MRG, BUG: Fix EDF header parsing

[larsoner]

1. Implements a variant of @a-hurst's code from EDF/EDF+ headers parsed incompletely/incorrectly #8544 (most of the PR)
2. Fixes dipy decorator (1 line)
3. Fixes str-as-int parsing from gitter (1 line)
4. Adds stub test with XXX for carrying over edf_info['subject_info'] to raw.info['subject_info'], which we should probably do to the extent possible (but in another PR)

@a-hurst the one thing I didn't follow was (what is keeping this WIP from my end):

My only other change was adding recording_info=meas_id, to the edf_info.update() call further down so the recording info gets carried over to the final Raw object.

Can you take a look, and clarify what needs to change in / be added to this PR? And let me know if you're happy with the GitHub URL I put for your name link?

@sappelhoff @cbrnr can you look to see if this makes sense in light of discussions in #8544?

Closes #8544
Closes #8551

[a-hurst]

Looks good, thanks for this! The part you're asking about refers to this section:

        # Populate edf_info
        edf_info.update(
            ch_names=ch_names, data_offset=header_nbytes,
            digital_max=digital_max, digital_min=digital_min,
            highpass=highpass, sel=sel, lowpass=lowpass, meas_date=meas_date,
            n_records=n_records, n_samps=n_samps, nchan=nchan,
            subject_info=patient, physical_max=physical_max,
            physical_min=physical_min, record_length=record_length,
            subtype=subtype, tal_idx=tal_idx)

In my local copy, I added recording_info=meas_id, after subject_info=patient, so that the recording info actually gets passed to the final raw object (before it was just discarded). I'm not sure if that's the proper naming convention or not, but I couldn't find any info on whether MNE standardized the naming and location of that kind of extended metadata.

Also, if you want to do some more extensive testing of the new header parsing you can base it off the tests I wrote for mne-bids here:

https://github.com/mne-tools/mne-bids/blob/18cc4f912495e8de8ab30d9c3077b19c923bebf4/mne_bids/tests/test_copyfiles.py#L116-L142

The MNE test .edf and .bdf files I looked at don't seem to have any of those header fields actually filled out, so manually adding the header info pre-test might be useful!

EDIT: To answer your other question, yes the Github link you used for me is fine!

[cbrnr]

LGTM, let me know if you want me to take a look at the missing feature once you've implemented it.

[sappelhoff]

LGTM overall once Clemens' comments are taken into account

[larsoner]

In my local copy, I added recording_info=meas_id, after subject_info=patient, so that the recording info actually gets passed to the final raw object (before it was just discarded).

It gets passed into a dict that ends up living only in a private attribute raw._raw_extras[0], so people should not rely on these values to exist (see extensive discussion here), so I'm -1 on adding these until we decide to make some EDF attributes public.

[larsoner]

@cbrnr made the tests more complete, should cover what you asked about

[cbrnr]

@larsoner did you push the changes?

[larsoner]

nope, forgot the -f so it didn't go through -- it's there now, though!