Remove internal usage of core.index.as_index in favor of cudf.Index (#…

…15851)

`cudf.Index.__init__` essentially calls `as_index` immediately internally. To avoid both from potentially diverging, the public `cudf.Index` should be preferred to ensure the public behaviors are used internally

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - Vyas Ramasubramani (https://github.com/vyasr)

URL: #15851

python/cudf/cudf/core/algorithms.py

@@ -6,7 +6,7 @@
 
 from cudf.core.column import as_column
 from cudf.core.copy_types import BooleanMask
-from cudf.core.index import RangeIndex, as_index
+from cudf.core.index import Index, RangeIndex
 from cudf.core.indexed_frame import IndexedFrame
 from cudf.core.scalar import Scalar
 from cudf.options import get_option
@@ -107,7 +107,7 @@ def factorize(values, sort=False, use_na_sentinel=True, size_hint=None):
         dtype="int64" if get_option("mode.pandas_compatible") else None,
     ).values
 
-    return labels, cats.values if return_cupy_array else as_index(cats)
+    return labels, cats.values if return_cupy_array else Index(cats)
 
 
 def _linear_interpolation(column, index=None):

python/cudf/cudf/core/column/methods.py

@@ -93,8 +93,6 @@ def _return_or_inplace(
                 else:
                     return cudf.Series(new_col, name=self._parent.name)
             elif isinstance(self._parent, cudf.BaseIndex):
-                return cudf.core.index.as_index(
-                    new_col, name=self._parent.name
-                )
+                return cudf.Index(new_col, name=self._parent.name)
             else:
                 return self._parent._mimic_inplace(new_col, inplace=False)

python/cudf/cudf/core/column/string.py

@@ -4391,7 +4391,7 @@ def code_points(self) -> SeriesOrIndex:
         if isinstance(self._parent, cudf.Series):
             return cudf.Series(new_col, name=self._parent.name)
         elif isinstance(self._parent, cudf.BaseIndex):
-            return cudf.core.index.as_index(new_col, name=self._parent.name)
+            return cudf.Index(new_col, name=self._parent.name)
         else:
             return new_col
 
@@ -4706,7 +4706,7 @@ def character_tokenize(self) -> SeriesOrIndex:
             index = self._parent.index.repeat(lengths)
             return cudf.Series(result_col, name=self._parent.name, index=index)
         elif isinstance(self._parent, cudf.BaseIndex):
-            return cudf.core.index.as_index(result_col, name=self._parent.name)
+            return cudf.Index(result_col, name=self._parent.name)
         else:
             return result_col
 

python/cudf/cudf/core/cut.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION.
+# Copyright (c) 2021-2024, NVIDIA CORPORATION.
 
 from collections import abc
 
@@ -292,7 +292,7 @@ def cut(
     )
 
     # we return a categorical index, as we don't have a Categorical method
-    categorical_index = cudf.core.index.as_index(col)
+    categorical_index = cudf.Index(col)
 
     if isinstance(orig_x, (pd.Series, cudf.Series)):
         # if we have a series input we return a series output

python/cudf/cudf/core/dataframe.py

@@ -712,7 +712,7 @@ def __init__(
                     data = data.reindex(index)
                     index = data.index
                 else:
-                    index = as_index(index)
+                    index = cudf.Index(index)
             else:
                 index = data.index
 
@@ -761,7 +761,7 @@ def __init__(
             if index is None:
                 self._index = RangeIndex(0)
             else:
-                self._index = as_index(index)
+                self._index = cudf.Index(index)
             if columns is not None:
                 rangeindex = isinstance(
                     columns, (range, pd.RangeIndex, cudf.RangeIndex)
@@ -875,7 +875,7 @@ def _init_from_series_list(self, data, columns, index):
             # When `index` is `None`, the final index of
             # resulting dataframe will be union of
             # all Series's names.
-            final_index = as_index(_get_union_of_series_names(data))
+            final_index = cudf.Index(_get_union_of_series_names(data))
         else:
             # When an `index` is passed, the final index of
             # resulting dataframe will be whatever
@@ -919,7 +919,7 @@ def _init_from_series_list(self, data, columns, index):
                         f"not match length of index ({index_length})"
                     )
 
-            final_index = as_index(index)
+            final_index = cudf.Index(index)
 
         series_lengths = list(map(len, data))
         data = numeric_normalize_types(*data)
@@ -943,7 +943,7 @@ def _init_from_series_list(self, data, columns, index):
             # Setting `final_columns` to self._index so
             # that the resulting `transpose` will be have
             # columns set to `final_columns`
-            self._index = as_index(final_columns)
+            self._index = cudf.Index(final_columns)
 
             transpose = self.T
         else:
@@ -987,9 +987,9 @@ def _init_from_list_like(self, data, index=None, columns=None):
         if index is None:
             index = RangeIndex(start=0, stop=len(data))
         else:
-            index = as_index(index)
+            index = cudf.Index(index)
 
-        self._index = as_index(index)
+        self._index = cudf.Index(index)
         # list-of-dicts case
         if len(data) > 0 and isinstance(data[0], dict):
             data = DataFrame.from_pandas(pd.DataFrame(data))
@@ -1095,7 +1095,7 @@ def _init_from_dict_like(
 
             self._index = RangeIndex(0, num_rows)
         else:
-            self._index = as_index(index)
+            self._index = cudf.Index(index)
 
         if len(data):
             self._data.multiindex = True
@@ -1410,7 +1410,7 @@ def __setitem__(self, arg, value):
                             new_columns, verify=False
                         )
                         if isinstance(value, (pd.Series, Series)):
-                            self._index = as_index(value.index)
+                            self._index = cudf.Index(value.index)
                         elif len(value) > 0:
                             self._index = RangeIndex(length)
                         return
@@ -1728,7 +1728,7 @@ def _concat(
         for cols in columns:
             table_index = None
             if 1 == first_data_column_position:
-                table_index = cudf.core.index.as_index(cols[0])
+                table_index = cudf.Index(cols[0])
             elif first_data_column_position > 1:
                 table_index = DataFrame._from_data(
                     data=dict(
@@ -1780,9 +1780,7 @@ def _concat(
             if not isinstance(out.index, MultiIndex) and isinstance(
                 out.index.dtype, cudf.CategoricalDtype
             ):
-                out = out.set_index(
-                    cudf.core.index.as_index(out.index._values)
-                )
+                out = out.set_index(cudf.Index(out.index._values))
         for name, col in out._data.items():
             out._data[name] = col._with_type_metadata(
                 tables[0]._data[name].dtype
@@ -2828,7 +2826,7 @@ def reindex(
         if columns is None:
             df = self
         else:
-            columns = as_index(columns)
+            columns = cudf.Index(columns)
             intersection = self._data.to_pandas_index().intersection(
                 columns.to_pandas()
             )
@@ -3245,7 +3243,7 @@ def _insert(self, loc, name, value, nan_as_null=None, ignore_index=True):
         if len(self) == 0:
             if isinstance(value, (pd.Series, Series)):
                 if not ignore_index:
-                    self.index = as_index(value.index)
+                    self.index = cudf.Index(value.index)
             elif (length := len(value)) > 0:
                 if num_cols != 0:
                     ca = self._data._from_columns_like_self(
@@ -5654,7 +5652,7 @@ def from_records(cls, data, index=None, columns=None, nan_as_null=False):
             }
 
         if not is_scalar(index):
-            new_index = as_index(index)
+            new_index = cudf.Index(index)
         else:
             new_index = None
 
@@ -5738,7 +5736,7 @@ def _from_arrays(cls, data, index=None, columns=None, nan_as_null=False):
             }
 
         if index is not None:
-            index = as_index(index)
+            index = cudf.Index(index)
 
         if isinstance(columns, (pd.Index, cudf.Index)):
             level_names = tuple(columns.names)
@@ -6171,7 +6169,7 @@ def count(self, axis=0, numeric_only=False):
                     for col in self._data.names
                 ]
             },
-            as_index(self._data.names),
+            cudf.Index(self._data.names),
         )
 
     _SUPPORT_AXIS_LOOKUP = {
@@ -6298,7 +6296,7 @@ def _reduce(
                         source._data.names, names=source._data.level_names
                     )
                 else:
-                    idx = as_index(source._data.names)
+                    idx = cudf.Index(source._data.names)
                 return Series._from_data({None: as_column(result)}, idx)
         elif axis == 1:
             return source._apply_cupy_method_axis_1(op, **kwargs)

python/cudf/cudf/core/dtypes.py

@@ -186,10 +186,10 @@ def categories(self) -> "cudf.core.index.Index":
         Index(['b', 'a'], dtype='object')
         """
         if self._categories is None:
-            return cudf.core.index.as_index(
+            return cudf.Index(
                 cudf.core.column.column_empty(0, dtype="object", masked=False)
             )
-        return cudf.core.index.as_index(self._categories, copy=False)
+        return cudf.Index(self._categories, copy=False)
 
     @property
     def type(self):

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

@@ -2800,15 +2800,13 @@ def keys(self):
         nkeys = len(self._key_columns)
 
         if nkeys == 0:
-            return cudf.core.index.as_index([], name=None)
+            return cudf.Index([], name=None)
         elif nkeys > 1:
             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]
-            )
+            return cudf.Index(self._key_columns[0], name=self.names[0])
 
     @property
     def values(self) -> cudf.core.frame.Frame:

python/cudf/cudf/core/index.py

@@ -1169,7 +1169,7 @@ def _concat(cls, objs):
             result = _concat_range_index(non_empties)
         else:
             data = concat_columns([o._values for o in non_empties])
-            result = as_index(data)
+            result = Index(data)
 
         names = {obj.name for obj in objs}
         if len(names) == 1:
@@ -1437,7 +1437,7 @@ def __repr__(self):
     def __getitem__(self, index):
         res = self._get_elements_from_column(index)
         if isinstance(res, ColumnBase):
-            res = as_index(res, name=self.name)
+            res = Index(res, name=self.name)
         return res
 
     @property  # type: ignore
@@ -1958,7 +1958,7 @@ def microsecond(self):
         >>> datetime_index.microsecond
         Index([0, 1, 2], dtype='int32')
         """  # noqa: E501
-        return as_index(
+        return Index(
             (
                 # Need to manually promote column to int32 because
                 # pandas-matching binop behaviour requires that this
@@ -2209,7 +2209,7 @@ def _get_dt_field(self, field):
             mask=out_column.base_mask,
             offset=out_column.offset,
         )
-        return as_index(out_column, name=self.name)
+        return Index(out_column, name=self.name)
 
     def _is_boolean(self):
         return False
@@ -2522,29 +2522,23 @@ def days(self):
         Number of days for each element.
         """
         # Need to specifically return `int64` to avoid overflow.
-        return as_index(
-            arbitrary=self._values.days, name=self.name, dtype="int64"
-        )
+        return Index(self._values.days, name=self.name, dtype="int64")
 
     @property  # type: ignore
     @_cudf_nvtx_annotate
     def seconds(self):
         """
         Number of seconds (>= 0 and less than 1 day) for each element.
         """
-        return as_index(
-            arbitrary=self._values.seconds, name=self.name, dtype="int32"
-        )
+        return Index(self._values.seconds, name=self.name, dtype="int32")
 
     @property  # type: ignore
     @_cudf_nvtx_annotate
     def microseconds(self):
         """
         Number of microseconds (>= 0 and less than 1 second) for each element.
         """
-        return as_index(
-            arbitrary=self._values.microseconds, name=self.name, dtype="int32"
-        )
+        return Index(self._values.microseconds, name=self.name, dtype="int32")
 
     @property  # type: ignore
     @_cudf_nvtx_annotate
@@ -2553,9 +2547,7 @@ def nanoseconds(self):
         Number of nanoseconds (>= 0 and less than 1 microsecond) for each
         element.
         """
-        return as_index(
-            arbitrary=self._values.nanoseconds, name=self.name, dtype="int32"
-        )
+        return Index(self._values.nanoseconds, name=self.name, dtype="int32")
 
     @property  # type: ignore
     @_cudf_nvtx_annotate
@@ -2693,7 +2685,7 @@ def codes(self):
         """
         The category codes of this categorical.
         """
-        return as_index(self._values.codes)
+        return Index(self._values.codes)
 
     @property  # type: ignore
     @_cudf_nvtx_annotate
@@ -3137,15 +3129,15 @@ def _concat_range_index(indexes: List[RangeIndex]) -> BaseIndex:
         elif step is None:
             # First non-empty index had only one element
             if obj.start == start:
-                result = as_index(concat_columns([x._values for x in indexes]))
+                result = Index(concat_columns([x._values for x in indexes]))
                 return result
             step = obj.start - start
 
         non_consecutive = (step != obj.step and len(obj) > 1) or (
             next_ is not None and obj.start != next_
         )
         if non_consecutive:
-            result = as_index(concat_columns([x._values for x in indexes]))
+            result = Index(concat_columns([x._values for x in indexes]))
             return result
         if step is not None:
             next_ = obj[-1] + step

python/cudf/cudf/core/indexed_frame.py

@@ -3638,7 +3638,7 @@ def _align_to_index(
         sort: bool = True,
         allow_non_unique: bool = False,
     ) -> Self:
-        index = cudf.core.index.as_index(index)
+        index = cudf.Index(index)
 
         if self.index.equals(index):
             return self
@@ -3713,7 +3713,7 @@ def _reindex(
                 raise ValueError(
                     "cannot reindex on an axis with duplicate labels"
                 )
-            index = cudf.core.index.as_index(
+            index = cudf.Index(
                 index, name=getattr(index, "name", self.index.name)
             )
 

python/cudf/cudf/core/multiindex.py

@@ -30,7 +30,6 @@
     BaseIndex,
     _get_indexer_basic,
     _lexsorted_equal_range,
-    as_index,
 )
 from cudf.core.join._join_helpers import _match_join_keys
 from cudf.utils.dtypes import is_column_like
@@ -824,7 +823,7 @@ def _index_and_downcast(self, result, index, index_key):
             # it into an Index and name the final index values according
             # to that column's name.
             *_, last_column = index._data.columns
-            out_index = as_index(last_column)
+            out_index = cudf.Index(last_column)
             out_index.name = index.names[-1]
             index = out_index
         elif out_index._num_columns > 1:
@@ -1082,7 +1081,9 @@ def get_level_values(self, level):
                 raise KeyError(f"Level not found: '{level}'")
         else:
             level_idx = colnames.index(level)
-        level_values = as_index(self._data[level], name=self.names[level_idx])
+        level_values = cudf.Index(
+            self._data[level], name=self.names[level_idx]
+        )
         return level_values
 
     def _is_numeric(self):