Testing Nautobot¶

Best practices for developing and maintaining Nautobot's automated test suites.

Unit tests are automated tests written and run to ensure that a section of the Nautobot application (known as the "unit") meets its design and behaves as intended and expected. Most commonly as a developer of or contributor to Nautobot you will be writing unit tests to exercise the code you have written. Unit tests are not meant to test how the application behaves, only the individual blocks of code, therefore use of mock data and phony connections is common in unit test code. As a guiding principle, unit tests should be fast, because they will be executed quite often.

Integration tests are automated tests written and run to ensure that the Nautobot application behaves as expected when being used as it would be in practice. By contrast to unit tests, where individual units of code are being tested, integration tests rely upon the server code actually running, and web UI clients or API clients to make real connections to the service to exercise actual workflows, such as navigating to the login page, filling out the username/passwords fields, and clicking the "Log In" button.

Integration testing is much more involved, and builds on top of the foundation laid by unit testing. As a guiding principle, integration tests should be comprehensive, because they are the last mile to asserting that Nautobot does what it is advertised to do. Without integration testing, we have to do it all manually, and that's no fun for anyone!

Migration tests are automated tests written and run to ensure that Nautobot database migrations (primarily "data" migrations in particular, as opposed to "schema" migrations) correctly handle various data scenarios. These are "before/after" tests that bring the database to a particular historical state, populate data into this database state, then run the database migrations (simulating an in-place update of Nautobot) and verify that the final state of the database is as expected.

Tagging Tests¶

By Nautobot convention, unit tests must be tagged with unit. The base test case class nautobot.core.testing.TestCase has this tag, therefore any test cases inheriting from that class do not need to be explicitly tagged. All existing view and API test cases in the Nautobot test suite inherit from this class.

By Nautobot convention, integration tests must be tagged with integration. The base test case class nautobot.core.testing.integration.SeleniumTestCase has this tag, therefore any test cases inheriting from that class do not need to be explicitly tagged. All existing integration test cases in the Nautobot test suite inherit from this class.

Changed in version 2.0.0

The base test classes moved from nautobot.utilities.testing to nautobot.core.testing.

Nautobot's Python-based migration tests are built around the django-test-migrations library and its MigratorTestCase class, which has the tag migration_test, therefore any test cases inheriting from that class do not need to be explicitly tagged.

Info

There are also a set of "holistic" migration tests executable by the invoke migration-test command; these "tests" consist of nothing more than historical SQL dumps (in the development/datasets/ directory) of a fully populated database state that can be used to populate an empty database before running a nautobot-server migrate command and confirming that it raises no errors. This sort of test is fairly crude; in most cases writing a specific MigratorTestCase subclass is to be preferred.

Base Classes and Code Location¶

Test Type	Base Class	Code Location	Test Execution
Unit	`nautobot.core.testing.TestCase` or subclass (see below)	`nautobot/APP/tests/test_*.py`	`invoke tests`
Integration	`nautobot.core.testing.integration.SeleniumTestCase`	`nautobot/APP/tests/integration/test_*.py`	`invoke tests --tag integration`
Migration	`django_test_migrations.contrib.unittest_case.MigratorTestCase`	`nautobot/APP/tests/migration/test_*.py`	`invoke tests --tag migration_test`

New unit tests must always inherit from nautobot.core.testing.TestCase or one of its subclasses. Do not use django.test.TestCase or unittest.TestCase.
- API view test cases should generally inherit from one or more of the classes in nautobot.core.testing.api.APIViewTestCases.
- Filterset test cases should generally inherit from nautobot.core.testing.filters.FilterTestCases.FilterTestCase.
- Form test cases should generally inherit from nautobot.core.testing.forms.FormTestCases.BaseFormTestCase.
- Model test cases should generally inherit from nautobot.core.testing.models.ModelTestCases.BaseModelTestCase.
- View test cases should generally inherit from one or more of the classes in nautobot.core.testing.views.ViewTestCases.
New integration tests must always inherit from nautobot.core.testing.integration.SeleniumTestCase. Do not use any other base class for integration tests.

Changed in version 2.0.0

The base test classes moved from nautobot.utilities.testing to nautobot.core.testing.

New migration tests should generally inherit from django_test_migrations.contrib.unittest_case.MigratorTestCase; if any other base class is used, you must explicitly tag it with the migration_test tag.

Generic Filter Tests¶

Added in version 2.0.0

Nautobot provides a set of generic tests for testing the behavior of FilterSets. These tests are located in nautobot.core.testing.filters.FilterTestCase and can be used to test some common filters in Nautobot.

Generic Boolean Filter Tests¶

When using FilterTestCase, all filters that are instances of nautobot.core.filters.RelatedMembershipBooleanFilter that are not using a custom filter method will be tested to verify that the filter returns the same results as the model's queryset. RelatedMembershipBooleanFilter filters will be tested for both True and False values.

Generic Multiple Choice Filter Tests¶

A generic_filter_tests attribute with a list of filters can be defined on the test class to run generic tests against multiple choice filters. The generic_filter_tests attribute should be in the following format:

generic_filter_tests = (
    # use a single item when the filter name matches the model field name
    ["model_field"],
    # use [filter_name, field_name] when the filter name does not match the model field name
    ["related_object_filter", "related_object__name"],
    # the field name is passed as a kwarg to the `queryset.filter` method, so the dunder syntax can be used to make nested queries
    ["related_object_filter", "related_object__id"],
)

Tags Filter Test¶

If the model being tested is a PrimaryModel, the tags filter will be automatically tested by passing at least two values to the filter and verifying that the result matches the equivalent queryset filter.

Integration Tests¶

Troubleshooting Integration Tests¶

Because integration tests normally involve interacting with Nautobot through a browser via Selenium and the Splinter wrapper library, they can be difficult to troubleshoot directly from the Python code when a failure occurs. A common troubleshooting technique is to add a breakpoint() at the appropriate place in the Python test code (i.e., immediately prior to the observed failure). When the breakpoint is hit and the test pauses, you can then use a VNC viewer application (such as macOS's "Screen Sharing" app) to connect to the running Selenium instance (localhost:15900 if using the Docker development environment; the default password if prompted is simply "secret"). This will allow you to interact live with the testing web browser in its current state and can often provide invaluable insight into the nature of any test failure.

Factories¶

Nautobot uses the factory_boy library as a way to generate randomized but plausible database data for use in unit and integration tests, or for convenience in populating a local development instance.

Factories for each Nautobot app's models are defined in the corresponding nautobot/APPNAME/factory.py files. Helper classes and functions for certain common patterns are defined in nautobot/core/factory.py. Factories can be used directly from nautobot-server nbshell so long as you have factory_boy installed. Examples:

>>> from nautobot.tenancy.factory import TenantFactory, TenantGroupFactory
>>> # Create a single TenantGroup instance
>>> TenantGroupFactory.create()
<TenantGroup: Peterson, Nunez and Miller>
>>> # Create 5 Tenant instances
>>> TenantFactory.create_batch(5)
[<Tenant: Smith-Vance>, <Tenant: Sanchez, Brown and Davis>, <Tenant: Benson and Sons>, <Tenant: Pennington PLC>, <Tenant: Perez and Sons>]
>>> # Create 5 more Tenant instances all with a specified "tenant_group" value
>>> TenantFactory.create_batch(5, tenant_group=TenantGroup.objects.first())
[<Tenant: Mercado, Wilson and Fuller>, <Tenant: Blackburn-Andrade>, <Tenant: Oliver-Ramirez>, <Tenant: Pugh-Clay>, <Tenant: Norman and Sons>]

Warning

factory_boy is only a development dependency of Nautobot. You cannot use the model factories in a production deployment of Nautobot unless you directly pip install factory_boy into such a deployment.

Nautobot's custom test runner class (nautobot.core.tests.runner.NautobotTestRunner) makes use of the various factories to pre-populate the test database with data before running any tests. This reduces the need for individual tests to define their own baseline data sets.

Info

Because Apps also commonly use Nautobot's test runner, the base Nautobot settings.py currently defaults TEST_USE_FACTORIES to False so as to not negatively impact App tests that may not be designed to account for the presence of pre-populated test data in the database. This configuration is overridden to True in nautobot/core/tests/nautobot_config.py for Nautobot's own tests.

Warning

Factories should generally not be called within test code, i.e. in a setUp() or setUpTestData() method. This is because factory output is stateful, that is to say the output of any given factory call will depend on the history of all previous factory calls since the process was started. This means that a call to a factory within a test case will depend on which other test cases have also called factories, and what order they were called in, as well as whether the initial test database population was done via factories or whether they were bypassed by reuse of cached test data (see below).

In short, we should only have one place in our tests where factories are called, and that's the generate_test_data management command. Individual tests should use standard create() or save() model methods, never factories.

Factory Caching¶

To reduce the time taken between multiple test runs, a new argument has been added to the nautobot-server test command: --cache-test-fixtures. When running tests with --cache-test-fixtures for the first time, after the factory data has been generated it will be saved to a factory_dump.json file in the development directory. On subsequent runs of unit or integration tests, if --cache-test-fixtures is again specified (hint: it is included by default when running invoke tests), the factory data will be loaded from the file instead of being generated again. This can significantly reduce the time taken to run tests.

Changed in version 2.2.7 — Hashing of migrations in the factory dump

The test runner now calculates a hash of applied database migrations and uses that as a key when creating/locating the factory data file. This serves as a way to avoid inadvertently using cached test data from the wrong branch or wrong set of migrations, and reduces the frequency with which you might need to manually delete the fixture file. For example, the set of migrations present in develop might result in a factory_dump.966e2e1ed4ae5f924d54.json, while those in next might result in factory_dump.72b71317c5f5c047493e.json - both files can coexist, and when you switch between branches during development, the correct one will automatically be selected.

Changed in version 2.3.4 — Factory caching is enabled by default in invoke tasks

Factory caching is now enabled by default with the invoke tests command. To bypass it, either use the --no-cache-test-fixtures argument to invoke tests, or manually remove the development/factory_dump.*.json cache file(s).

Tip

Although changes to the set of migrations defined will automatically invalidate an existing factory dump, there are two other cases where you will currently need to manually remove the file in order to force regeneration of the factory data:

When the contents of an existing migration file are modified (the hashing implementation currently can't detect this change).
When the definition of a factory is changed or a new factory is added.

Test Code Style¶

Use more specific/feature-rich test assertion methods where available (e.g. self.assertInHTML(fragment, html) rather than self.assertTrue(re.search(fragment, html)) or assert re.search(fragment, html) is not None).
Keep test case scope (especially in unit tests) small. Split test functions into smaller tests where possible; otherwise, use self.subTest() to delineate test blocks as appropriate.