NeurodataWithoutBorders
Convention to specify SIUnit
While going through the File tutorial, I noticed the SIunit placeholder was used few times.
Looking for examples, I read through the TimeSeries class definition and also the schema documentation
From the schema doc:
The data stored can be in any unit. The attributes of the data field must indicate the SI unit that the data relates to (or appropriate counterpart, such as color-space) and the multiplier necessary to convert stored values to the specified SI unit.
From the TimeSeries class definition:
https://github.com/NeurodataWithoutBorders/pynwb/blob/67a3bca2c37c42cb963b2ea1509952567bdf35cd/src/pynwb/base.py#L106-L108
That led me to think the unit must be specified in their plural form: Seconds instead of Second.
That said, looking at Wikipedia, the unit names are in their singular form:
- second
- metre
- ...
What should be the recommended form ? Also should it be lower or upper cased ?
cc: @nicain
How about we use the standard abbreviations for the units? Like s / V in their official casing?
Personally I think "seconds" is more descriptive than "s", but that's just my personal feeling, but I'm fine with whatever makes most sense for neuroscientists. Once you decide, please add a short section to the format docs to describe it:
https://github.com/NeurodataWithoutBorders/nwb-schema/blob/dev/docs/format/source/format_description.rst
We should probably also have a dict somewhere in PyNWB where standard unit strings are stored so that users can a) look up the the correct strings there and b) we can do error checking on assignment.
what about using something external that provides support for units already rather than writing it?
e.g. https://pypi.python.org/pypi/units/
Thanks everyone for the feedback
Personally I think "seconds" is more descriptive than "s",
:+1:
I'm fine with whatever makes most sense for neuroscientists
:+1:
dict somewhere in PyNWB
That would work (or also something like pynwb.units.second, ... ), at first I think we need to agree on what should be the convention for units and then come up with a technical solution.
I wonder how unit are currently stored in files. @NeurodataWithoutBorders/hck04 ?
something external ... provides support for units
That would be an option, but the unit package referenced uses single letter unit which are not very expressive
https://bitbucket.org/adonohue/units/src/b187fbf3c1d403df9326670e7a00e2522fc74215/units/si.py?at=default&fileviewer=file-view-default#si.py-6:27
Some other inconsistencies with units:
- SpatialSeries is hardcoded to meters instead of just being the default
- Unspecified default values are inconsistently defined (None, or 'None')
I think single-letter units are fine as long as we have a standard set of units and their single-letter equivalents. For visualization, we are much more likely to want the single-letter versions. Ideally, the API can give us access to both versions. I'd like to be able to do something like:
>>> data.unit
<Unit 'Second'>
>>> str(data.unit)
's'
