Get nappy working with Xarray

[agstephens]

This is quite a big task, because there is a lot of code that translates NASA Ames concepts/objects to/from NetCDF-style concepts. However, it is more complex than that because we previously used cdms as the library to convert to NetCDF. Now, we are moving to xarray.

NOTE: Initially, we only need the Xarray-to-NASAAmes|CSV functionality to work!

Here is an overview of the order of tasks:

• [x] Create a new branch and basically replace all references to cdms|CDMS|Cdms with xarray|xr - https://github.com/cedadev/nappy/tree/xarray-port
• [x] Check that the basic unit tests still work: https://github.com/cedadev/nappy/blob/xarray-port/tests/test_na_file.py
• [x] Rewrite the following files so that each line is run as a proper unit test that wraps calling the cli. In all cases we just want to check the output file has been written, and check that the output file generated matching those in the tests/expected_outputs/ directory:

  tests/tests-na2nc.py.txt
  tests/tests-nc2csv.py.txt
  tests/tests-nc2na.py.txt

• [x] Take the nappy/nappy_api.py file contents and create a tests/test_nappy_api.py - create unit tests for each of the examples given.
• [x] Run the tests/test_nappy_api.py tests - they will fail... Now start to modify the contents of the nappy.nc_interface modules to get them working using Xarray.
• [x] Some for the files reference a package called: cdms_utils (http://proj.badc.rl.ac.uk/cows/browser/cows_support/cdms_utils/trunk/lib/cdms_utils). We just need to create the following functions that are actually used by nappy, and put them in a local nappy/nc_interface/xarray_utils.py module. There might be some stuff we can steal from: https://github.com/roocs/roocs-utils/blob/master/roocs_utils/xarray_utils/xarray_utils.py . The required functions are:

  
  cdms_utils.var_utils.getBestName(var)
  cdms_utils.var_utils.getMissingValue(var)

cdms_utils.axis_utils.isUniformlySpaced(axis) cdms_utils.axis_utils.areAxesIdentical(ax1, ax2) cdms_utils.axis_utils.isAxisRegularlySpacedSubsetOf(ax1, ax2)

- [x] Then it's a brute force (let's get it working approach!)