Writing tests

To help keep Bokeh maintainable, all Pull Requests that add or update code should include new or updated tests. While exceptions are possible, a Pull Request without adequate tests will generally not be considered ready to merge.

Follow those general guidelines to decide which kinds of tests to update or add:

When you edit Bokeh’s Python code

Check whether you should update or add Python tests

When your change anything related to BokehJS

Check whether you should update or add JavaScript tests

When you fix a bug from Bokeh’s issue tracker:

Check whether you should add a regression test

Before writing any tests, you should make sure to go through the relevant passages in Running tests. This chapter on writing tests assumes you know how to run the tests you are working on.

Writing Python tests

If all or parts of your changes affect Bokeh’s Python code, you should add or update the relevant unit and integration tests.

These tests are located in the tests folder. See Run Python tests for information on how to run them.

General information on contributing to Bokeh’s Python code and models is available in Contributing Python code.

Python unit tests

Python unit tests help maintain the basic functionality of the Python portion of Bokeh. They are located in tests/unit/bokeh. The folder structure resembles the structure of Bokeh’s Python models. The name of each test file begins with test_, followed by the module’s name.

Follow these general guidelines when writing Python unit tests:

Import the model under test with as

Always import the specific model you are testing using the import as syntax. Use the model’s initials to name your import. For example:

import bokeh.plotting.graph as bpg
Use absolute imports

Bokeh’s unit tests should be as relocatable and unambiguous as possible. Therefore, you should use absolute imports (from bokeh.embed import components) in test files whenever possible. Don’t use relative imports (from ..document import Document).

Use pytest (not unittest)

All new tests should use and assume pytest for all testing-related aspects, such as test running, fixtures, or parameterized testing. Please do not use the unittest module of the Python standard library.

Python integration tests

Bokeh’s Python-focused integration tests help make sure that Bokeh’s Python code works as intended with the TypeScript code of BokehJS.

The Python integration tests use Selenium with ChromeDriver. The test scripts are located in tests/integration. The folder structure resembles the structure of Bokeh’s Python models.

Python integration tests use pytest fixtures to handle the web driver configuration and interaction with Selenium. Depending on which context you want to test an object in, choose from bokeh_model_page, single_plot_page, or bokeh_server_page. See bokeh/_testing/plugins/project.py for more details.

Follow these guidelines when adding or updating Python integration tests:

Keep your code as simple as possible

Try to only include things that are essential to your test. Focus your test on one specific functionality. If possible, write several small tests instead of one complex one.

Use the bokeh.models API whenever possible

Try to use Bokeh’s low-level bokeh.models interface instead of the more high-level bokeh.plotting interface.

Writing JavaScript tests (BokehJS)

To maintain the functionality of all BokehJS components, Bokeh includes various tests written in TypeScript. If all or parts of your changes affect the JavaScript code of BokehJS, you should add or update the relevant BokehJS tests.

Bokeh’s JavaScript tests use a custom testing framework that requires Google Chrome or Chromium. You need a recent version of one of these browsers available on your system to work with these tests.

Like several other testing frameworks such as Mocha or Jasmine, the BokehJS testing framework uses describe() and it() functions to set up tests. Assertions are made with expect() and assert(). Use assert() when you need to narrow types.

The BokehJS tests are located in bokehjs/test. See Run JavaScript tests for information on how to run them.

General information on contributing to BokehJS is available in Contributing to BokehJS.

BokehJS unit tests

The BokehJS unit tests help make sure that the individual sections of BokehJS function as expected. The unit tests for BokehJS are located in the bokehjs/test/unit/ folder and sub-folders.

The basic structure for writing tests for Bokeh’s testing framework is in parts inspired by the Chai “expect” assertion style. See the API documentation of the Chai Assertion Library for some general ideas.

Use expect() together with the following elements to create assertions for the BokehJS testing framework:

  • to and be: tokens to improve readability of assertions and connect elements

  • not: negates the following assertions

  • throw: asserts that an error is thrown. Accepts the following optional parameters: error_type (filter by Error) and pattern (filter by regular expression or string).

  • equal: asserts (deep) value equality. Expects an operand to compare to.

  • similar: asserts similarity within a defined tolerance, based on the same value equality as equal. Expects an operand to compare to as well as an optional number as tolerance.

  • identical: asserts strict equality (===). Expects an operand to compare to.

  • instanceof: asserts that the tested element is an instance of the given constructor. Expects a Constructor to test against.

  • undefined: asserts strict equality (===) to undefined

  • null: asserts strict equality (===) to null

  • true: asserts strict equality (===) to true

  • false: asserts strict equality (===) to false

  • NaN: asserts that the tested element is NaN

  • empty: asserts a length of 0 (for example, an empty string or an iterable that does not contain any retrievable values)

  • below: asserts that the tested element is below (<) a value. Expects a number to compare to.

  • above: asserts that the tested element is below (>) a value. Expects a number to compare to.

For example:

expect(m.name).to.be.null
expect(grid0).to.be.instanceof(Column)
expect(h.msgid).to.not.be.equal(h2.msgid)

BokehJS visual tests

BokehJS uses visual regression tests as integration tests. These baseline comparison tests help make sure that Bokeh’s visual output is consistent with the output expected by design. Any BokehJS-related pull requests that result in changes to the visual output generated by BokehJS should include visual baseline comparison tests.

In the background, the BokehJS testing framework runs a headless browser and takes screenshots of the browser’s output. The testing framework then compares the visual output to each test’s individual baseline files.

Each test in test:integration consists of two types of baseline comparisons:

Textual baseline comparison

For each test, the testing framework compares the pixel location of certain elements in the visual output to pixel locations in the baseline data. This baseline data is stored as plain text in each test’s respective .blf file.

Visual baseline comparison

For each test, the testing framework does a pixel-by-pixel comparison of a screenshot and a baseline image. These baseline images are stored as .png files. In contrast to textual baseline comparisons, visual baseline comparisons are platform-dependent. Even minor differences in font rendering, for example, will make the pixel-by-pixel comparison fail.

The visual baseline comparison tests are located in the bokehjs/test/integration/ folder and sub-folders. Bokeh’s CI runs these tests on Linux, macOS, and Windows environments. The baseline files for each environment are located in the bokehjs/test/baselines/ folder.

Follow these steps to write new visual tests or update existing tests:

  1. Create or update visual testing scripts:

    To write a visual test for the BokehJS testing framework, start by importing the display() and fig() functions from the testing framework’s _util module (located in bokehjs/test/integration/):

    import {display, fig} from "./_util"
    

    When writing tests, replace the standard BokehJS show() function with the display() function in _util. The display() function accepts the same arguments as show() but also captures the visual output for comparison.

    Similarly, replace the standard BokehJS figure() function with the fig() function in _util. The fig() function expects an array of [width, height] as the first argument, followed by the same arguments as figure(). However, to keep visual tests as efficient as possible, you should only use width and height wherever you can.

    Keep the width and height of your testing plot as small as possible while still being able to see the details you want to test with the naked eye. Try to keep the number of elements on your plot to a minimum.

    Follow this general pattern for visual tests:

    describe("Your Object", () => {
      it("should show certain behavior", async () => {
        const p = fig([width, height], {figure_attrs})
    
        ...
    
        await display(p)
      })
    })
    

    To change the sensitivity of a visual test, you have the option to set a threshold value. The threshold value represents the amounts of pixels by which a test image can differ from the baseline image before a test fails. To set a threshold value, use it.allowing(threshold). For example:

    describe("Your Object", () => {
      it.allowing(16)("should show certain behavior", async () => {
    

    Always run node make lint before committing TypeScript files.

  2. Run tests locally:

    Run node make tests to test your changes on your system. To only run integration tests, use node make test:integration.

    If you want to run a specific test only, use the -k argument and supply a search string. The search string is case-sensitive. The BokehJS testing framework tries to match your search string to the strings defined in the code’s describe() and it() functions. For example:

    $ node make test:integration -k 'Legend annotation'
    

    The first time you run a new or updated visual test, the BokehJS testing framework will notify you that baseline files are missing or outdated. At this point, it will also generate all missing or outdated baseline files for your operating system. The baseline files will be in a subfolder of bokehjs/test/baselines/.

    Use the BokehJS devtools server to review your local test results. Optionally, you can use any PNG viewer to inspect the generated PNG files. Adjust your testing code until the test’s visual output matches your expectations.

  3. Generate CI baselines and commit test:

    As a final step before pushing your visual tests to Bokeh’s GitHub repository, you need to generate and commit the baseline files using Bokeh’s CI.

    The baseline files are platform-dependent. This is why the CI will only work reliably if you upload baseline files that were created by the CI, not locally created files.

    Before generating new baseline images with Bokeh’s CI, rebase your branch to make sure all tests are up to date.

    Follow these steps to generate the necessary baseline files and upload them to Bokeh’s CI:

    1. Push your changes to GitHub and wait for CI to finish.

    2. The CI will expectedly fail because baseline images are either missing (in case you created new tests) or outdated (in case you updated existing tests).

    3. After the CI has finished running, go to Bokeh’s GitHubCI page. Find the most recent test run for your PR and download the associated bokehjs-report artifact.

    4. Unzip the downloaded artifact file into the root folder of your local Bokeh repository.

    5. Use the devtools server to review the baseline files the CI has created for each platform: first, go to /integration/report?platform=linux, then to /integration/report?platform=macos, and finally to /integration/report?platform=windows.

    6. If you did not detect any unintentional differences, commit all new or modified *.blf and *.png files from the folders bokehjs/test/baselines/linux, bokehjs/test/baselines/macos, and bokehjs/test/baselines/windows.

    7. Push your changes to GitHub again and verify that the tests pass this time.

Note

Make sure to only push baseline files to the CI that the CI created for your specific pull request. Do not include any locally created baseline files in your pull request.

After downloading and unpacking the baseline files from the CI, check your local bokehjs/test/baselines directory for any modified files that are not part of your changes. Make sure only to commit baseline files that are necessary for your pull request. Reset the baselines directory after every failed test run with git clean or git clean -f.

If you encounter any problems with the steps described here, feel free to get in touch at the Bokeh Discourse or Bokeh’s contributor Slack.

BokehJS regression tests

Additionally, BokehJS uses regression tests with its unit and integration tests. Regression tests are located in bokehjs/test/unit/regressions.ts and bokehjs/test/integration/regressions.ts.

You should add a regression test whenever you fix a bug related to BokehJS. Write your regression test so that it fails in case the bug you fixed occurs again.

Add your testing function to the outermost describe() function that has "Bug" passed to it as its description argument. Add the bug’s issue number to your test’s describe() function and provide a short description of the fixed bug in your it() function.

For example:

describe("in issue #9522", () => {
  it("disallows arrow to be positioned correctly in stacked layouts", async () => {

    ...

    await display(row([p1, p2]))
  })
})

Working with examples tests

Bokeh’s example tests are based on examples found in the examples and sphinx/source/docs/user_guide/examples folders.

When you add a new example to one of these folders, they are usually included in the examples tests automatically. Edit tests/examples.yaml to explicitly include or exclude specific examples.

See Run examples tests for more information on running the examples tests.