BUG: isnull doesnt handle PeriodNaT properly

Author: sinhrks

pandas/core/common.py

@@ -221,6 +221,8 @@ def _isnull_new(obj):
         return obj._constructor(obj._data.isnull(func=isnull))
     elif isinstance(obj, list) or hasattr(obj, '__array__'):
         return _isnull_ndarraylike(np.asarray(obj))
+    elif isinstance(obj, pd.Period):
+        return obj.ordinal == tslib.iNaT
     else:
         return obj is None
 
@@ -291,7 +293,6 @@ def _isnull_ndarraylike(obj):
                 values = values.values
             result = values.isnull()
         else:
-
             # Working around NumPy ticket 1542
             shape = values.shape
 
@@ -302,7 +303,7 @@ def _isnull_ndarraylike(obj):
                 vec = lib.isnullobj(values.ravel())
                 result[...] = vec.reshape(shape)
 
-    elif is_datetimelike(obj):
+    elif dtype in _DATELIKE_DTYPES or isinstance(obj, pd.PeriodIndex):
         # this is the NaT pattern
         result = values.view('i8') == tslib.iNaT
     else:

pandas/lib.pyx

@@ -218,6 +218,8 @@ cpdef checknull(object val):
         return get_timedelta64_value(val) == NPY_NAT
     elif is_array(val):
         return False
+    # elif is_periodnat(val):
+    #     return True
     else:
         return _checknull(val)
 
@@ -278,6 +280,8 @@ def item_from_zerodim(object val):
     """
     return util.unbox_if_zerodim(val)
 
+cdef is_periodnat(object val):
+    return hasattr(val, 'freq') and getattr(val, 'ordinal', None) == iNaT
 
 @cython.wraparound(False)
 @cython.boundscheck(False)
@@ -290,7 +294,7 @@ def isnullobj(ndarray[object] arr):
     result = np.zeros(n, dtype=np.uint8)
     for i from 0 <= i < n:
         val = arr[i]
-        result[i] = val is NaT or _checknull(val)
+        result[i] = val is NaT or _checknull(val) or is_periodnat(val)
     return result.view(np.bool_)
 
 @cython.wraparound(False)

pandas/tests/test_common.py

@@ -7,6 +7,7 @@
 import nose
 from nose.tools import assert_equal
 import numpy as np
+import pandas as pd
 from pandas.tslib import iNaT, NaT
 from pandas import Series, DataFrame, date_range, DatetimeIndex, Timestamp, Float64Index
 from pandas import compat
@@ -143,6 +144,18 @@ def test_isnull_nat():
     exp = np.array([True])
     assert(np.array_equal(result, exp))
 
+    result = isnull(pd.Period('NaT', freq='M'))
+    assert(result)
+    result = isnull([pd.Period('NaT', freq='M')])
+    exp = np.array([True])
+    assert(np.array_equal(result, exp))
+
+    result = notnull(pd.Period('NaT', freq='M'))
+    assert(not result)
+    result = notnull([pd.Period('NaT', freq='M')])
+    exp = np.array([False])
+    assert(np.array_equal(result, exp))
+
 def test_isnull_datetime():
     assert (not isnull(datetime.now()))
     assert notnull(datetime.now())
@@ -166,6 +179,30 @@ def test_isnull_datetime():
     mask = isnull(pidx[1:])
     assert(not mask.any())
 
+def test_isnull_period():
+    assert (not isnull(pd.Period('2011-01', freq='M')))
+    assert notnull(pd.Period('2011-01', freq='M'))
+
+    idx = pd.period_range('1/1/1990', periods=20)
+    assert(notnull(idx).all())
+
+    idx = pd.PeriodIndex(['NaT', '2011-01', '2011-02'], freq='M')
+    mask = isnull(idx)
+    assert(mask[0])
+    assert(not mask[1:].any())
+
+    mask = isnull(idx.asobject)
+    assert(mask[0])
+    assert(not mask[1:].any())
+
+    mask = notnull(idx)
+    assert(not mask[0])
+    assert(mask[1:].all())
+
+    mask = notnull(idx.asobject)
+    assert(not mask[0])
+    assert(mask[1:].all())
+
 
 class TestIsNull(tm.TestCase):
     def test_0d_array(self):

vb_suite/panel_methods.py

@@ -8,8 +8,8 @@
 # shift
 
 setup = common_setup + """
-index = date_range(start="2000", freq="D", periods=1000)
-panel = Panel(np.random.randn(100, len(index), 1000))
+index = date_range(start="2000", freq="D", periods=100)
+panel = Panel(np.random.randn(100, len(index), 100))
 """
 
 panel_shift = Benchmark('panel.shift(1)', setup,