lint & format weights/*.py part 7

[jGaboardi]

This PR:

• formats and lints weights/*.py – part 7 of $n$

This is a bit more than I had thought it would be. Can split into 2 PRs if desired, @martinfleis

[codecov]

Codecov Report

Merging #655 (f078de3) into main (c655e99) will not change coverage.
The diff coverage is 97.1%.

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #655   +/-   ##
=====================================
  Coverage   84.7%   84.7%           
=====================================
  Files        139     139           
  Lines      14859   14859           
=====================================
  Hits       12591   12591           
  Misses      2268    2268           

Files	Coverage Δ
libpysal/weights/tests/test_adjlist.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_contiguity.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_distance.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_nx.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_raster.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_spatial_lag.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_spintW.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test_weights_IO.py	100.0% <100.0%> (ø)
libpysal/weights/tests/test__contW_lists.py	97.7% <97.2%> (-<0.1%)	⬇️
libpysal/weights/raster.py	62.4% <80.0%> (ø)

[martinfleis]

This changes the type, doesn't it? Can you test that?

[jGaboardi]

So after looking into this a bit, there's a bit of strangeness going on with pytest-cov. The report claims those lines are covered, but after some tinkering locally it appears they are actually not. Going to investigate further.

[jGaboardi]

So (unless I am missing something exceedingly obvious), indeed those lines were not being tested... and they fail with the change... BUT they also fail on main when actually tested... This adds another level of anxiety if we can't trust the coverage reports from pytest...

from libpysal.weights import raster
pytest.importorskip("xarray")
da2_missing_in = raster.testDataArray((1, 4, 4), missing_vals=True)
w2_missing = raster.da2W(da2_missing_in, "rook", n_jobs=-1)
da2_missing_out = raster.w2da(
    da2_missing_in.data.flatten(), w2_missing, da2_missing_in.attrs, None
)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[2], line 6
      4 da2_missing_in = raster.testDataArray((1, 4, 4), missing_vals=True)
      5 w2_missing = raster.da2W(da2_missing_in, "rook", n_jobs=-1)
----> 6 da2_missing_out = raster.w2da(
      7     da2_missing_in.data.flatten(), w2_missing, da2_missing_in.attrs, None
      8 )

File ~/libpysal/libpysal/weights/raster.py:343, in w2da(data, w, attrs, coords)
    341     raise TypeError("w must be an instance of weights.W")
    342 if hasattr(w, "index"):
--> 343     da = _index2da(data, w.index, attrs, coords)
    344 else:
    345     raise AttributeError(
    346         "This method requires `w` object to include `index` "
    347         "attribute that is built as a `pandas.MultiIndex` object."
    348     )

File ~/libpysal/libpysal/weights/raster.py:570, in _index2da(data, index, attrs, coords)
    568 else:
    569     data_complete = np.empty(shape, data.dtype)
--> 570 data_complete[indexer] = data
    571 coords = {}
    572 for dim, lev in zip(dims, idx.levels, strict=True):

ValueError: shape mismatch: value array of shape (16,) could not be broadcast to indexing result of shape (6,)

[jGaboardi]

So I suppose I'll revert and create an issue for this?

[martinfleis]

Yes, that sounds the best

[jGaboardi]

Do we think all changes to raster should be reverted or just that one?

[martinfleis]

just this line and add noqa, the rest is fine

[jGaboardi]

Done