Skip to content

Latest commit

 

History

History
37 lines (35 loc) · 2.51 KB

CONVERTING.md

File metadata and controls

37 lines (35 loc) · 2.51 KB

Notes on converting tests from unittest to pytest

  • Base fixtures are provided in the core/dbt/tests/fixtures/project.py
  • You can override any fixture by specifying it in your test script
  • You can't use any test utility from the legacy tests that starts with 'self'.
  • Most special case assert functions will work with a simple 'assert'
  • Every integration test needs to use the 'project' fixture
  • Files are not copied (unless done explicitly in the test) so if you need to access a local file (like seed.sql) you need to get the path from the project fixture, (project.test_data_dir) or the 'test_data_dir' fixture (for specifying the location of data files.)
  • Table comparison methods have been moved to TableComparison in test/tables.py
  • Fixtures are for test setup, and are specified in the test signature. You can't call fixtures in the middle of a test function.
  • Information from the fixture setup that might be needed later in the test is provided by the project fixture return class (TestProjInfo)
  • Every fixture has a scope, which means that you can call it multiple times in the same scope and it will return the same thing. Default scope is 'function', you can also scope fixtures to 'module' and 'session'.
  • All fixtures are run before and after the test function they're attached to. If you have teardown pieces in the fixture, do a 'yield' after the setup, and the part after the 'yield' will be run at teardown.
  • 'run_dbt', 'run_sql', and 'get_manifest' are provided by the core/dbt/tests/util.py file
  • You will probably want to make separate test files out of tests that use substantially different projects. If they're only different by a file or two, you could write out individual files instead and keep them in the same file.
  • You can also import file strings from other test cases
  • old: self.get_models_in_schema, new: get_tables_in_schema
  • somewhat easier way to get the legacy files into a more usable format: tail -n +1 models/* > copy_models.out
  • some of the legacy tests used a 'default_project' method to change (for example) the seeds directory to load a different seed. Don't do that. Copying a file is probably a better option.
  • If there are more than 50 lines of fixture strings, they should be defined in a fixtures.py and then imported. We definitely don't do this everywhere right now but should move to this model.

Integration test directories that have been converted

  • 001_simple_copy_tests => moved to 'basic'
  • 003_simple_reference_tests => moved to 'basic'