Possible regression for reading EDF files in version 0.17

External Email - Use Caution

I have data files in EDF format that contain both SEEG and surface electrodes. Our naming strategy for SEEG electrodes results in some duplicate names of surface EEG electrodes. Prior versions of MNE raised a runtime warning and appended sequential digits to the electrode names. The current version now raises an error such as "ValueError: Channel EEG F1-Ref-0 has no associated original unit."

If dropping support for duplicate electrode names was the desired intent, it would probably be better to change the runtime warning into an error so the problem is clear.

If this was not the intent, I have a test file demonstrating the problem and an EDF specific solution that I can make into a pull request. The new behavior appears to have occurred as result of changes meant to make the original units available in the final raw data structure. Please let me know if I should proceed with making a pull request. Thanks.

Larry

External Email - Use Caution

hi Larry,

it was not our intent to break this. Can you open a PR or at least
share the problematic file?
We'll push the fix asap in a new minor release.

thanks
Alex