test_algorithm module

test_my_algorithm(notebook, x)

Run numerical experiments of my_implementation against some input.

my_implementation represents a numerical method we are developing. For the sake of example, it simply computes the square of the input.

Here is what’s happening line by line:

@pytest.mark.experiment

[optional] mark this test as an experiment. This does nothing on its own, but will make it easy to separate our experiments from other tests.

@pytest.mark.parametrize(...

Generate some input. pytest-experiments automatically records the input to the experiment function so it’s best practice to use pytest fixtures and parameters to set up the input data. The experiment will run once for each value we specify.

def test_my_algorithm(notebook, x):

Define the experiment. Since we will use pytest to execute our experiments, we must prefix the experiment name with test_. We also will need to bring in the notebook fixture, which does all of the heavy lifting behind the scenes. Lastly, x is the parameter we defined in the previous line.

y = my_implementation(x)

We call our implementation on the input and store the result in a local variable y.

assert y >= 0

We know that regardless of what x is, y should never be negative. If it is, our code is wrong and our experiment should fail. It is important to have some basic validity checks in our experiments.

notebook.record(result=y)

Tell the notebook to record the value of y under the key “y”. Note that the values must be JSON-serializable. See the documentation for details but as a rule of thumb, this includes most python built-in datatypes.