BUG: Fix na_position type in IndexEngine (#61062)

* Modify an existing test to cover the issue with na_pos > 128.

* Change na_position type from int8_t and int64_t consistently to Py_ssize_t.

* Add What's New entry.

* Sort whatsnew entries alphabetically

* Improve the whatsnew entry.

* Move whatsnew entry from v2.3.0.rst to v3.0.0.rst.

* Update doc/source/whatsnew/v3.0.0.rst

Co-authored-by: Matthew Roeschke <[email protected]>

* Undo remove '-'.

* Sort whatsnew entries alphabetically.

---------

Co-authored-by: avm19 <[email protected]>
Co-authored-by: Matthew Roeschke <[email protected]>

Author: 3 people

doc/source/whatsnew/v3.0.0.rst

@@ -696,6 +696,7 @@ Indexing
 - Bug in :meth:`DataFrame.__getitem__` returning modified columns when called with ``slice`` in Python 3.12 (:issue:`57500`)
 - Bug in :meth:`DataFrame.from_records` throwing a ``ValueError`` when passed an empty list in ``index`` (:issue:`58594`)
 - Bug in :meth:`DataFrame.loc` with inconsistent behavior of loc-set with 2 given indexes to Series (:issue:`59933`)
+- Bug in :meth:`Index.get_indexer` and similar methods when ``NaN`` is located at or after position 128 (:issue:`58924`)
 - Bug in :meth:`MultiIndex.insert` when a new value inserted to a datetime-like level gets cast to ``NaT`` and fails indexing (:issue:`60388`)
 - Bug in printing :attr:`Index.names` and :attr:`MultiIndex.levels` would not escape single quotes (:issue:`60190`)
 

pandas/_libs/hashtable.pxd

@@ -41,7 +41,7 @@ cdef class HashTable:
 
 cdef class UInt64HashTable(HashTable):
     cdef kh_uint64_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, uint64_t val)
@@ -51,7 +51,7 @@ cdef class UInt64HashTable(HashTable):
 
 cdef class Int64HashTable(HashTable):
     cdef kh_int64_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, int64_t val)
@@ -61,7 +61,7 @@ cdef class Int64HashTable(HashTable):
 
 cdef class UInt32HashTable(HashTable):
     cdef kh_uint32_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, uint32_t val)
@@ -71,7 +71,7 @@ cdef class UInt32HashTable(HashTable):
 
 cdef class Int32HashTable(HashTable):
     cdef kh_int32_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, int32_t val)
@@ -81,7 +81,7 @@ cdef class Int32HashTable(HashTable):
 
 cdef class UInt16HashTable(HashTable):
     cdef kh_uint16_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, uint16_t val)
@@ -91,7 +91,7 @@ cdef class UInt16HashTable(HashTable):
 
 cdef class Int16HashTable(HashTable):
     cdef kh_int16_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, int16_t val)
@@ -101,7 +101,7 @@ cdef class Int16HashTable(HashTable):
 
 cdef class UInt8HashTable(HashTable):
     cdef kh_uint8_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, uint8_t val)
@@ -111,7 +111,7 @@ cdef class UInt8HashTable(HashTable):
 
 cdef class Int8HashTable(HashTable):
     cdef kh_int8_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, int8_t val)
@@ -121,7 +121,7 @@ cdef class Int8HashTable(HashTable):
 
 cdef class Float64HashTable(HashTable):
     cdef kh_float64_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, float64_t val)
@@ -131,7 +131,7 @@ cdef class Float64HashTable(HashTable):
 
 cdef class Float32HashTable(HashTable):
     cdef kh_float32_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, float32_t val)
@@ -141,7 +141,7 @@ cdef class Float32HashTable(HashTable):
 
 cdef class Complex64HashTable(HashTable):
     cdef kh_complex64_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, complex64_t val)
@@ -151,7 +151,7 @@ cdef class Complex64HashTable(HashTable):
 
 cdef class Complex128HashTable(HashTable):
     cdef kh_complex128_t *table
-    cdef int64_t na_position
+    cdef Py_ssize_t na_position
     cdef bint uses_mask
 
     cpdef get_item(self, complex128_t val)

pandas/_libs/hashtable_class_helper.pxi.in

@@ -535,7 +535,7 @@ cdef class {{name}}HashTable(HashTable):
             int ret = 0
             {{c_type}} val
             khiter_t k
-            int8_t na_position = self.na_position
+            Py_ssize_t na_position = self.na_position
 
         if self.uses_mask and mask is None:
             raise NotImplementedError  # pragma: no cover
@@ -567,7 +567,7 @@ cdef class {{name}}HashTable(HashTable):
             Int64Vector self_locs = Int64Vector()
             Int64VectorData *l
             Int64VectorData *sl
-            int8_t na_position = self.na_position
+            Py_ssize_t na_position = self.na_position
 
         l = &locs.data
         sl = &self_locs.data
@@ -609,7 +609,7 @@ cdef class {{name}}HashTable(HashTable):
             {{c_type}} val
             khiter_t k
             intp_t[::1] locs = np.empty(n, dtype=np.intp)
-            int8_t na_position = self.na_position
+            Py_ssize_t na_position = self.na_position
 
         if self.uses_mask and mask is None:
             raise NotImplementedError  # pragma: no cover

pandas/tests/libs/test_hashtable.py

@@ -149,18 +149,19 @@ def test_map_locations(self, table_type, dtype, writable):
     def test_map_locations_mask(self, table_type, dtype, writable):
         if table_type == ht.PyObjectHashTable:
             pytest.skip("Mask not supported for object")
-        N = 3
+        N = 129  # must be > 128 to test GH#58924
         table = table_type(uses_mask=True)
         keys = (np.arange(N) + N).astype(dtype)
         keys.flags.writeable = writable
-        table.map_locations(keys, np.array([False, False, True]))
+        mask = np.concatenate([np.repeat(False, N - 1), [True]], axis=0)
+        table.map_locations(keys, mask)
         for i in range(N - 1):
             assert table.get_item(keys[i]) == i
 
         with pytest.raises(KeyError, match=re.escape(str(keys[N - 1]))):
             table.get_item(keys[N - 1])
 
-        assert table.get_na() == 2
+        assert table.get_na() == N - 1
 
     def test_lookup(self, table_type, dtype, writable):
         N = 3