Unit testing support.

Unlike most Skylib files, this exports four modules:

  • unittest contains functions to declare and define unit tests for ordinary Starlark functions;
  • analysistest contains functions to declare and define tests for analysis phase behavior of a rule, such as a given target's providers or registered actions;
  • loadingtest contains functions to declare and define tests for loading phase behavior, such as macros and native.*;
  • asserts contains the assertions used within tests.

See https://bazel.build/extending/concepts for background about macros, rules, and the different phases of a build.

unittest_toolchain

ATTRIBUTES

NameDescriptionTypeMandatoryDefault
nameA unique name for this target.Namerequired
escape_chars_withDictionary of characters that need escaping in test failure message to prefix appended to escape those characters. For example, {"%": "%", ">": "^"} would replace % with %% and > with ^> in the failure message before that is included in success_templ.Dictionary: String -> Stringoptional{}
escape_other_chars_withString to prefix every character in test failure message which is not a key in escape_chars_with before including that in success_templ. For example, "" would prefix every character in the failure message (except those in the keys of escape_chars_with) with \.Stringoptional""
failure_templTest script template with a single %s. That placeholder is replaced with the lines in the failure message joined with the string specified in join_with. The resulting script should print the failure message and exit with non-zero status.Stringrequired
file_extFile extension for test script, including leading dot.Stringrequired
join_onString used to join the lines in the failure message before including the resulting string in the script specified in failure_templ.Stringrequired
success_templTest script generated when the test passes. Should exit with status 0.Stringrequired

analysistest.begin

Begins an analysis test.

This should be the first function called in an analysis test implementation function. It initializes a “test environment” that is used to collect assertion failures so that they can be reported and logged at the end of the test.

PARAMETERS

NameDescriptionDefault Value
ctxThe Starlark context. Pass the implementation function's ctx argument in verbatim.none

RETURNS

A test environment struct that must be passed to assertions and finally to analysistest.end. Do not rely on internal details about the fields in this struct as it may change.

analysistest.end

Ends an analysis test and logs the results.

This must be called and returned at the end of an analysis test implementation function so that the results are reported.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by analysistest.begin.none

RETURNS

A list of providers needed to automatically register the analysis test result.

analysistest.fail

Unconditionally causes the current test to fail.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
msgThe message to log describing the failure.none

analysistest.make

Creates an analysis test rule from its implementation function.

An analysis test verifies the behavior of a “real” rule target by examining and asserting on the providers given by the real target.

Each analysis test is defined in an implementation function that must then be associated with a rule so that a target can be built. This function handles the boilerplate to create and return a test rule and captures the implementation function's name so that it can be printed in test feedback.

An example of an analysis test:

def _your_test(ctx):
  env = analysistest.begin(ctx)

  # Assert statements go here

  return analysistest.end(env)

your_test = analysistest.make(_your_test)

Recall that names of test rules must end in _test.

PARAMETERS

NameDescriptionDefault Value
implThe implementation function of the unit test.none
expect_failureIf true, the analysis test will expect the target_under_test to fail. Assertions can be made on the underlying failure using asserts.expect_failureFalse
attrsAn optional dictionary to supplement the attrs passed to the unit test's rule() constructor.{}
fragmentsAn optional list of fragment names that can be used to give rules access to language-specific parts of configuration.[]
config_settingsA dictionary of configuration settings to change for the target under test and its dependencies. This may be used to essentially change ‘build flags’ for the target under test, and may thus be utilized to test multiple targets with different flags in a single build{}
extra_target_under_test_aspectsAn optional list of aspects to apply to the target_under_test in addition to those set up by default for the test harness itself.[]
docA description of the rule that can be extracted by documentation generating tools.""

RETURNS

A rule definition that should be stored in a global whose name ends in _test.

analysistest.target_actions

Returns a list of actions registered by the target under test.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by analysistest.begin.none

RETURNS

A list of actions registered by the target under test

analysistest.target_bin_dir_path

Returns ctx.bin_dir.path for the target under test.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by analysistest.begin.none

RETURNS

Output bin dir path string.

analysistest.target_under_test

Returns the target under test.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by analysistest.begin.none

RETURNS

The target under test.

asserts.equals

Asserts that the given expected and actual values are equal.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
expectedThe expected value of some computation.none
actualThe actual value returned by some computation.none
msgAn optional message that will be printed that describes the failure. If omitted, a default will be used.None

asserts.expect_failure

Asserts that the target under test has failed with a given error message.

This requires that the analysis test is created with analysistest.make() and expect_failures = True is specified.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by analysistest.begin.none
expected_failure_msgThe error message to expect as a result of analysis failures.""

asserts.false

Asserts that the given condition is false.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
conditionA value that will be evaluated in a Boolean context.none
msgAn optional message that will be printed that describes the failure. If omitted, a default will be used."Expected condition to be false, but was true."

asserts.new_set_equals

Asserts that the given expected and actual sets are equal.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
expectedThe expected set resulting from some computation.none
actualThe actual set returned by some computation.none
msgAn optional message that will be printed that describes the failure. If omitted, a default will be used.None

asserts.set_equals

Asserts that the given expected and actual sets are equal.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
expectedThe expected set resulting from some computation.none
actualThe actual set returned by some computation.none
msgAn optional message that will be printed that describes the failure. If omitted, a default will be used.None

asserts.true

Asserts that the given condition is true.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
conditionA value that will be evaluated in a Boolean context.none
msgAn optional message that will be printed that describes the failure. If omitted, a default will be used."Expected condition to be true, but was false."

loadingtest.equals

Creates a test case for asserting state at LOADING phase.

PARAMETERS

NameDescriptionDefault Value
envLoading test env created from loadingtest.makenone
test_caseName of the test casenone
expectedExpected value to testnone
actualActual value received.none

RETURNS

None, creates test case

loadingtest.make

Creates a loading phase test environment and test_suite.

PARAMETERS

NameDescriptionDefault Value
namename of the suite of tests to createnone

RETURNS

loading phase environment passed to other loadingtest functions

register_unittest_toolchains

Registers the toolchains for unittest users.

unittest.begin

Begins a unit test.

This should be the first function called in a unit test implementation function. It initializes a “test environment” that is used to collect assertion failures so that they can be reported and logged at the end of the test.

PARAMETERS

NameDescriptionDefault Value
ctxThe Starlark context. Pass the implementation function's ctx argument in verbatim.none

RETURNS

A test environment struct that must be passed to assertions and finally to unittest.end. Do not rely on internal details about the fields in this struct as it may change.

unittest.end

Ends a unit test and logs the results.

This must be called and returned at the end of a unit test implementation function so that the results are reported.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none

RETURNS

A list of providers needed to automatically register the test result.

unittest.fail

Unconditionally causes the current test to fail.

PARAMETERS

NameDescriptionDefault Value
envThe test environment returned by unittest.begin.none
msgThe message to log describing the failure.none

unittest.make

Creates a unit test rule from its implementation function.

Each unit test is defined in an implementation function that must then be associated with a rule so that a target can be built. This function handles the boilerplate to create and return a test rule and captures the implementation function's name so that it can be printed in test feedback.

The optional attrs argument can be used to define dependencies for this test, in order to form unit tests of rules.

The optional toolchains argument can be used to define toolchain dependencies for this test.

An example of a unit test:

def _your_test(ctx):
  env = unittest.begin(ctx)

  # Assert statements go here

  return unittest.end(env)

your_test = unittest.make(_your_test)

Recall that names of test rules must end in _test.

PARAMETERS

NameDescriptionDefault Value
implThe implementation function of the unit test.none
attrsAn optional dictionary to supplement the attrs passed to the unit test's rule() constructor.{}
docA description of the rule that can be extracted by documentation generating tools.""
toolchainsAn optional list to supplement the toolchains passed to the unit test's rule() constructor.[]

RETURNS

A rule definition that should be stored in a global whose name ends in _test.

unittest.suite

Defines a test_suite target that contains multiple tests.

After defining your test rules in a .bzl file, you need to create targets from those rules so that blaze test can execute them. Doing this manually in a BUILD file would consist of listing each test in your load statement and then creating each target one by one. To reduce duplication, we recommend writing a macro in your .bzl file to instantiate all targets, and calling that macro from your BUILD file so you only have to load one symbol.

You can use this function to create the targets and wrap them in a single test_suite target. If a test rule requires no arguments, you can simply list it as an argument. If you wish to supply attributes explicitly, you can do so using partial.make(). For instance, in your .bzl file, you could write:

def your_test_suite():
  unittest.suite(
      "your_test_suite",
      your_test,
      your_other_test,
      partial.make(yet_another_test, timeout = "short"),
  )

Then, in your BUILD file, simply load the macro and invoke it to have all of the targets created:

load("//path/to/your/package:tests.bzl", "your_test_suite")
your_test_suite()

If you pass N unit test rules to unittest.suite, N + 1 targets will be created: a test_suite target named ${name} (where ${name} is the name argument passed in here) and targets named ${name}_test_${i}, where ${i} is the index of the test in the test_rules list, which is used to uniquely name each target.

PARAMETERS

NameDescriptionDefault Value
nameThe name of the test_suite target, and the prefix of all the test target names.none
test_rulesA list of test rules defines by unittest.test.none