Create hypothesis.extra.xarray

[raybellwaves]

I use hypothesis.extra.pandas for creation of pandas.DataFrame. e.g. https://hypothesis.readthedocs.io/en/latest/numpy.html#hypothesis.extra.pandas.data_frames

from hypothesis.extra.pandas import column, data_frames
df = data_frames([column("A", dtype=int), column("B", dtype=float)]).example()

I would be great to have something similar for xarray. I'm not sure what the API would look like but here is a mock up

from hypothesis.extra.xarray import data_array, data_set
ds = data_set(
    [
        data_array("A", shape=(2, 2), dtype=int),
        data_array("B", shape=(2, 2), dtype=float),
    ]
).example()

which could wrap the numpy functions

import xarray as xr
from hypothesis.extra.numpy import arrays
xr.Dataset(
    {
        "A": xr.DataArray(arrays(dtype=int, shape=(2, 2)).example()),
        "B": xr.DataArray(arrays(dtype=float, shape=(2, 2)).example()),
    }
)

[Zac-HD]

Hi, Ray - I'm a keen Xarray user myself, and have some good news: as of a few weeks ago, Xarray ships strategies themselves! The stable set is basically just Variables at the moment, but people are actively working on the rest of the key types and I'm optimistic about shipping something awesome in the coming weeks.

I'm also pretty excited about the new pattern of shipping the integrations on the upstream-library side rather than in Hypothesis itself - it's a great way to encourage more people to try property-based testing, and ensures that the maintainers with deep expertise in e.g. Xarray are the ones maintaining the integration. (by contrast, hypothesis.extra.django rather suffers from there being no Django users among the Hypothesis maintainers)