Remove _source_index from MultiIndex

python/cudf/cudf/core/dataframe.py

@@ -3983,7 +3983,7 @@ def sort_index(
                     ]
                 else:
                     labels = [self.index._get_level_label(level)]
-                inds = self.index._source_data[labels].argsort(
+                inds = self.index.to_frame(index=False)[labels].argsort(
                     ascending=ascending, na_position=na_position
                 )
             else:

python/cudf/cudf/core/frame.py

@@ -1856,7 +1856,7 @@ def sample(
             if isinstance(self, cudf.MultiIndex):
                 # TODO: Need to update this once MultiIndex is refactored,
                 # should be able to treat it similar to other Frame object
-                result = cudf.Index(self._source_data[gather_map])
+                result = cudf.Index(self.to_frame(index=False)[gather_map])
             else:
                 result = self[gather_map]
                 if not keep_index:
@@ -3168,9 +3168,13 @@ def _reindex(
             index = cudf.core.index.as_index(index)
 
             if isinstance(index, cudf.MultiIndex):
-                idx_dtype_match = (
-                    df.index._source_data.dtypes == index._source_data.dtypes
-                ).all()
+                idx_dtype_match = all(
+                    left_dtype == right_dtype
+                    for left_dtype, right_dtype in zip(
+                        (col.dtype for col in df.index._data.columns),
+                        (col.dtype for col in index._data.columns),
+                    )
+                )
             else:
                 idx_dtype_match = df.index.dtype == index.dtype
 
@@ -5220,7 +5224,7 @@ def _drop_rows_by_labels(
         # 1. Merge Index df and data df along column axis:
         # | id | ._index df | data column(s) |
         idx_nlv = obj._index.nlevels
-        working_df = obj._index._source_data
+        working_df = obj._index.to_frame(index=False)
         working_df.columns = [i for i in range(idx_nlv)]
         for i, col in enumerate(obj._data):
             working_df[idx_nlv + i] = obj._data[col]

python/cudf/cudf/core/groupby/groupby.py

@@ -1335,12 +1335,9 @@ def keys(self):
         if nkeys == 0:
             return cudf.core.index.as_index([], name=None)
         elif nkeys > 1:
-            return cudf.MultiIndex(
-                source_data=cudf.DataFrame(
-                    dict(zip(range(nkeys), self._key_columns))
-                ),
-                names=self.names,
-            )
+            return cudf.MultiIndex._from_data(
+                dict(zip(range(nkeys), self._key_columns))
+            )._set_names(self.names)
         else:
             return cudf.core.index.as_index(
                 self._key_columns[0], name=self.names[0]

python/cudf/cudf/core/index.py

@@ -2203,7 +2203,7 @@ def as_index(arbitrary, **kwargs) -> BaseIndex:
     elif isinstance(arbitrary, pd.MultiIndex):
         return cudf.MultiIndex.from_pandas(arbitrary)
     elif isinstance(arbitrary, cudf.DataFrame):
-        return cudf.MultiIndex(source_data=arbitrary)
+        return cudf.MultiIndex.from_frame(arbitrary)
     return as_index(
         column.as_column(arbitrary, dtype=kwargs.get("dtype", None)), **kwargs
     )