From 610cb56f85bbebe16b6dfc9b2c4fcecc2506f186 Mon Sep 17 00:00:00 2001
From: Patrick Hoefler <61934744+phofl@users.noreply.github.com>
Date: Thu, 31 Aug 2023 11:34:06 +0200
Subject: [PATCH 1/3] REGR: Merge raising when left merging on arrow string
 index

---
 doc/source/whatsnew/v2.1.1.rst           |  2 +-
 pandas/core/reshape/merge.py             |  8 ++++++--
 pandas/tests/reshape/merge/test_merge.py | 11 +++++++++++
 3 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/doc/source/whatsnew/v2.1.1.rst b/doc/source/whatsnew/v2.1.1.rst
index 66f6d59d08cb2..53e3325684c64 100644
--- a/doc/source/whatsnew/v2.1.1.rst
+++ b/doc/source/whatsnew/v2.1.1.rst
@@ -13,7 +13,7 @@ including other versions of pandas.
 
 Fixed regressions
 ~~~~~~~~~~~~~~~~~
--
+- Fixed regression in :func:`merge` when merging over a PyArrow string index (:issue:`54894`)
 
 .. ---------------------------------------------------------------------------
 .. _whatsnew_211.bug_fixes:
diff --git a/pandas/core/reshape/merge.py b/pandas/core/reshape/merge.py
index 13bc1008698b2..71e91b574a9e7 100644
--- a/pandas/core/reshape/merge.py
+++ b/pandas/core/reshape/merge.py
@@ -2436,8 +2436,12 @@ def _factorize_keys(
             length = len(dc.dictionary)
 
             llab, rlab, count = (
-                pc.fill_null(dc.indices[slice(len_lk)], length).to_numpy(),
-                pc.fill_null(dc.indices[slice(len_lk, None)], length).to_numpy(),
+                pc.fill_null(dc.indices[slice(len_lk)], length)
+                .to_numpy()
+                .astype(np.intp),
+                pc.fill_null(dc.indices[slice(len_lk, None)], length)
+                .to_numpy()
+                .astype(np.intp),
                 len(dc.dictionary),
             )
             if how == "right":
diff --git a/pandas/tests/reshape/merge/test_merge.py b/pandas/tests/reshape/merge/test_merge.py
index 02d7e2059e8e1..7d923fd59fc3f 100644
--- a/pandas/tests/reshape/merge/test_merge.py
+++ b/pandas/tests/reshape/merge/test_merge.py
@@ -2947,3 +2947,14 @@ def test_merge_ea_int_and_float_numpy():
 
     result = df2.merge(df1)
     tm.assert_frame_equal(result, expected.astype("float64"))
+
+
+def test_merge_arrow_string_index():
+    # GH#54894
+    left = DataFrame({"a": ["a", "b"]}, dtype="string[pyarrow]")
+    right = DataFrame({"b": 1}, index=Index(["a", "c"], dtype="string[pyarrow]"))
+    result = left.merge(right, left_on="a", right_index=True, how="left")
+    expected = DataFrame(
+        {"a": Series(["a", "b"], dtype="string[pyarrow]"), "b": [1, np.nan]}
+    )
+    tm.assert_frame_equal(result, expected)

From 5b2b7ab96391d1d24050d34fda13872da2f231b7 Mon Sep 17 00:00:00 2001
From: Patrick Hoefler <61934744+phofl@users.noreply.github.com>
Date: Thu, 31 Aug 2023 12:47:21 +0200
Subject: [PATCH 2/3] Fix

---
 pandas/tests/reshape/merge/test_merge.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/pandas/tests/reshape/merge/test_merge.py b/pandas/tests/reshape/merge/test_merge.py
index 7d923fd59fc3f..9cada6964c094 100644
--- a/pandas/tests/reshape/merge/test_merge.py
+++ b/pandas/tests/reshape/merge/test_merge.py
@@ -2951,6 +2951,7 @@ def test_merge_ea_int_and_float_numpy():
 
 def test_merge_arrow_string_index():
     # GH#54894
+    pytest.importorskip("pyarrow")
     left = DataFrame({"a": ["a", "b"]}, dtype="string[pyarrow]")
     right = DataFrame({"b": 1}, index=Index(["a", "c"], dtype="string[pyarrow]"))
     result = left.merge(right, left_on="a", right_index=True, how="left")

From 4f1a996422b3c27dc04097a935a84029e5a56465 Mon Sep 17 00:00:00 2001
From: Patrick Hoefler <61934744+phofl@users.noreply.github.com>
Date: Thu, 31 Aug 2023 22:59:24 +0200
Subject: [PATCH 3/3] Add copy false

---
 pandas/core/reshape/merge.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/pandas/core/reshape/merge.py b/pandas/core/reshape/merge.py
index 71e91b574a9e7..6098ff2698e2e 100644
--- a/pandas/core/reshape/merge.py
+++ b/pandas/core/reshape/merge.py
@@ -2438,10 +2438,10 @@ def _factorize_keys(
             llab, rlab, count = (
                 pc.fill_null(dc.indices[slice(len_lk)], length)
                 .to_numpy()
-                .astype(np.intp),
+                .astype(np.intp, copy=False),
                 pc.fill_null(dc.indices[slice(len_lk, None)], length)
                 .to_numpy()
-                .astype(np.intp),
+                .astype(np.intp, copy=False),
                 len(dc.dictionary),
             )
             if how == "right":