Fix highpass-spatial-filter return dtype

[alejoe91]

Fixes #2311

@JoeZiminski @h-mayorquin

[JoeZiminski]

This looks great, I'll quickly test it now

[JoeZiminski]

Hey Alessio I couldn't push to your fork, how are such issues usually managed?

The test I added the below test in test_highpass_spatial_file.py. I had to add the properties so the data could be returned scaled, but I think maybe there is a better way. It would be cool to generalise this test to all preprocessing steps, but my guess is this would be tricky!

@pytest.mark.parametrize("dtype", [np.int16, np.float32, np.float64])
def test_dtype_stability(dtype):
    """
    Check that the dtype of the recording and
    output data is as expected, as data is cast to float32
    during filtering.
    """
    num_chan = 32
    si_recording = generate_recording(num_channels=num_chan, durations=[2])
    si_recording.set_property("gain_to_uV", np.ones(num_chan))
    si_recording.set_property("offset_to_uV", np.ones(num_chan))
    si_recording = spre.astype(si_recording, dtype)

    assert si_recording.dtype == dtype

    highpass_spatial_filter = spre.highpass_spatial_filter(si_recording, n_channel_pad=2)

    assert highpass_spatial_filter.dtype == dtype

    filtered_data_unscaled = highpass_spatial_filter.get_traces(return_scaled=False)

    assert filtered_data_unscaled.dtype == dtype

    filtered_data_scaled = highpass_spatial_filter.get_traces(return_scaled=True)

    assert filtered_data_scaled.dtype == np.float32

[alejoe91]

I think that only maintainers can push! I'll add the suggested test :)

[JoeZiminski]

Great thanks, LGTM!

[h-mayorquin]

Looks good to me. Let's open an issue for the get_traces when the dtype is already float.