API: empty map should not infer (#18517)

closes #18509

Author: jreback

doc/source/whatsnew/v0.22.0.txt

@@ -107,7 +107,7 @@ Other API Changes
 - :class:`CacheableOffset` and :class:`WeekDay` are no longer available in the ``pandas.tseries.offsets`` module (:issue:`17830`)
 - `tseries.frequencies.get_freq_group()` and `tseries.frequencies.DAYS` are removed from the public API (:issue:`18034`)
 - :func:`Series.truncate` and :func:`DataFrame.truncate` will raise a ``ValueError`` if the index is not sorted instead of an unhelpful ``KeyError`` (:issue:`17935`)
-- :func:`Index.map` can now accept ``Series`` and dictionary input objects (:issue:`12756`, :issue:`18482`).
+- :func:`Index.map` can now accept ``Series`` and dictionary input objects (:issue:`12756`, :issue:`18482`, :issue:`18509`).
 - :func:`Dataframe.unstack` will now default to filling with ``np.nan`` for ``object`` columns. (:issue:`12815`)
 - :class:`IntervalIndex` constructor will raise if the ``closed`` parameter conflicts with how the input data is inferred to be closed (:issue:`18421`)
 - Inserting missing values into indexes will work for all types of indexes and automatically insert the correct type of missing value (``NaN``, ``NaT``, etc.) regardless of the type passed in (:issue:`18295`)

pandas/core/indexes/base.py

@@ -2897,25 +2897,9 @@ def map(self, mapper, na_action=None):
                                           names=names)
 
         attributes['copy'] = False
-
-        # we want to try to return our original dtype
-        # ints infer to integer, but if we have
-        # uints, would prefer to return these
-        if is_unsigned_integer_dtype(self.dtype):
-            inferred = lib.infer_dtype(new_values)
-            if inferred == 'integer':
-                attributes['dtype'] = self.dtype
-
-        elif not new_values.size:
+        if not new_values.size:
             # empty
             attributes['dtype'] = self.dtype
-        elif isna(new_values).all():
-            # all nan
-            inferred = lib.infer_dtype(self)
-            if inferred in ['datetime', 'datetime64',
-                            'timedelta', 'timedelta64',
-                            'period']:
-                new_values = [libts.NaT] * len(new_values)
 
         return Index(new_values, **attributes)
 

pandas/tests/indexes/common.py

@@ -1009,7 +1009,13 @@ def test_searchsorted_monotonic(self, indices):
     def test_map(self):
         # callable
         index = self.create_index()
-        expected = index
+
+        # we don't infer UInt64
+        if isinstance(index, pd.UInt64Index):
+            expected = index.astype('int64')
+        else:
+            expected = index
+
         result = index.map(lambda x: x)
         tm.assert_index_equal(result, expected)
 
@@ -1024,9 +1030,14 @@ def test_map_dictlike(self, mapper):
         if isinstance(index, (pd.CategoricalIndex, pd.IntervalIndex)):
             pytest.skip("skipping tests for {}".format(type(index)))
 
-        expected = index
-
         identity = mapper(index.values, index)
+
+        # we don't infer to UInt64 for a dict
+        if isinstance(index, pd.UInt64Index) and isinstance(identity, dict):
+            expected = index.astype('int64')
+        else:
+            expected = index
+
         result = index.map(identity)
         tm.assert_index_equal(result, expected)
 

pandas/tests/indexes/datetimelike.py

@@ -1,5 +1,6 @@
 """ generic datetimelike tests """
 import pytest
+import numpy as np
 import pandas as pd
 from .common import Base
 import pandas.util.testing as tm
@@ -72,6 +73,6 @@ def test_map_dictlike(self, mapper):
 
         # empty map; these map to np.nan because we cannot know
         # to re-infer things
-        expected = pd.Index([pd.NaT] * len(self.index))
+        expected = pd.Index([np.nan] * len(self.index))
         result = self.index.map(mapper([], []))
         tm.assert_index_equal(result, expected)

pandas/tests/indexes/test_base.py

@@ -885,9 +885,7 @@ def test_map_dictlike(self, mapper):
             expected = Index(np.arange(len(index), 0, -1))
 
             # to match proper result coercion for uints
-            if name == 'uintIndex':
-                expected = expected.astype('uint64')
-            elif name == 'empty':
+            if name == 'empty':
                 expected = Index([])
 
             result = index.map(mapper(expected, index))

pandas/tests/series/test_apply.py

@@ -377,6 +377,14 @@ def test_map(self):
         exp = Series([np.nan, 'B', 'C', 'D'])
         tm.assert_series_equal(a.map(c), exp)
 
+    @pytest.mark.parametrize("index", tm.all_index_generator(10))
+    def test_map_empty(self, index):
+        s = Series(index)
+        result = s.map({})
+
+        expected = pd.Series(np.nan, index=s.index)
+        tm.assert_series_equal(result, expected)
+
     def test_map_compat(self):
         # related GH 8024
         s = Series([True, True, False], index=[1, 2, 3])

pandas/tests/test_resample.py

@@ -816,21 +816,23 @@ def test_resample_empty_dataframe(self):
 
             # test size for GH13212 (currently stays as df)
 
-    def test_resample_empty_dtypes(self):
+    @pytest.mark.parametrize("index", tm.all_timeseries_index_generator(0))
+    @pytest.mark.parametrize(
+        "dtype",
+        [np.float, np.int, np.object, 'datetime64[ns]'])
+    def test_resample_empty_dtypes(self, index, dtype):
 
         # Empty series were sometimes causing a segfault (for the functions
         # with Cython bounds-checking disabled) or an IndexError.  We just run
         # them to ensure they no longer do.  (GH #10228)
-        for index in tm.all_timeseries_index_generator(0):
-            for dtype in (np.float, np.int, np.object, 'datetime64[ns]'):
-                for how in downsample_methods + upsample_methods:
-                    empty_series = Series([], index, dtype)
-                    try:
-                        getattr(empty_series.resample('d'), how)()
-                    except DataError:
-                        # Ignore these since some combinations are invalid
-                        # (ex: doing mean with dtype of np.object)
-                        pass
+        for how in downsample_methods + upsample_methods:
+            empty_series = Series([], index, dtype)
+            try:
+                getattr(empty_series.resample('d'), how)()
+            except DataError:
+                # Ignore these since some combinations are invalid
+                # (ex: doing mean with dtype of np.object)
+                pass
 
     def test_resample_loffset_arg_type(self):
         # GH 13218, 15002

pandas/util/testing.py

@@ -1695,7 +1695,8 @@ def all_index_generator(k=10):
     """
     all_make_index_funcs = [makeIntIndex, makeFloatIndex, makeStringIndex,
                             makeUnicodeIndex, makeDateIndex, makePeriodIndex,
-                            makeTimedeltaIndex, makeBoolIndex,
+                            makeTimedeltaIndex, makeBoolIndex, makeRangeIndex,
+                            makeIntervalIndex,
                             makeCategoricalIndex]
     for make_index_func in all_make_index_funcs:
         yield make_index_func(k=k)