Resolve external warnings #1208
Conversation
Resolves an large amount of warnings of the type DeprecationWarning: `np.float` is a deprecated alias for the builtin `float` that were present in scikit-learn when using NumPy > 1.20. scikit-learn>=1.0 is also ABI stable and follows semantic versioning, making maintaining a range of versions easier.
Resolves the following warning: tests/test_osmnx.py::test_features /home/runner/work/osmnx/osmnx/osmnx/features.py:684: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame. Try using .loc[row_indexer,col_indexer] = value instead See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy gdf.loc[:, "geometry"] = gdf["geometry"].make_valid()
Resolved a VisibleDeprecationWarning by specifying dtype=object when creating a NumPy array from potentially ragged nested sequences in the _verify_edge_attribute function. This ensures that the array can handle irregular shapes without raising deprecation warnings, maintaining compatibility with future NumPy versions.
tests/test_osmnx.py::test_routing
/home/runner/work/osmnx/osmnx/osmnx/routing.py:473: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.
values_float = (np.array(tuple(G.edges(data=attr)))[:, 2]).astype(float)
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #1208 +/- ##
=======================================
Coverage 98.31% 98.31%
=======================================
Files 24 24
Lines 2369 2370 +1
=======================================
+ Hits 2329 2330 +1
Misses 40 40 ☔ View full report in Codecov by Sentry. |
| @@ -680,7 +680,7 @@ def _filter_features( | |||
| Filtered GeoDataFrame of features. | |||
| """ | |||
| # remove any null or empty geometries then fix any invalid geometries | |||
| gdf = gdf[~(gdf["geometry"].isna() | gdf["geometry"].is_empty)] | |||
There was a problem hiding this comment.
Isn't the SettingWithCopyWarning here a false positive? That is, we don't care about writing back to the original dataframe, because we're overwriting the variable itself. Given the expensiveness of .copy(), it may be better to filter/disable the warning on this line instead (assuming this is indeed a false positive).
There was a problem hiding this comment.
Also, regarding:
Addressed a VisibleDeprecationWarning in the _verify_edge_attribute function by explicitly setting dtype=object for a NumPy array to handle ragged nested sequences, maintaining future compatibility.
How did you produce this warning? I'm using numpy 2 and haven't seen it. I'm trying to figure it out, but dtype=object doesn't make sense to me. This shouldn't be a ragged array, but rather always a (u, v, attr_value).
There was a problem hiding this comment.
All warnings were from a CI run at that time on the main branch.
Unfortunately, at this time I don't have the capacity to follow up on this, so feel free to close or modify this PR.
This PR addresses resolves several 3 issues caused by external dependencies, that were present one or multiple times in osmnx CI test runs.
Upgrade scikit-learn to >=1.0:
scikit-learn>=1.0, resolving deprecation warnings related tonp.floatwhen using NumPy > 1.20.Resolve
SettingWithCopyWarningin Pandas:SettingWithCopyWarninginfeatures.pyby ensuring a copy of the DataFrame slice is made before modifications.Fix
VisibleDeprecationWarningin NumPy:VisibleDeprecationWarningin the_verify_edge_attributefunction by explicitly settingdtype=objectfor a NumPy array to handle ragged nested sequences, maintaining future compatibility.Some more details are in the commit messages of each commit.
CI is passing on my test branch, where I triggered the CI explicitly.
There's one external issue left, but I think that has to be fixed upstream (tracking: geopandas/geopandas#3403). The other issues look to be triggered by OSMnx (or its tests) itself.