streamlined approach to creating smoke tests by executing noteboks "as is" from the tests and asserting on values of local notebook vars

[slayoo]

@claresinger @edejong-caltech @abulenok @AgnieszkaMakulska

We've introduced a new utility that will likely make it easier to accompany notebooks with smoke tests. It's a function called notebook_vars that executes a given notebook and returns a dictionary mapping notebook variables to their values at the end of notebook execution: https://open-atmos.github.io/PySDM/PySDM_examples/utils/notebook_vars.html

Here's an new test using this approach just contributed by @bhiogade (https://github.com/open-atmos/PySDM/blob/main/tests/smoke_tests/parcel/graf_et_al_2019/test_fig_4.py):

from pathlib import Path

import pytest
from PySDM_examples import Graf_2019
from PySDM_examples.utils import notebook_vars

from PySDM.physics import si

PLOT = False


@pytest.fixture(scope="session", name="variables")
def variables_fixture():
    return notebook_vars(
        file=Path(Graf_2019.__file__).parent / "figure_4.ipynb", plot=PLOT
    )


@pytest.mark.parametrize(
    "level_name, expected_height", (("CB", 1 * si.km), ("0C", 2.24 * si.km))
)
def test_fig_4(variables, level_name, expected_height):
    # arrange
    tolerance = 50 * si.m

    # act
    actual = variables["levels"][level_name] + variables["alt_initial"]

    # assert
    assert abs(expected_height - actual) < tolerance

As exemplified above, the new tool automagically handles the plot flag, so debugging the failing tests one can easily turn plotting on, getting all the notebook plots, without ever starting Jupyter. The relevant notebook looks like this: