Register for the online Blockly Developer Summit April 28-29, 2021!

Unit Tests

After changing or adding code, you should run existing unit tests and consider writing more. All tests are executed on the uncompressed versions of the code.

There are two sets of unit tests: JS tests and block generator tests.

JS Tests

The JS tests confirm the operation of internal JavaScript functions in Blockly's core. We use Mocha to run unit tests, Sinon to stub dependencies, and Chai to make assertions about the code.

Running Tests

In both blockly and blockly-samples, npm run test will run the unit tests. In blockly, this will also run other tests such as linting and compilation. You can also open tests/mocha/index.html in a browser to interactively run all mocha tests.

Writing Tests

We use the Mocha TDD interface to run tests. Tests are organized into suites, which can contain both additional sub-suites and/or tests. Generally, each component of Blockly (such as toolbox or workspace) has its own test file which contains one or more suites. Each suite can have a setup and teardown method that will be called before and after, respectively, each test in that suite.

Test Helpers

We have a number of helper functions specific to Blockly that may be useful when writing tests. These can be found in core and in blockly-samples.

The helper functions include sharedTestSetup and sharedTestTeardown which are required to be called before and after your tests (see Requirements section).

  • Sets up sinon fake timers (in some tests you will need to use this.clock.runAll).
  • Stubs to fire immediately (configurable).
  • Sets up automatic cleanup of blockTypes defined though defineBlocksWithJsonArray.
  • Declares a few properties on the this context that are meant to be accessible:
    • this.clock (but should not be restored else it will cause issues in sharedTestTeardown)
    • this.eventsFireStub
    • this.sharedCleanup (to be used with addMessageToCleanup and addBlockTypeToCleanup) (NOTE: you don't need to use addBlockTypeToCleanup if you defined the block using defineBlocksWithJsonArray)

The function has one optional options parameter to configure setup. Currently, it's only used to determine whether to stub to fire immediately (will stub by default).

  • Disposes of workspace this.workspace (depending on where it was defined, see Test Requirements section for more information).
  • Restores all stubs.
  • Cleans up all block types added though defineBlocksWithJsonArray and addBlockTypeToCleanup.
  • Cleans up all messages added though addMessageToCleanup.

Test Requirements

  • Each test must call; as the first line in the setup of the outermost suite and; as the last line in the teardown of the outermost suite for a file.
  • If you need a workspace with a generic toolbox, you can use one of the preset toolboxes on the test index.html. See below for an example.
  • You must properly dispose of this.workspace. In most tests, you will define this.workspace in the outermost suite and use it for all subsequent tests, but in some cases you might define or redefine it in an inner suite (for example, one of your tests requires a workspace with different options than you originally set up). It must be disposed of at the end of the test.
    • If you define this.workspace in the outermost suite and never redefine it, no further action is needed. It will be automatically disposed of by sharedTestTeardown.
    • If you define this.workspace for the first time in an inner suite (i.e. you did not define it in the outermost suite), you must manually dispose of it by calling, this.workspace) in the teardown of that suite.
    • If you define this.workpace in the outermost suite, but then redefine it in an inner test suite, you must first call, this.workspace) before redefining it to tear down the original workspace defined in the top level suite. You must also manually dispose the new value by calling, this.workspace) again in the teardown of this inner suite.

Test Structure

Unit tests generally follow a set structure, which can be summarized as arrange, act, assert.

  1. Arrange: Set up the state of the world and any necessary conditions for the behavior under test.
  2. Act: Call the code under test to trigger the behavior being tested.
  3. Assert: Make assertions about the return value or interactions with mocked objects in order to verify correctness.

In a simple test, there may not be any behavior to arrange, and the act and assert stages can be combined by inlining the call to the code under test in the assertion. For more complex cases, your tests will be more readable if you stick to these 3 stages.

Here is an example test file (simplified from the real thing).

suite('Flyout', function() {
  setup(function() {;
    this.toolboxXml = document.getElementById('toolbox-simple');
    this.workspace = Blockly.inject('blocklyDiv',
          toolbox: this.toolboxXml

  teardown(function() {;

  suite('simple flyout', function() {
    setup(function() {
      this.flyout = this.workspace.getFlyout();
    test('y is always 0', function() {
      // Act and assert stages combined for simple test case
      chai.assert.equal(this.flyout.getY(), 0, 'y coordinate in vertical flyout is 0');
    test('x is right of workspace if flyout at right', function() {
      // Arrange
      sinon.stub(this.flyout.targetWorkspace, 'getMetrics').returns({
        viewWidth: 100,
      this.flyout.targetWorkspace.toolboxPosition = Blockly.TOOLBOX_AT_RIGHT;
      this.flyout.toolboxPosition_ = Blockly.TOOLBOX_AT_RIGHT;

      // Act
      var x = this.flyout.getX();

      // Assert
      chai.assert.equal(x, 100, 'x is right of workspace');

Things to note from this example:

  • A suite can contain other suites that have additional setup and teardown methods.
  • Each suite and test has a descriptive name.
  • Chai assertions are used to make assertions about the code.
    • You can supply an optional string argument that will be displayed if the test fails. This makes it easier to debug broken tests.
    • The order of the parameters is chai.assert.equal(actualValue, expectedValue, optionalMessage). If you swap actual and expected, the error messages won't make sense.
  • Sinon is used to stub methods when you don't want to call the real code. In this example, we don't want to call the real metrics function because it isn't relevant to this test. We only care about how the results are used by the method under test. Sinon stubs the getMetrics function to return a canned response which we can easily check for in our test assertions.
  • The setup methods for each suite should contain only generic setup that applies to all tests. If a test for a particular behavior relies on a certain condition, that condition should clearly be stated in the relevant test.

Debugging Tests

  • You can open the tests in a browser and use the developer tools to set breakpoints and investigate if your tests are unexpectedly failing (or unexpectedly passing!).
  • Set .only() or .skip() on a test or suite to run only that set of tests, or skip a test. For example:

    suite.only('Workspace', function () {
      suite('updateToolbox', function () {
        test('test name', function () {
          // ...
        test.skip('test I don’t care about', function () {
          // ...

    Remember to remove these before committing your code.

Block Generator Tests

Each block has its own unit tests. These tests verify that blocks generate code than functions as intended.

  1. Load tests/generators/index.html in Firefox or Safari. Note that Chrome and Opera have security restrictions that prevent loading the tests from the local "file://" system (Issues 41024 and 47416).
  2. Choose the relevant part of the system to test from the drop-down menu, and click "Load". Blocks should appear in the workspace.
  3. Click on "JavaScript".
    Copy and run the generated code in a JavaScript console. If the output ends with "OK", the test has passed.
  4. Click on "Python".
    Copy and run the generated code in a Python interpreter. If the output ends with "OK", the test has passed.
  5. Click on "PHP".
    Copy and run the generated code in a PHP interpreter. If the output ends with "OK", the test has passed.
  6. Click on "Lua".
    Copy and run the generated code in a Lua interpreter. If the output ends with "OK", the test has passed.
  7. Click on "Dart".
    Copy and run the generated code in a Dart interpreter. If the output ends with "OK", the test has passed.

Editing Block Generator Tests

  1. Load tests/generators/index.html in a browser.
  2. Choose the relevant part of the system from the drop-down menu, and click "Load". Blocks should appear in the workspace.
  3. Make any changes or additions to the blocks.
  4. Click on "XML".
  5. Copy the generated XML into the appropriate file in tests/generators/.