Numpy support should be added as in GeoCAT-ncomp
erogluorhan opened this issue · 4 comments
Numpy
support should be added to the GeoCAT-temp functionality because:
-
While
xArray
is having an increasing use among the atmospheric science community, it requires some dedicated training, tutorial, etc. to overcome a learning curve. On the other hand,Numpy
has been a Python data access/analysis standard among almost everybody for a long while. -
Not all of the use cases with GeoCAT-comp stack have to includes big data or related tasks such as parallelization that would require
Dask
, soNumpy
would provide a very practical way of running GeoCAT-comp functionality for users to work on small data sets.
@pilotchute @anissa111 Due to above points, I strongly prefer to provide Numpy
support in our f2py
effort as we did so far in GeoCAT-ncomp; however, I would like to hear your thoughts, if any.
If/when the team decides on Numpy
support, I can add it into the current GeoCAT-temp structure, or @pilotchute if you would like to handle it on your own, you are welcome as well. Instead of implementing it the same way as we did it GeoCAT-ncomp (i.e. every single function had similar if
-checks for Numpy
vs xArray
handling), we might want to implement a helper function to handle conversion of Numpy
inputs into xArray
data structure. Thoughts?
- Implement numpy support for linint2
- make sure that
xi
yi
are available as optionals in method signature - setup check for
fi
being numpy, andxi
yi
existant - build Xarray from
fi xi yi
- make sure that
- Test numpy support
Numpy support implemented for linint2, tested on 8 cores, completes at same speed as single chunk and pre-chunked Xarrays.