Unit Test¶

In DeepReg, we use pytest (not unittest) for unit tests to ensure a certain code quality and to facilitate the code maintenance.

For testing the code locally,

If you only need to test with python 3.7, please execute pytest at the repository root:
```
pytest test/
```
If you want to test with all supported python versions, please execute tox at the repository root:
```
tox
```

Moreover, the testing is checked automatically via GitHub workflows and Codecov is used to monitor the test coverage. While checking the Codecov report in file mode, generally a line highlighted by red means it is not covered by test. Please check the Codecov documentation for more details.

Test requirement¶

We would like to achieve 100% test coverage. In general, tests should be

thorough, covering different scenarios.
independent, different scenarios are not tested together.
clean and compact, for instance,
- Use parameterized test to reduce code redundancy.
- Use is_equal_np and is_equal_tf provided in test/unit/util.py to compare arrays or tensors.

The detailed requirements are as follows:

Test all functions in python classes.
Test the trigger of all warning and errors in functions.
Test the correctness of output values for functions.
Test at least the correctness of output shapes for TensorFlow functions.

Example unit test¶

We provide here an example to help understanding the requirements.

import pytest


def subtract(x: int) -> int:
    """
    A function subtracts one from a non-negative integer.
    :param x: a non-negative integer
    :return: x - 1
    """
    assert isinstance(x, int), f"input {x} is not int"
    assert x >= 0, f"input {x} is negative"
    return x - 1


class TestSubtract:
    @pytest.mark.parametrize("x,expected",[(0, -1), (1,0)])
    def test_value(self, x, expected):
        got = subtract(x=x)
        assert got == expected

    @pytest.mark.parametrize("x,msg", [(-1, "is negative"), (0.0, "is not int")])
    def test_err(self, x, msg):
        with pytest.raises(AssertionError) as err_info:
            subtract(x=x)
        assert msg in str(err_info.value)

where

we group multiple test functions for subtract under the same class TestSubtract.
we parameterize test to test different inputs.
we catch errors using pytest.raises and check error messages.

For further usage like fixture and other functionalities, please check pytest documentation or existing tests in DeepReg. You can also raise an issue for any questions.