FIX: ctf data can sometimes have no date time.

[kingjr]

I encountered a bug when reading a CTF dataset with an empty string in the res4['data_date'] and res4['data_time'] fields.

This sets the field to January 1st of the year 1001 (earlier make a date parsing bug)

[larsoner]

I would rather set the date to 1) the epoch (0., 0.) in FIF form, or 2) None, but (2) probably makes working with things like annotations harder. (0., 0.) seems like a better choice than 1001. Or were you trying to avoid possible collisions with anonymized dates...? In that case I could be persuaded :)

[kingjr]

new commit: I had to use a higher date, else there is an error at reading time after saving the ctf as a fif.

Thanks Eric. I did not understand option 1), but I agree that option 2) would be best: we don't have a date, so we shouldn't try to force it. However, it will likely require a lot of downstream corrections.

I'll be using this branch, as it unblocks my pipeline, but feel free to take over.

[larsoner]

1970/01/01 okay for you @kingjr ?

@agramfort @hoechenberger @drammock feel free to merge if happy

[kingjr]

LGTM, thanks Eric!

[drammock]

LGTM, just one tiny suggestion to make the logger.info message clearer.