PERF: Return RangeIndex from RangeIndex.join when possible (#57651)

* PERF: Return RangeIndex from RangeIndex.join when possible

* whatsnew number

* Fix indexer

Author: mroeschke

doc/source/whatsnew/v3.0.0.rst

@@ -255,6 +255,7 @@ Performance improvements
 - Performance improvement in :meth:`MultiIndex.equals` for equal length indexes (:issue:`56990`)
 - Performance improvement in :meth:`RangeIndex.__getitem__` with a boolean mask returning a :class:`RangeIndex` instead of a :class:`Index` when possible. (:issue:`57588`)
 - Performance improvement in :meth:`RangeIndex.append` when appending the same index (:issue:`57252`)
+- Performance improvement in :meth:`RangeIndex.join` returning a :class:`RangeIndex` instead of a :class:`Index` when possible. (:issue:`57651`)
 - Performance improvement in :meth:`RangeIndex.take` returning a :class:`RangeIndex` instead of a :class:`Index` when possible. (:issue:`57445`)
 - Performance improvement in ``DataFrameGroupBy.__len__`` and ``SeriesGroupBy.__len__`` (:issue:`57595`)
 - Performance improvement in indexing operations for string dtypes (:issue:`56997`)

pandas/core/indexes/base.py

@@ -4588,7 +4588,6 @@ def _get_leaf_sorter(
         )
         return join_index, left_indexer, right_indexer
 
-    @final
     def _join_monotonic(
         self, other: Index, how: JoinHow = "left"
     ) -> tuple[Index, npt.NDArray[np.intp] | None, npt.NDArray[np.intp] | None]:

pandas/core/indexes/range.py

@@ -55,6 +55,7 @@
     from pandas._typing import (
         Axis,
         Dtype,
+        JoinHow,
         NaPosition,
         Self,
         npt,
@@ -890,6 +891,41 @@ def symmetric_difference(
             result = result.rename(result_name)
         return result
 
+    def _join_monotonic(
+        self, other: Index, how: JoinHow = "left"
+    ) -> tuple[Index, npt.NDArray[np.intp] | None, npt.NDArray[np.intp] | None]:
+        # This currently only gets called for the monotonic increasing case
+        if not isinstance(other, type(self)):
+            maybe_ri = self._shallow_copy(other._values)
+            if not isinstance(maybe_ri, type(self)):
+                return super()._join_monotonic(other, how=how)
+            other = maybe_ri
+
+        if self.equals(other):
+            ret_index = other if how == "right" else self
+            return ret_index, None, None
+
+        if how == "left":
+            join_index = self
+            lidx = None
+            ridx = other.get_indexer(join_index)
+        elif how == "right":
+            join_index = other
+            lidx = self.get_indexer(join_index)
+            ridx = None
+        elif how == "inner":
+            join_index = self.intersection(other)
+            lidx = self.get_indexer(join_index)
+            ridx = other.get_indexer(join_index)
+        elif how == "outer":
+            join_index = self.union(other)
+            lidx = self.get_indexer(join_index)
+            ridx = other.get_indexer(join_index)
+
+        lidx = None if lidx is None else ensure_platform_int(lidx)
+        ridx = None if ridx is None else ensure_platform_int(ridx)
+        return join_index, lidx, ridx
+
     # --------------------------------------------------------------------
 
     # error: Return type "Index" of "delete" incompatible with return type

pandas/tests/indexes/ranges/test_join.py

@@ -1,4 +1,5 @@
 import numpy as np
+import pytest
 
 from pandas import (
     Index,
@@ -175,3 +176,54 @@ def test_join_self(self, join_type):
         index = RangeIndex(start=0, stop=20, step=2)
         joined = index.join(index, how=join_type)
         assert index is joined
+
+
+@pytest.mark.parametrize(
+    "left, right, expected, expected_lidx, expected_ridx, how",
+    [
+        [RangeIndex(2), RangeIndex(3), RangeIndex(2), None, [0, 1], "left"],
+        [RangeIndex(2), RangeIndex(2), RangeIndex(2), None, None, "left"],
+        [RangeIndex(2), RangeIndex(20, 22), RangeIndex(2), None, [-1, -1], "left"],
+        [RangeIndex(2), RangeIndex(3), RangeIndex(3), [0, 1, -1], None, "right"],
+        [RangeIndex(2), RangeIndex(2), RangeIndex(2), None, None, "right"],
+        [
+            RangeIndex(2),
+            RangeIndex(20, 22),
+            RangeIndex(20, 22),
+            [-1, -1],
+            None,
+            "right",
+        ],
+        [RangeIndex(2), RangeIndex(3), RangeIndex(2), [0, 1], [0, 1], "inner"],
+        [RangeIndex(2), RangeIndex(2), RangeIndex(2), None, None, "inner"],
+        [RangeIndex(2), RangeIndex(1, 3), RangeIndex(1, 2), [1], [0], "inner"],
+        [RangeIndex(2), RangeIndex(3), RangeIndex(3), [0, 1, -1], [0, 1, 2], "outer"],
+        [RangeIndex(2), RangeIndex(2), RangeIndex(2), None, None, "outer"],
+        [
+            RangeIndex(2),
+            RangeIndex(2, 4),
+            RangeIndex(4),
+            [0, 1, -1, -1],
+            [-1, -1, 0, 1],
+            "outer",
+        ],
+    ],
+)
+@pytest.mark.parametrize("right_type", [RangeIndex, lambda x: Index(list(x))])
+def test_join_preserves_rangeindex(
+    left, right, expected, expected_lidx, expected_ridx, how, right_type
+):
+    result, lidx, ridx = left.join(right_type(right), how=how, return_indexers=True)
+    tm.assert_index_equal(result, expected, exact=True)
+
+    if expected_lidx is None:
+        assert lidx is expected_lidx
+    else:
+        exp_lidx = np.array(expected_lidx, dtype=np.intp)
+        tm.assert_numpy_array_equal(lidx, exp_lidx)
+
+    if expected_ridx is None:
+        assert ridx is expected_ridx
+    else:
+        exp_ridx = np.array(expected_ridx, dtype=np.intp)
+        tm.assert_numpy_array_equal(ridx, exp_ridx)