Merge pull request #8726 from jreback/inf

COMPAT: Compat issue is DataFrame.dtypes when options.mode.use_inf_as_null is True (GH8722)

Author: jreback

doc/source/whatsnew/v0.15.1.txt

@@ -212,7 +212,7 @@ Bug Fixes
 - Bug in duplicated/drop_duplicates with a Categorical (:issue:`8623`)
 - Bug in ``Categorical`` reflected comparison operator raising if the first argument was a numpy array scalar (e.g. np.int64) (:issue:`8658`)
 - Bug in Panel indexing with a list-like (:issue:`8710`)
-
+- Compat issue is ``DataFrame.dtypes`` when ``options.mode.use_inf_as_null`` is True (:issue:`8722`)
 
 
 

pandas/lib.pyx

@@ -289,8 +289,8 @@ def isnullobj(ndarray[object] arr):
     n = len(arr)
     result = np.zeros(n, dtype=np.uint8)
     for i from 0 <= i < n:
-        arobj = arr[i]
-        result[i] = arobj is NaT or _checknull(arobj)
+        val = arr[i]
+        result[i] = val is NaT or _checknull(val)
     return result.view(np.bool_)
 
 @cython.wraparound(False)
@@ -303,10 +303,10 @@ def isnullobj_old(ndarray[object] arr):
     n = len(arr)
     result = np.zeros(n, dtype=np.uint8)
     for i from 0 <= i < n:
-        result[i] = util._checknull_old(arr[i])
+        val = arr[i]
+        result[i] = val is NaT or util._checknull_old(val)
     return result.view(np.bool_)
 
-
 @cython.wraparound(False)
 @cython.boundscheck(False)
 def isnullobj2d(ndarray[object, ndim=2] arr):
@@ -323,20 +323,6 @@ def isnullobj2d(ndarray[object, ndim=2] arr):
                 result[i, j] = 1
     return result.view(np.bool_)
 
-@cython.wraparound(False)
-@cython.boundscheck(False)
-def isnullobj_old(ndarray[object] arr):
-    cdef Py_ssize_t i, n
-    cdef object val
-    cdef ndarray[uint8_t] result
-
-    n = len(arr)
-    result = np.zeros(n, dtype=np.uint8)
-    for i from 0 <= i < n:
-        result[i] = util._checknull_old(arr[i])
-    return result.view(np.bool_)
-
-
 @cython.wraparound(False)
 @cython.boundscheck(False)
 def isnullobj2d_old(ndarray[object, ndim=2] arr):

pandas/src/util.pxd

@@ -76,7 +76,7 @@ cdef inline bint _checknull_old(object val):
     cdef double INF = <double> np.inf
     cdef double NEGINF = -INF
     try:
-        return val is None or val != val or val == INF or val == NEGINF
+        return val is None or (cpython.PyFloat_Check(val) and (val != val or val == INF or val == NEGINF))
     except ValueError:
         return False
 

pandas/tests/test_frame.py

@@ -6783,6 +6783,12 @@ def test_dtypes(self):
                           index=result.index)
         assert_series_equal(result, expected)
 
+        # compat, GH 8722
+        with option_context('use_inf_as_null',True):
+            df = DataFrame([[1]])
+            result = df.dtypes
+            assert_series_equal(result,Series({0:np.dtype('int64')}))
+
     def test_convert_objects(self):
 
         oops = self.mixed_frame.T.T

pandas/tests/test_lib.py

@@ -5,7 +5,7 @@
 import pandas as pd
 from pandas.lib import isscalar, item_from_zerodim
 import pandas.util.testing as tm
-
+from pandas.compat import u
 
 class TestIsscalar(tm.TestCase):
     def test_isscalar_builtin_scalars(self):
@@ -16,7 +16,7 @@ def test_isscalar_builtin_scalars(self):
         self.assertTrue(isscalar(np.nan))
         self.assertTrue(isscalar('foobar'))
         self.assertTrue(isscalar(b'foobar'))
-        self.assertTrue(isscalar(u'foobar'))
+        self.assertTrue(isscalar(u('efoobar')))
         self.assertTrue(isscalar(datetime(2014, 1, 1)))
         self.assertTrue(isscalar(date(2014, 1, 1)))
         self.assertTrue(isscalar(time(12, 0)))
@@ -38,7 +38,7 @@ def test_isscalar_numpy_array_scalars(self):
         self.assertTrue(isscalar(np.int32(1)))
         self.assertTrue(isscalar(np.object_('foobar')))
         self.assertTrue(isscalar(np.str_('foobar')))
-        self.assertTrue(isscalar(np.unicode_(u'foobar')))
+        self.assertTrue(isscalar(np.unicode_(u('foobar'))))
         self.assertTrue(isscalar(np.bytes_(b'foobar')))
         self.assertTrue(isscalar(np.datetime64('2014-01-01')))
         self.assertTrue(isscalar(np.timedelta64(1, 'h')))