Ensure compatibility with pandas CoW

RELEASE_NOTES.rst

@@ -4,6 +4,7 @@ Next release
 All changes
 -----------
 
+- Ensure compatibility with pandas upcoming 3.0 Copy-on-Write behaviour (:pull:`842`).
 - Update tutorial Westeros multinode to include code-based hints for in-depth questions (:pull:`798`).
 - :func:`.make_df` can now create partly-filled :class:`DataFrames <pandas.DataFrame>` for indexed sets; not only parameters (:pull:`784`).
 - New function :func:`.util.copy_model` that exposes the behaviour of the :program:`message-ix copy-model` CLI command to other Python code (:pull:`784`).

message_ix/macro.py

@@ -272,6 +272,8 @@ def extrapolate(
         The index does *not* have a ``year`` dimension; the data are implicitly for
         `ym1`.
     """
+    from sys import version_info
+
     from scipy.optimize import curve_fit
 
     def f(x, b, m):
@@ -286,8 +288,11 @@ def fitted_intercept(df: pd.DataFrame) -> float:
 
     # Apply fitted_intercept to grouped data
     groupby_cols = set(model_data.columns) - {"year", "value"}
+    apply_kwargs = {"include_groups": False} if version_info.minor > 8 else {}
     result = (
-        model_data.groupby(list(groupby_cols)).apply(fitted_intercept).rename("value")
+        model_data.groupby(list(groupby_cols))
+        .apply(fitted_intercept, **apply_kwargs)
+        .rename("value")
     )
 
     # Convert "commodity" and "level" to "sector"