Open ericphanson opened 12 months ago
I've realized this totally contradicts the docstring of TimestampAnnotatedList
which explicitly rounds for supposed EDF+ compliance. However I keep trying to find where in the spec it says these onsets and durations must be 8 ascii characters and fail to do so. Would love some help @ararslan
This does not help with my actual bug (#74) although I do think we should make sure small durations / onsets are supported. IMO that doesn't need to write them in full precision like this PR though.
I incorporated this into https://github.com/beacon-biosignals/EDF.jl/pull/75, there choosing 100ms precision matching EDFlib. But we could allow more precision there too.
https://www.edfplus.info/specs/edfplus.html#tal
We are currently subjecting our onsets and durations to the strict 8-char limitation of EDF headers, but that does not seem necessary.
Note this is mostly orthogonal to #70, since that PR does not use scientific notation in TALs anyway (but it does allow underflow-to-zero there, which we can get rid of this way).