ENH : add units parameter to read_raw_edf in case units is missing from the file

[agramfort]

related to mne-tools/mne-bids#1045 and mne-tools/mne-bids#1044

now with the data from ds003775 you can just do:

raw = mne.io.read_raw_edf(bp, units='uV')

and in mne-bids

raw = read_raw_bids(bp, extra_params={'units': 'uV'})

what do you think @sappelhoff @hoechenberger ?

[hoechenberger]

@agramfort Looks good, but I'd call the parameter unit (singular), since it's just one unit (i.e., the same one) for all channels in the most common scenario (user passes a string). WDYT?

An alternative approach could also be to introduce a parameter scaling instead? And accept an an int or float by which we'll multiply the input data? I feel this would fit better into the existing MNE-Python API, but it's slightly less convenient for users (?)

What bugs me a little bit about calling the parameter unit(s) is that we don't actually perform a unit conversion, but simply adjust the scaling factor. Hence the idea of having scaling instead.

[sappelhoff]

LGTM!

I like "units" (plural), because one can also pass a dict of units (one for each channel).
Regarding "scaling(s)" --> true, we are not doing unit conversions ... but won't get_scalings warn/error when a user passes a unit that does not correspond to the unit the channel type is in?

Still, I'd be fine with bot units and scalings.

[hoechenberger]

I like "units" (plural), because one can also pass a dict of units (one for each channel).

But I believe we usually follow the convention that variable names for dictionaries are singular, while names for lists are plural, no?

I'm not vetoing "units", just thinking out loud.

[agramfort]

@hoechenberger do a git grep in the repo with units. You will see that we call units many parameters that are dict of units for the different channels. I think this naming is consistent with mne

[hoechenberger]

Thanks @agramfort!

[hoechenberger]

Oh, I merged without requesting a test.

@agramfort Care to add a small test case in a followup PR?