BUG: fix conversion of WKB to geometries in case of using ArrowDtype and missing values #321
Conversation
…and missing values
pyogrio/geopandas.py
Outdated
| if PANDAS_GE_20 and wkb_values.dtype != object: | ||
| # for example ArrowDtype will otherwise create numpy array with pd.NA | ||
| wkb_values = wkb_values.to_numpy(na_value=None) |
There was a problem hiding this comment.
I decided to this only conditionally, because otherwise it would impact performance for the default case (where the binary pyarrow array was already converted to an object dtype column with None values. When specifying an na_value in that case, it would check for missing values and overwrite them with None, which be an unnecessary operation).
Checking for "not object dtype" instead of explicitly checking for "is ArrowDtype" felt the most future proof (eg when pandas adds a proper binary dtype (using Arrow under the hood), it might have the same issue but without being an ArrowDtype instance)
There was a problem hiding this comment.
Thanks @jorisvandenbossche !
Can you please add a changelog entry for 0.7.2 and update the release date (also, I'm not available to do a release today)
Closes #319