TESTING.md

The testing suite is executed using pytest with all the tests inside the tests/ folder

A few notes on running and writing tests

Requirements

The test suite makes use of packages not included in requirements.txt. To install these additional packages, use

pip install -r requirements-dev.txt

Slow vs fast tests

The majority of tests are designed to take together ~ 30 min of wall time, but a few tests take a long time. To exclude the slow tests you can run the pytest as

pytest -m 'not slow'

Long tests include notebook tests and some high-dimensional validation tests

Testing in parallel

If you have many CPUs, you can run the tests in parallel using

pytest  -m 'not slow' --dist loadgroup  -n 10

(if you have 10 CPUs). This way the test suite should finish in a few minutes.

Random seeds

By default the testing suite uses the same random seed for all the tests to ensure the consistency. It is occasionally useful to run with different seeds. For this you can set the environment variable DYNESTY_RANDOM_SEED to any integer value. If you are writing a new test you need to use the the get_rstate() function (from tests/utils.py)

Test printing

By default most of the tests will not print any output. You can enable the printing of progress by setting the environment variable DYNESTY_TEST_PRINTING=1 and use -s flag of pytest

Jupyter notebooks tests

To test that the Jupyter notebooks run, you can run the tests/test_notebooks.py test. These tests rely on dynesty being installed in the Python environment referred to by the default python3 Jupyter kernel.