Fix two pandas 3.0 incompatibilities (StringDtype groupby, Series positional indexing)

changelog/321.fix.md

@@ -0,0 +1 @@
+Restored compatibility with pandas 3.0 by replacing two calls that pandas 3 no longer accepts: `numpy.issubdtype(col.dtype, numpy.number)` in `RunGroupBy.__init__` (raised on `StringDtype` meta columns) and `Series[0]` positional indexing in `_xarray._many_to_one` (raised `KeyError: 0`). The previously-failing `ScmRun.groupby` and `ScmRun.to_nc` paths now run on both pandas 2 and pandas 3.

src/scmdata/_xarray.py

@@ -198,7 +198,10 @@ def _many_to_one(df, col1, col2):
     # thanks https://stackoverflow.com/a/59091549
     checker = df[[col1, col2]].drop_duplicates()
 
-    max_count = checker.groupby(col2).count().max()[0]
+    # ``.iloc[0]`` rather than ``[0]``: pandas 3.0 removed positional
+    # integer indexing on label-indexed Series, so ``[0]`` would raise
+    # ``KeyError: 0`` on the Series returned by the chained ``.max()``.
+    max_count = checker.groupby(col2).count().max().iloc[0]
     if max_count < 1:  # pragma: no cover # emergency valve
         raise AssertionError
 

src/scmdata/groupby.py

@@ -57,8 +57,12 @@ def __init__(
         m = run.meta.reset_index(drop=True)
         self.na_fill_value = float(na_fill_value)
 
-        # Work around the bad handling of NaN values in groupbys
-        if any([np.issubdtype(m[c].dtype, np.number) for c in m]):
+        # Work around the bad handling of NaN values in groupbys.
+        # pd.api.types.is_numeric_dtype accepts every dtype scmdata
+        # ever emits; np.issubdtype(..., np.number) raises on
+        # pandas 3.0's default StringDtype with
+        # ``TypeError: Cannot interpret '<StringDtype(...)>'``.
+        if any([pd.api.types.is_numeric_dtype(m[c]) for c in m]):
             if (m == na_fill_value).any(axis=None):
                 raise ValueError(
                     "na_fill_value conflicts with data value. Choose a na_fill_value "