special setup and teardown functions, which run before and after each test case, respectively. Use these to load fixtures, set up your environment, and clean up when you’re done.

You can also define setup_file and teardown_file, which will run once before the first test’s setup and after the last test’s teardown for the containing file. Variables that are exported in setup_file will be visible to all following functions (setup, the test itself, teardown, teardown_file).

Similarly, there is setup_suite (and teardown_suite) which run once before (and after) all tests of the test run.

Note: As setup_suite and teardown_suite are intended for all files in a suite, they must be defined in a separate setup_suite.bash file. Automatic discovery works by searching for setup_suite.bash in the folder of the first *.bats file of the suite. If this automatism does not work for your usecase, you can work around by specifying --setup-suite-file on the bats command. If you have a setup_suite.bash, it must define setup_suite! However, defining teardown_suite is optional.

Example of setup/{,_file,_suite} (and teardown{,_file,_suite}) call order For example the following call order would result from two files (file 1 with tests 1 and 2, and file 2 with test3) with a corresponding `setup_suite.bash` file being tested:
setup_suite # from setup_suite.bash
  setup_file # from file 1, on entering file 1
    setup
      test1
    teardown
    setup
      test2
    teardown
  teardown_file # from file 1, on leaving file 1
  setup_file # from file 2,  on enter file 2
    setup
      test3
    teardown
  teardown_file # from file 2,  on leaving file 2
teardown_suite # from setup_suite.bash

Note that the teardown* functions can fail a test, if their return code is nonzero. This means, using return 1 or having the last command in teardown fail, will fail the teardown. Unlike @test, failing commands within teardown won’t trigger failure as ERREXIT is disabled.

Example of different teardown failure modes
teardown() {
  false # this will fail the test, as it determines the return code
}

teardown() {
  false # this won't fail the test ...
  echo some more code # ... and this will be executed too!
}

teardown() {
  return 1 # this will fail the test, but the rest won't be executed
  echo some more code
}

teardown() {
  if true; then
    false # this will also fail the test, as it is the last command in this function
  else
    true
  fi
}

bats_require_minimum_version <Bats version number>

Added in v1.7.0

Code for newer versions of Bats can be incompatible with older versions. In the best case this will lead to an error message and a failed test suite. In the worst case, the tests will pass erroneously, potentially masking a failure.

Use bats_require_minimum_version <Bats version number> to avoid this. It communicates in a concise manner, that you intend the following code to be run under the given Bats version or higher.

Additionally, this function will communicate the current Bats version floor to subsequent code, allowing e.g. Bats’ internal warning to give more informed warnings.

Note: By default, calling bats_require_minimum_version with versions before Bats 1.7.0 will fail regardless of the required version as the function is not available. However, you can use the bats-backports plugin to make your code usable with older versions, e.g. during migration while your CI system is not yet upgraded.

Code outside of test cases

In general you should avoid code outside tests, because each test file will be evaluated many times. However, there are situations in which this might be useful, e.g. when you want to check for dependencies and fail immediately if they’re not present.

In general, you should avoid printing outside of @test, setup* or teardown* functions. Have a look at section printing to the terminal for more details.

File descriptor 3 (read this if Bats hangs)

Bats makes a separation between output from the code under test and output that forms the TAP stream (which is produced by Bats internals). This is done in order to produce TAP-compliant output. In the Printing to the terminal section, there are details on how to use file descriptor 3 to print custom text properly.

A side effect of using file descriptor 3 is that, under some circumstances, it can cause Bats to block and execution to seem dead without reason. This can happen if a child process is spawned in the background from a test. In this case, the child process will inherit file descriptor 3. Bats, as the parent process, will wait for the file descriptor to be closed by the child process before continuing execution. If the child process takes a lot of time to complete (eg if the child process is a sleep 100 command or a background service that will run indefinitely), Bats will be similarly blocked for the same amount of time.

To prevent this from happening, close FD 3 explicitly when running any command that may launch long-running child processes, e.g. command_name 3>&- .

Printing to the terminal

Bats produces output compliant with version 12 of the TAP protocol. The produced TAP stream is by default piped to a pretty formatter for human consumption, but if Bats is called with the -t flag, then the TAP stream is directly printed to the console.

This has implications if you try to print custom text to the terminal. As mentioned in File descriptor 3, bats provides a special file descriptor, &3, that you should use to print your custom text. Here are some detailed guidelines to refer to:

  • Printing from within a test function:

    • First you should consider if you want the text to be always visible or only when the test fails. Text that is output directly to stdout or stderr (file descriptor 1 or 2), ie echo 'text' is considered part of the test function output and is printed only on test failures for diagnostic purposes, regardless of the formatter used (TAP or pretty).

    • To have text printed unconditionally from within a test function you need to redirect the output to file descriptor 3, eg echo 'text' >&3. This output will become part of the TAP stream. You are encouraged to prepend text printed this way with a hash (eg echo '# text' >&3) in order to produce 100% TAP compliant output. Otherwise, depending on the 3rd-party tools you use to analyze the TAP stream, you can encounter unexpected behavior or errors.

  • Printing from within the setup* or teardown* functions: The same hold true as for printing with test functions.

  • Printing outside test or setup*/teardown* functions:

    • You should avoid printing in free code: Due to the multiple executions contexts (setup_file, multiple @tests) of test files, output will be printed more than once.

    • Regardless of where text is redirected to (stdout, stderr or file descriptor 3) text is immediately visible in the terminal, as it is not piped into the formatter.

    • Text printed to stdout may interfere with formatters as it can make output non-compliant with the TAP spec. The reason for this is that such output will be produced before the plan line is printed, contrary to the spec that requires the plan line to be either the first or the last line of the output.

    • Due to internal pipes/redirects, output to stderr is always printed first.

Special variables

There are several global variables you can use to introspect on Bats tests:

  • $BATS_RUN_COMMAND is the run command used in your test case.

  • $BATS_TEST_FILENAME is the fully expanded path to the Bats test file.

  • $BATS_TEST_DIRNAME is the directory in which the Bats test file is located.

  • $BATS_TEST_NAMES is an array of function names for each test case.

  • $BATS_TEST_NAME is the name of the function containing the current test case.

  • BATS_TEST_NAME_PREFIX will be prepended to the description of each test on stdout and in reports.

  • $BATS_TEST_DESCRIPTION is the description of the current test case.

  • BATS_TEST_RETRIES is the maximum number of additional attempts that will be made on a failed test before it is finally considered failed. The default of 0 means the test must pass on the first attempt.

  • BATS_TEST_TIMEOUT is the number of seconds after which a test (including setup) will be aborted and marked as failed. Updates to this value in setup() or @test cannot change the running timeout countdown, so the latest useful update location is setup_file().

  • $BATS_TEST_NUMBER is the (1-based) index of the current test case in the test file.

  • $BATS_SUITE_TEST_NUMBER is the (1-based) index of the current test case in the test suite (over all files).

  • $BATS_TEST_TAGS the tags of the current test.

  • $BATS_TMPDIR is the base temporary directory used by bats to create its temporary files / directories. (default: $TMPDIR. If $TMPDIR is not set, /tmp is used.)

  • $BATS_RUN_TMPDIR is the location to the temporary directory used by bats to store all its internal temporary files during the tests. (default: $BATS_TMPDIR/bats-run-$BATS_ROOT_PID-XXXXXX)

  • $BATS_FILE_EXTENSION (default: bats) specifies the extension of test files that should be found when running a suite (via bats [-r] suite_folder/)

  • $BATS_SUITE_TMPDIR is a temporary directory common to all tests of a suite. Could be used to create files required by multiple tests.

  • $BATS_FILE_TMPDIR is a temporary directory common to all tests of a test file. Could be used to create files required by multiple tests in the same test file.

  • $BATS_TEST_TMPDIR is a temporary directory unique for each test. Could be used to create files required only for specific tests.

  • $BATS_VERSION is the version of Bats running the test.

Libraries and Add-ons

Bats supports loading external assertion libraries and helpers. Those under bats-core are officially supported libraries (integration tests welcome!):

and some external libraries, supported on a “best-effort” basis: