COMPAT: avoid warnings from numeric/string-like comparisons

Author: jreback

pandas/core/common.py

@@ -232,7 +232,7 @@ def _isnull_ndarraylike(obj):
     values = getattr(obj, 'values', obj)
     dtype = values.dtype
 
-    if dtype.kind in ('O', 'S', 'U'):
+    if is_string_dtype(dtype):
         if is_categorical_dtype(values):
             from pandas import Categorical
             if not isinstance(values, Categorical):
@@ -243,7 +243,7 @@ def _isnull_ndarraylike(obj):
             # Working around NumPy ticket 1542
             shape = values.shape
 
-            if dtype.kind in ('S', 'U'):
+            if is_string_like_dtype(dtype):
                 result = np.zeros(values.shape, dtype=bool)
             else:
                 result = np.empty(shape, dtype=bool)
@@ -267,11 +267,11 @@ def _isnull_ndarraylike_old(obj):
     values = getattr(obj, 'values', obj)
     dtype = values.dtype
 
-    if dtype.kind in ('O', 'S', 'U'):
+    if is_string_dtype(dtype):
         # Working around NumPy ticket 1542
         shape = values.shape
 
-        if values.dtype.kind in ('S', 'U'):
+        if is_string_like_dtype(dtype):
             result = np.zeros(values.shape, dtype=bool)
         else:
             result = np.empty(shape, dtype=bool)
@@ -2208,13 +2208,17 @@ def is_numeric_v_string_like(a, b):
 
     is_a_numeric_array = is_a_array and is_numeric_dtype(a)
     is_b_numeric_array = is_b_array and is_numeric_dtype(b)
+    is_a_string_array = is_a_array and is_string_like_dtype(a)
+    is_b_string_array = is_b_array and is_string_like_dtype(b)
 
     is_a_scalar_string_like = not is_a_array and is_string_like(a)
     is_b_scalar_string_like = not is_b_array and is_string_like(b)
 
     return (
         is_a_numeric_array and is_b_scalar_string_like) or (
-        is_b_numeric_array and is_a_scalar_string_like
+        is_b_numeric_array and is_a_scalar_string_like) or (
+        is_a_numeric_array and is_b_string_array) or (
+        is_b_numeric_array and is_a_string_array
         )
 
 def is_datetimelike_v_numeric(a, b):
@@ -2257,6 +2261,15 @@ def is_numeric_dtype(arr_or_dtype):
             and not issubclass(tipo, (np.datetime64, np.timedelta64)))
 
 
+def is_string_dtype(arr_or_dtype):
+    dtype = _get_dtype(arr_or_dtype)
+    return dtype.kind in ('O', 'S', 'U')
+
+def is_string_like_dtype(arr_or_dtype):
+    # exclude object as its a mixed dtype
+    dtype = _get_dtype(arr_or_dtype)
+    return dtype.kind in ('S', 'U')
+
 def is_float_dtype(arr_or_dtype):
     tipo = _get_dtype_type(arr_or_dtype)
     return issubclass(tipo, np.floating)

pandas/core/internals.py

@@ -4419,11 +4419,14 @@ def _putmask_smart(v, m, n):
     try:
         nn = n[m]
         nn_at = nn.astype(v.dtype)
-        comp = (nn == nn_at)
-        if is_list_like(comp) and comp.all():
-            nv = v.copy()
-            nv[m] = nn_at
-            return nv
+
+        # avoid invalid dtype comparisons
+        if not is_numeric_v_string_like(nn, nn_at):
+            comp = (nn == nn_at)
+            if is_list_like(comp) and comp.all():
+                nv = v.copy()
+                nv[m] = nn_at
+                return nv
     except (ValueError, IndexError, TypeError):
         pass
 

pandas/tests/test_series.py

@@ -5112,26 +5112,49 @@ def test_cov(self):
 
     def test_copy(self):
 
-        for deep in [False, True]:
+        for deep in [None, False, True]:
             s = Series(np.arange(10),dtype='float64')
-            s2 = s.copy(deep=deep)
+
+            # default deep is True
+            if deep is None:
+                s2 = s.copy()
+            else:
+                s2 = s.copy(deep=deep)
+
             s2[::2] = np.NaN
 
-            # Did not modify original Series
-            self.assertTrue(np.isnan(s2[0]))
-            self.assertFalse(np.isnan(s[0]))
+            if deep is None or deep is True:
+                # Did not modify original Series
+                self.assertTrue(np.isnan(s2[0]))
+                self.assertFalse(np.isnan(s[0]))
+            else:
+
+                # we DID modify the original Series
+                self.assertTrue(np.isnan(s2[0]))
+                self.assertTrue(np.isnan(s[0]))
 
         # GH 11794
         # copy of tz-aware
         expected = Series([Timestamp('2012/01/01', tz='UTC')])
         expected2 = Series([Timestamp('1999/01/01', tz='UTC')])
 
-        for deep in [False, True]:
+        for deep in [None, False, True]:
             s = Series([Timestamp('2012/01/01', tz='UTC')])
-            s2 = s.copy()
+
+            if deep is None:
+                s2 = s.copy()
+            else:
+                s2 = s.copy(deep=deep)
+
             s2[0] = pd.Timestamp('1999/01/01', tz='UTC')
-            assert_series_equal(s, expected)
-            assert_series_equal(s2, expected2)
+
+            # default deep is True
+            if deep is None or deep is True:
+                assert_series_equal(s, expected)
+                assert_series_equal(s2, expected2)
+            else:
+                assert_series_equal(s, expected2)
+                assert_series_equal(s2, expected2)
 
     def test_count(self):
         self.assertEqual(self.ts.count(), len(self.ts))