FIX: Revert

mne/time_frequency/tests/test_spectrum.py

@@ -306,6 +306,8 @@ def test_unaggregated_spectrum_to_data_frame(raw, long_format, method, output):
     pytest.importorskip("pandas")
     from pandas.testing import assert_frame_equal
 
+    from mne.utils.dataframe import _inplace
+
     # aggregated spectrum → dataframe
     orig_df = raw.compute_psd(method=method).to_data_frame(long_format=long_format)
     # unaggregated welch or complex multitaper →
@@ -327,7 +329,7 @@ def test_unaggregated_spectrum_to_data_frame(raw, long_format, method, output):
     orig_df = orig_df.loc[subset(orig_df["freq"])]
     # sort orig_df, because at present we can't actually prevent pandas from
     # sorting at the agg step *sigh*
-    orig_df.sort_values(by=grouping_cols, ignore_index=True)
+    _inplace(orig_df, "sort_values", by=grouping_cols, ignore_index=True)
     # aggregate
     df = df.drop(columns=drop_cols)
     gb = df.groupby(grouping_cols, as_index=False, observed=False)

mne/utils/dataframe.py

@@ -4,10 +4,13 @@
 # License: BSD-3-Clause
 # Copyright the MNE-Python contributors.
 
+from inspect import signature
+
 import numpy as np
 
 from ..defaults import _handle_default
 from ._logging import logger, verbose
+from .check import check_version
 
 
 @verbose
@@ -47,6 +50,24 @@ def _convert_times(times, time_format, meas_date=None, first_time=0):
     return times
 
 
+def _inplace(df, method, **kwargs):
+    # Handle transition: inplace=True (pandas <1.5) → copy=False (>=1.5)
+    # and 3.0 warning:
+    # E   DeprecationWarning: The copy keyword is deprecated and will be removed in a
+    #     future version. Copy-on-Write is active in pandas since 3.0 which utilizes a
+    #     lazy copy mechanism that defers copies until necessary. Use .copy() to make
+    #     an eager copy if necessary.
+    _meth = getattr(df, method)  # used for set_index() and rename()
+
+    if check_version("pandas", "3.0"):
+        return _meth(**kwargs)
+    elif "copy" in signature(_meth).parameters:
+        return _meth(**kwargs, copy=False)
+    else:
+        _meth(**kwargs, inplace=True)
+        return df
+
+
 @verbose
 def _build_data_frame(
     inst,
@@ -74,16 +95,16 @@ def _build_data_frame(
         df.insert(i, k, v)
     # build Index
     if long_format:
-        df.set_index(keys=default_index)
+        df = _inplace(df, "set_index", keys=default_index)
         df.columns.name = col_kind
     elif index is not None:
-        df.set_index(keys=index)
+        df = _inplace(df, "set_index", keys=index)
         if set(index) == set(default_index):
             df.columns.name = col_kind
     # long format
     if long_format:
         df = df.stack().reset_index()
-        df.rename(columns={0: "value"})
+        df = _inplace(df, "rename", columns={0: "value"})
         # add column for channel types (as appropriate)
         ch_map = (
             None
@@ -101,7 +122,7 @@ def _build_data_frame(
             df.insert(col_index, "ch_type", ch_type)
         # restore index
         if index is not None:
-            df.set_index(keys=index)
+            df = _inplace(df, "set_index", keys=index)
         # convert channel/vertex/ch_type columns to factors
         to_factor = [
             c for c in df.columns.tolist() if c not in ("freq", "time", "value")