testing.md

Testing

Setup

make direnv    # install direnv and its shell hook
direnv allow   # load the extension-selected workspace and compute
make dev       # create the Python 3.12 development environment
make test      # build, install, and test the wheel

Environment flow: VS Code extension → .databricks/.databricks.env → direnv → Make → pytest. Databricks Connect and the SDK therefore use the workspace and compute selected in the extension. See Developer setup.

What make test does

1. Build and install the wheel with its dependencies.
2. Run all files in tests/.
3. Write .junittest.xml, coverage/html/, and coverage/xml/xml.xml.

Important options:

• -n8: pytest-xdist runs eight workers.
• --exitfirst -vv: stop on failure and show every test.
• --cov=dbrdemo: pytest-cov reports coverage without enforcing a threshold.

Databricks test libraries

• Databricks Labs pytester provides reusable pytest helpers such as make_random.
• Databricks Labs Blueprint provides the formatted logger enabled by tests/conftest.py.

Test design

• Local tests cover docs, CLIs, and skill installation.
• Connected tests cover Spark, the SDK, Unity Catalog tables, volumes, and files.
• temporary_schema creates a random schema per test and always removes it.
• Random names keep parallel xdist workers isolated.

The test identity needs permission to create schemas, tables, and volumes in the current catalog.

Debug one test

.venv/bin/pytest -n0 -vv tests/foobar_test.py::test_write_foobar