Remove or split up Frame methods that use the index

python/cudf/cudf/core/column_accessor.py

@@ -342,6 +342,26 @@ def select_by_label(self, key: Any) -> ColumnAccessor:
                     return self._select_by_label_with_wildcard(key)
             return self._select_by_label_grouped(key)
 
+    def get_labels_by_index(self, index: Any) -> tuple:
+        """Get the labels corresponding to the provided column indices.
+
+        Parameters
+        ----------
+        index : integer, integer slice, or list-like of integers
+            The column indexes.
+
+        Returns
+        -------
+        tuple
+        """
+        if isinstance(index, slice):
+            start, stop, step = index.indices(len(self._data))
+            return self.names[start:stop:step]
+        elif pd.api.types.is_integer(index):
+            return (self.names[index],)
+        else:
+            return tuple(self.names[i] for i in index)
+
     def select_by_index(self, index: Any) -> ColumnAccessor:
         """
         Return a ColumnAccessor composed of the columns
@@ -355,13 +375,7 @@ def select_by_index(self, index: Any) -> ColumnAccessor:
         -------
         ColumnAccessor
         """
-        if isinstance(index, slice):
-            start, stop, step = index.indices(len(self._data))
-            keys = self.names[start:stop:step]
-        elif pd.api.types.is_integer(index):
-            keys = (self.names[index],)
-        else:
-            keys = tuple(self.names[i] for i in index)
+        keys = self.get_labels_by_index(index)
         data = {k: self._data[k] for k in keys}
         return self.__class__(
             data, multiindex=self.multiindex, level_names=self.level_names,

python/cudf/cudf/core/dataframe.py

@@ -372,9 +372,9 @@ class _DataFrameIlocIndexer(_DataFrameIndexer):
     def _getitem_tuple_arg(self, arg):
         # Iloc Step 1:
         # Gather the columns specified by the second tuple arg
-        columns_df = self._frame._get_columns_by_index(arg[1])
-
-        columns_df._index = self._frame._index
+        columns_df = self._frame._from_data(
+            self._frame._data.select_by_index(arg[1]), self._frame._index
+        )
 
         # Iloc Step 2:
         # Gather the rows specified by the first tuple arg
@@ -422,9 +422,9 @@ def _getitem_tuple_arg(self, arg):
 
     @_cudf_nvtx_annotate
     def _setitem_tuple_arg(self, key, value):
-        columns = self._frame._get_columns_by_index(key[1])
-
-        for col in columns:
+        # TODO: Determine if this usage is prevalent enough to expose this
+        # selection logic at a higher level than ColumnAccessor.
+        for col in self._frame._data.get_labels_by_index(key[1]):
             self._frame[col].iloc[key[0]] = value
 
     def _getitem_scalar(self, arg):
@@ -612,7 +612,7 @@ def __init__(
                 new_df = self._from_arrays(data, index=index, columns=columns)
 
             self._data = new_df._data
-            self.index = new_df._index
+            self._index = new_df._index
         elif hasattr(data, "__array_interface__"):
             arr_interface = data.__array_interface__
             if len(arr_interface["descr"]) == 1:
@@ -621,7 +621,7 @@ def __init__(
             else:
                 new_df = self.from_records(data, index=index, columns=columns)
             self._data = new_df._data
-            self.index = new_df._index
+            self._index = new_df._index
         else:
             if is_list_like(data):
                 if len(data) > 0 and is_scalar(data[0]):
@@ -632,7 +632,7 @@ def __init__(
                     new_df = DataFrame(data=data, index=index)
 
                     self._data = new_df._data
-                    self.index = new_df._index
+                    self._index = new_df._index
                 elif len(data) > 0 and isinstance(data[0], Series):
                     self._init_from_series_list(
                         data=data, columns=columns, index=index
@@ -650,6 +650,11 @@ def __init__(
                     data, index=index, columns=columns, nan_as_null=nan_as_null
                 )
 
+        if self._data.nrows > 0 and self._data.nrows != len(self._index):
+            raise ValueError(
+                f"Shape of passed values is {self.shape}, indices imply "
+                f"({len(self._index)}, {self._num_columns})"
+            )
         if dtype:
             self._data = self.astype(dtype)._data
 
@@ -855,10 +860,10 @@ def _from_data(
         data: MutableMapping,
         index: Optional[BaseIndex] = None,
         columns: Any = None,
+        *args,
+        **kwargs,
     ) -> DataFrame:
-        out = super()._from_data(data, index)
-        if index is None:
-            out.index = RangeIndex(out._data.nrows)
+        out = super()._from_data(data=data, index=index)
         if columns is not None:
             out.columns = columns
         return out