get_labels_groupby is only used with Int64HashTable, thus define it only for int64. What should happen for other types (encoding nans as negative values is not possible with all types) depends on how it is used (but it is not used right now)

Author: realead

pandas/_libs/hashtable_class_helper.pxi.in

@@ -681,6 +681,7 @@ cdef class {{name}}HashTable(HashTable):
                                  ignore_na=True, return_inverse=True)
         return labels
 
+    {{if dtype == 'int64'}}
     @cython.boundscheck(False)
     def get_labels_groupby(self, const {{dtype}}_t[:] values):
         cdef:
@@ -701,16 +702,9 @@ cdef class {{name}}HashTable(HashTable):
                 val = {{to_c_type}}(values[i])
 
                 # specific for groupby
-                {{if dtype == 'complex64' or dtype== 'complex128'}}
-                # TODO: what should be done here?
-                if val.real < 0:
-                    labels[i] = -1
-                    continue
-                {{elif dtype != 'uint64'}}
                 if val < 0:
                     labels[i] = -1
                     continue
-                {{endif}}
 
                 k = kh_get_{{dtype}}(self.table, val)
                 if k != self.table.n_buckets:
@@ -730,6 +724,7 @@ cdef class {{name}}HashTable(HashTable):
         arr_uniques = uniques.to_array()
 
         return np.asarray(labels), arr_uniques
+    {{endif}}
 
 {{endfor}}
 

pandas/tests/libs/test_hashtable.py

@@ -156,6 +156,16 @@ def test_tracemalloc_for_empty(self, table_type, dtype):
             assert get_allocated_khash_memory() == 0
 
 
+def test_get_labels_groupby_for_Int64():
+    table = ht.Int64HashTable()
+    vals = np.array([1, 2, -1, 2, 1, -1], dtype=np.int64)
+    arr, unique = table.get_labels_groupby(vals)
+    expected_arr = np.array([0, 1, -1, 1, 0, -1], dtype=np.int64)
+    expected_unique = np.array([1, 2], dtype=np.int64)
+    tm.assert_numpy_array_equal(arr, expected_arr)
+    tm.assert_numpy_array_equal(unique, expected_unique)
+
+
 def test_tracemalloc_works_for_StringHashTable():
     N = 1000
     keys = np.arange(N).astype(np.compat.unicode).astype(np.object_)