BUG: read raw eyelink file that switches from binocular to monocular mode

[scott-huberty]

This is an issue as PR.

A file from one of our studies causes read_raw_eyelink to break. The file starts recording binocular data and switches to monocular data mid-way through the recording. This is something that EyeLink acquisition software allows.

Although I can't share the file, I’ve created a test that simulates the error. I'm still investigating a fix.

Potential Solutions

• Add a Check for this scenario (binocular to monocular switch or vice versa) and raise a more informative error. The user can then regenerate their ASCII file, outputting monocular data only, which in theory should then be readable by MNE. I think this minimizes technical debt for us.
• Devise a workaround to handle this situation, such as adding a parameter to read_raw_eyelink that reads data for one eye only. This is probably (a lot?) more development work.
• Try to convince EyeLink (SR Research) to open-source a python reader for their format 😆

CC @anomalosepia

[larsoner]

To me naively the easiest solution is actually to treat the data as binocular, but set one of the eyes to nan when it switches to monocular. Then if you want a "monocular instance" you just raw.drop_channels(["eye_right"]) or whatever (assuming left is the one that was on the whole time).

[scott-huberty]

To me naively the easiest solution is actually to treat the data as binocular, but set one of the eyes to nan when it switches to monocular. Then if you want a "monocular instance" you just raw.drop_channels(["eye_right"]) or whatever (assuming left is the one that was on the whole time).

Hmm yeah could do.. I'll give it a try and will see how much complexity it adds.