While working on #76 I realized that I can't put np.datetime64('NaT') into the ds.time. However, this causes problems because sometimes timestamps really are missing. We need to figure out a consistent way to address this issue. It shouldn't be too bad because we should be able to interpolate or extrapolate pretty easily (the sample rate is known after all), but I do think we should issue a warning whenever this happens, and also probably have a flag somewhere that indicates where the bad timestamps were. Maybe a boolean array bad_time variable in the dataset?
While working on #76 I realized that I can't put
np.datetime64('NaT')
into theds.time
. However, this causes problems because sometimes timestamps really are missing. We need to figure out a consistent way to address this issue. It shouldn't be too bad because we should be able to interpolate or extrapolate pretty easily (the sample rate is known after all), but I do think we should issue a warning whenever this happens, and also probably have a flag somewhere that indicates where the bad timestamps were. Maybe a boolean arraybad_time
variable in the dataset?