In previous version of this software, there used to be some assumptions about the ordering of channels. This was removed at some point in the main mne package, but the changes were never propagated here. This commit corrects this behaviour and attempt to follow the structure of upstream changes.
Reference issue
Fixes #530 .
What does this implement/fix?
Previous version made the assumption that the different wavelengths for a channel were adjacent. This was changed in the upstream version of mne-python and never updated here. This caused some kinds of data to be parsed incorrectly, leading to missing entries in time base quality functions. This corrects this incorrect behavior.
Additional information
At the present, I haven't made any changes to the testing of these functions, as it is a faulty assumption to say that there will never be zero entries for a channel. More thought should be given as to how this is tested by a maintainer to avoid future regression, though this fix should prevent any regardless.
Additionally, the old order of channels being adjacent results in different behavior in the function
Channels used to be plotted with wavelength adjacent, as show below.
The new behavior due to the previously mentioned upstream changes to channel ordering is shown below.
Personally, I don't find this very ergonomic, as the channels with differing wavelength are no longer adjacent, even though their values are the same. This behavior isn't technically incorrect, but perhaps undesirable given the purpose of this being to evaluate signal quality (these are different datasets being plotted btw, but that isn't relevant to this point).
larsoner
commented
6 months ago
In previous version of this software, there used to be some assumptions about the ordering of channels. This was removed at some point in the main mne package, but the changes were never propagated here. This commit corrects this behaviour and attempt to follow the structure of upstream changes.
Reference issue
Fixes #530 .
What does this implement/fix?
Previous version made the assumption that the different wavelengths for a channel were adjacent. This was changed in the upstream version of mne-python and never updated here. This caused some kinds of data to be parsed incorrectly, leading to missing entries in time base quality functions. This corrects this incorrect behavior.
Additional information
At the present, I haven't made any changes to the testing of these functions, as it is a faulty assumption to say that there will never be zero entries for a channel. More thought should be given as to how this is tested by a maintainer to avoid future regression, though this fix should prevent any regardless.
Additionally, the old order of channels being adjacent results in different behavior in the function
https://github.com/mne-tools/mne-nirs/blob/3ba42cd47bb4f5e6bc6c15d4677bb2b2177cb6ae/mne_nirs/visualisation/_plot_quality_metrics.py#L11
Channels used to be plotted with wavelength adjacent, as show below.
The new behavior due to the previously mentioned upstream changes to channel ordering is shown below.
Personally, I don't find this very ergonomic, as the channels with differing wavelength are no longer adjacent, even though their values are the same. This behavior isn't technically incorrect, but perhaps undesirable given the purpose of this being to evaluate signal quality (these are different datasets being plotted btw, but that isn't relevant to this point).