BUG: groupby.agg returns incorrect results for uint64 cols (#26310)

Author: Matias Heikkilä

doc/source/whatsnew/v0.25.0.rst

@@ -258,6 +258,7 @@ Performance Improvements
 Bug Fixes
 ~~~~~~~~~
 
+- Bug where groupby.agg (first, last, min, etc...) returns incorrect results for uint64 columns. (:issue:`26310`)
 
 
 Categorical

pandas/core/dtypes/common.py

@@ -90,9 +90,11 @@ def ensure_categorical(arr):
 def ensure_int64_or_float64(arr, copy=False):
     """
     Ensure that an dtype array of some integer dtype
-    has an int64 dtype if possible
+    has an int64 dtype if possible.
     If it's not possible, potentially because of overflow,
     convert the array to float64 instead.
+    If the array is explicitly of type uint64 the type
+    will remain unchanged.
 
     Parameters
     ----------
@@ -107,11 +109,16 @@ def ensure_int64_or_float64(arr, copy=False):
     out_arr : The input array cast as int64 if
               possible without overflow.
               Otherwise the input array cast to float64.
+              If the array is explicitly of type uint64 the type
+              will remain unchanged.
     """
     try:
         return arr.astype('int64', copy=copy, casting='safe')
     except TypeError:
-        return arr.astype('float64', copy=copy)
+        try:
+            return arr.astype('uint64', copy=copy, casting='safe')
+        except TypeError:
+            return arr.astype('float64', copy=copy)
 
 
 def classes(*klasses):

pandas/tests/groupby/aggregate/test_aggregate.py

@@ -313,3 +313,14 @@ def test_order_aggregate_multiple_funcs():
     expected = pd.Index(['sum', 'max', 'mean', 'ohlc', 'min'])
 
     tm.assert_index_equal(result, expected)
+
+
+def test_uint64_type_handling():
+    # GH 26310
+    df1 = pd.DataFrame({'x': 6903052872240755750, 'y': [1, 2]})
+    df1.groupby('y').agg({'x': 'first'})
+    df2 = df1
+    df2.x = df2.x.astype(np.uint64)
+    df2.groupby('y').agg({'x': 'first'})
+
+    tm.assert_frame_equal(df1, df2)