Support NamedTuples for named arguments in parametrizations.

[tobiasraabe]

Workarounds

typing.NamedTuple or collections.namedtuple can be used to achieve named arguments per iteration of the parametrization.

from pathlib import Path
from typing import NamedTuple


class Task(NamedTuple):
    depends_on: Path
    produces: Path


@pytask.mark.parametrize("depends_on, produces", [
    Task(depends_on="first_dataset.pkl", produces="first_plot.png"),
    Task(depends_on="second_dataset.pkl", produces="second_plot.png"),
])
def task_plot_data(depends_on, produces):
    df = pd.read_pickle(depends_on)
    ax = df.plot()
    plt.savefig(produces)

• Document example usage in the documentation.
• Implement a check for whether the arguments of the namedtuple match the signature of the parametrization.

Originally posted by @tobiasraabe in #206 (comment)

[tobiasraabe]

This will not be pursued. #229 implements a new approach to parametrizations which will eventually supersede pytest's approach.

We keep this issue open for documentation. Probably until the pytest approach is deprecated.