CLN: replace _interleave_dtype with _find_common_type

xref pandas-dev#15736   xref pandas-dev#12780

Author: Jeff Reback <jeff@reback.net>

Closes pandas-dev#15765 from jreback/common_types and squashes the following commits:

d472646 [Jeff Reback] try removing restriction on windows
8d07cae [Jeff Reback] CLN: replace _interleave_dtype with _find_common_type

doc/source/whatsnew/v0.20.0.txt

@@ -826,7 +826,7 @@ Bug Fixes
 
 
 - Bug in the display of ``.info()`` where a qualifier (+) would always be displayed with a ``MultiIndex`` that contains only non-strings (:issue:`15245`)
-- Bug in ``.replace()`` may result in incorrect dtypes. (:issue:`12747`)
+- Bug in ``.replace()`` may result in incorrect dtypes. (:issue:`12747`, :issue:`15765`)
 
 - Bug in ``.asfreq()``, where frequency was not set for empty ``Series`` (:issue:`14320`)
 

pandas/core/internals.py

@@ -9,7 +9,8 @@
 
 from pandas.core.base import PandasObject
 
-from pandas.types.dtypes import DatetimeTZDtype, CategoricalDtype
+from pandas.types.dtypes import (ExtensionDtype, DatetimeTZDtype,
+                                 CategoricalDtype)
 from pandas.types.common import (_TD_DTYPE, _NS_DTYPE,
                                  _ensure_int64, _ensure_platform_int,
                                  is_integer,
@@ -4496,55 +4497,13 @@ def _interleaved_dtype(blocks):
     if not len(blocks):
         return None
 
-    counts = defaultdict(list)
-    for x in blocks:
-        counts[type(x)].append(x)
-
-    have_int = len(counts[IntBlock]) > 0
-    have_bool = len(counts[BoolBlock]) > 0
-    have_object = len(counts[ObjectBlock]) > 0
-    have_float = len(counts[FloatBlock]) > 0
-    have_complex = len(counts[ComplexBlock]) > 0
-    have_dt64 = len(counts[DatetimeBlock]) > 0
-    have_dt64_tz = len(counts[DatetimeTZBlock]) > 0
-    have_td64 = len(counts[TimeDeltaBlock]) > 0
-    have_cat = len(counts[CategoricalBlock]) > 0
-    # TODO: have_sparse is not used
-    have_sparse = len(counts[SparseBlock]) > 0  # noqa
-    have_numeric = have_float or have_complex or have_int
-    has_non_numeric = have_dt64 or have_dt64_tz or have_td64 or have_cat
-
-    if (have_object or
-        (have_bool and
-         (have_numeric or have_dt64 or have_dt64_tz or have_td64)) or
-        (have_numeric and has_non_numeric) or have_cat or have_dt64 or
-            have_dt64_tz or have_td64):
-        return np.dtype(object)
-    elif have_bool:
-        return np.dtype(bool)
-    elif have_int and not have_float and not have_complex:
-        # if we are mixing unsigned and signed, then return
-        # the next biggest int type (if we can)
-        lcd = _find_common_type([b.dtype for b in counts[IntBlock]])
-        kinds = set([i.dtype.kind for i in counts[IntBlock]])
-        if len(kinds) == 1:
-            return lcd
-
-        if lcd == 'uint64' or lcd == 'int64':
-            return np.dtype('int64')
-
-        # return 1 bigger on the itemsize if unsinged
-        if lcd.kind == 'u':
-            return np.dtype('int%s' % (lcd.itemsize * 8 * 2))
-        return lcd
-
-    elif have_int and have_float and not have_complex:
-        return np.dtype('float64')
-    elif have_complex:
-        return np.dtype('c16')
-    else:
-        introspection_blks = counts[FloatBlock] + counts[SparseBlock]
-        return _find_common_type([b.dtype for b in introspection_blks])
+    dtype = _find_common_type([b.dtype for b in blocks])
+
+    # only numpy compat
+    if isinstance(dtype, ExtensionDtype):
+        dtype = np.object
+
+    return dtype
 
 
 def _consolidate(blocks):

pandas/tests/indexing/test_coercion.py

@@ -1183,19 +1183,9 @@ def _assert_replace_conversion(self, from_key, to_key, how):
 
         result = obj.replace(replacer)
 
-        # buggy on windows for bool/int64
-        if (from_key == 'bool' and
-                to_key == 'int64' and
-                tm.is_platform_windows()):
-            pytest.skip("windows platform buggy: {0} -> {1}".format
-                        (from_key, to_key))
-
-        if ((from_key == 'float64' and to_key in ('bool', 'int64')) or
+        if ((from_key == 'float64' and to_key in ('int64')) or
             (from_key == 'complex128' and
-             to_key in ('bool', 'int64', 'float64')) or
-
-            # GH12747 The result must be int?
-           (from_key == 'int64' and to_key in ('bool'))):
+             to_key in ('int64', 'float64'))):
 
             # buggy on 32-bit
             if tm.is_platform_32bit():

pandas/tests/series/test_replace.py

@@ -152,8 +152,8 @@ def check_replace(to_rep, val, expected):
         tr, v = [3, 4], [3.5, pd.Timestamp('20130101')]
         check_replace(tr, v, e)
 
-        # casts to float
-        e = pd.Series([0, 1, 2, 3.5, 1])
+        # casts to object
+        e = pd.Series([0, 1, 2, 3.5, True], dtype='object')
         tr, v = [3, 4], [3.5, True]
         check_replace(tr, v, e)
 

pandas/tests/types/test_cast.py

@@ -238,6 +238,20 @@ def test_numpy_dtypes(self):
             ((np.object, np.float32), np.object),
             ((np.object, np.int16), np.object),
 
+            # bool with int
+            ((np.dtype('bool'), np.int64), np.object),
+            ((np.dtype('bool'), np.int32), np.object),
+            ((np.dtype('bool'), np.int16), np.object),
+            ((np.dtype('bool'), np.int8), np.object),
+            ((np.dtype('bool'), np.uint64), np.object),
+            ((np.dtype('bool'), np.uint32), np.object),
+            ((np.dtype('bool'), np.uint16), np.object),
+            ((np.dtype('bool'), np.uint8), np.object),
+
+            # bool with float
+            ((np.dtype('bool'), np.float64), np.object),
+            ((np.dtype('bool'), np.float32), np.object),
+
             ((np.dtype('datetime64[ns]'), np.dtype('datetime64[ns]')),
              np.dtype('datetime64[ns]')),
             ((np.dtype('timedelta64[ns]'), np.dtype('timedelta64[ns]')),

pandas/types/cast.py

@@ -892,12 +892,28 @@ def _possibly_cast_to_datetime(value, dtype, errors='raise'):
 
 
 def _find_common_type(types):
-    """Find a common data type among the given dtypes."""
+    """
+    Find a common data type among the given dtypes.
+
+    Parameters
+    ----------
+    types : list of dtypes
+
+    Returns
+    -------
+    pandas extension or numpy dtype
+
+    See Also
+    --------
+    numpy.find_common_type
+
+    """
 
     if len(types) == 0:
         raise ValueError('no types given')
 
     first = types[0]
+
     # workaround for find_common_type([np.dtype('datetime64[ns]')] * 2)
     # => object
     if all(is_dtype_equal(first, t) for t in types[1:]):
@@ -912,4 +928,14 @@ def _find_common_type(types):
     if all(is_timedelta64_dtype(t) for t in types):
         return np.dtype('timedelta64[ns]')
 
+    # don't mix bool / int or float or complex
+    # this is different from numpy, which casts bool with float/int as int
+    has_bools = any(is_bool_dtype(t) for t in types)
+    if has_bools:
+        has_ints = any(is_integer_dtype(t) for t in types)
+        has_floats = any(is_float_dtype(t) for t in types)
+        has_complex = any(is_complex_dtype(t) for t in types)
+        if has_ints or has_floats or has_complex:
+            return np.object
+
     return np.find_common_type(types, [])