ARROW-3953: [Python] Compat with pandas 0.24 rename of MultiIndex labels -> codes

python/pyarrow/pandas_compat.py

@@ -726,6 +726,14 @@ def _pandas_type_to_numpy_type(pandas_type):
         return np.dtype(pandas_type)
 
 
+def _get_multiindex_codes(mi):
+    # compat for pandas < 0.24 (MI labels renamed to codes).
+    if isinstance(mi, pd.MultiIndex):
+        return mi.codes if hasattr(mi, 'codes') else mi.labels
+    else:
+        return None
+
+
 def _reconstruct_columns_from_metadata(columns, column_indexes):
     """Construct a pandas MultiIndex from `columns` and column index metadata
     in `column_indexes`.
@@ -752,7 +760,7 @@ def _reconstruct_columns_from_metadata(columns, column_indexes):
     # Get levels and labels, and provide sane defaults if the index has a
     # single level to avoid if/else spaghetti.
     levels = getattr(columns, 'levels', None) or [columns]
-    labels = getattr(columns, 'labels', None) or [
+    labels = _get_multiindex_codes(columns) or [
         pd.RangeIndex(len(level)) for level in levels
     ]
 
@@ -779,7 +787,7 @@ def _reconstruct_columns_from_metadata(columns, column_indexes):
 
         new_levels.append(level)
 
-    return pd.MultiIndex(levels=new_levels, labels=labels, names=columns.names)
+    return pd.MultiIndex(new_levels, labels, names=columns.names)
 
 
 def _table_to_blocks(options, block_table, memory_pool, categories):
@@ -796,7 +804,7 @@ def _table_to_blocks(options, block_table, memory_pool, categories):
 def _flatten_single_level_multiindex(index):
     if isinstance(index, pd.MultiIndex) and index.nlevels == 1:
         levels, = index.levels
-        labels, = index.labels
+        labels, = _get_multiindex_codes(index)
 
         # Cheaply check that we do not somehow have duplicate column names
         if not index.is_unique:

python/pyarrow/tests/test_convert_pandas.py

@@ -176,6 +176,16 @@ def test_multiindex_columns_unicode(self):
         df = pd.DataFrame([(1, 'a'), (2, 'b'), (3, 'c')], columns=columns)
         _check_pandas_roundtrip(df, preserve_index=True)
 
+    def test_multiindex_doesnt_warn(self):
+        # ARROW-3953: pandas 0.24 rename of MultiIndex labels to codes
+        columns = pd.MultiIndex.from_arrays([['one', 'two'], ['X', 'Y']])
+        df = pd.DataFrame([(1, 'a'), (2, 'b'), (3, 'c')], columns=columns)
+
+        with pytest.warns(None) as record:
+            _check_pandas_roundtrip(df, preserve_index=True)
+
+        assert len(record) == 0
+
     def test_integer_index_column(self):
         df = pd.DataFrame([(1, 'a'), (2, 'b'), (3, 'c')])
         _check_pandas_roundtrip(df, preserve_index=True)