issue_comments

Thanks @Illviljan

Feedback from meeting: - let's merge - we should point to ds.head().load() if people actually want to see these values. @Illviljan Can you add this to the whats-new note please?

right, this seems to be an old (but unfortunately still unfixed) issue in numpy: numpy/numpy#9464

I have 64bit windows and python appears to be running in 64bit, still I get int32 for some reason:

FWIW, I think the issue is not unix vs windows but 32bit OS vs 64bit OS.

In any case, if we want to be consistent I guess we have to specify the dtype explicitly? So instead of np.arange(300) we'd get np.arange(300, dtype=np.int64).

``` @requires_netCDF4 def test_repr_file_collapsed(tmp_path) -> None: arr_to_store = xr.DataArray(np.arange(300), dims="test") arr_to_store.to_netcdf(tmp_path / "test.nc", engine="netcdf4")

    with xr.open_dataarray(tmp_path / "test.nc") as arr, xr.set_options(
        display_expand_data=False
    ):
        actual = arr.__repr__()
        expected = dedent(
            """\
        <xarray.DataArray (test: 300)>
        [300 values with dtype=int32]
        Dimensions without coordinates: test"""
        )

      assert actual == expected

E AssertionError: assert '<xarray.Data...dinates: test' == '<xarray.Data...dinates: test' E Skipping 47 identical leading characters in diff, use -v to show E - dtype=int32] E ? ^^ E + dtype=int64] E ? ^^ E Dimensions without coordinates: test ``` Once again differences with windows and mac/linux... I suppose this wasn't tested previously.