BUG: Fix broken isnull() on string classifiers.

Adds a special case in NullFilter to handle LabelArrays correctly.

tests/pipeline/test_classifier.py

@@ -21,7 +21,7 @@ unicode_dtype = np.dtype('U3')
 class ClassifierTestCase(BasePipelineTestCase):
 
     @parameter_space(mv=[-1, 0, 1, 999])
-    def test_isnull(self, mv):
+    def test_integral_isnull(self, mv):
 
         class C(Classifier):
             dtype = int64_dtype
@@ -51,6 +51,41 @@ class ClassifierTestCase(BasePipelineTestCase):
             mask=self.build_mask(self.ones_mask(shape=data.shape)),
         )
 
+    @parameter_space(mv=['0', None])
+    def test_string_isnull(self, mv):
+
+        class C(Classifier):
+            dtype = categorical_dtype
+            missing_value = mv
+            inputs = ()
+            window_length = 0
+
+        c = C()
+
+        # There's no significance to the values here other than that they
+        # contain a mix of missing and non-missing values.
+        raw = np.asarray(
+            [['',    'a',  'ab', 'ba'],
+             ['z',  'ab',   'a', 'ab'],
+             ['aa', 'ab',    '', 'ab'],
+             ['aa',  'a',  'ba', 'ba']],
+            dtype=categorical_dtype,
+        )
+        data = LabelArray(raw, missing_value=mv)
+
+        self.check_terms(
+            terms={
+                'isnull': c.isnull(),
+                'notnull': c.notnull()
+            },
+            expected={
+                'isnull': np.equal(raw, mv),
+                'notnull': np.not_equal(raw, mv),
+            },
+            initial_workspace={c: data},
+            mask=self.build_mask(self.ones_mask(shape=data.shape)),
+        )
+
     @parameter_space(compval=[0, 1, 999])
     def test_eq(self, compval):
 

zipline/lib/rank.pyx

@@ -28,9 +28,9 @@ from zipline.utils.numpy_utils import (
 import_array()
 
 
-cpdef ismissing(ndarray data, object missing_value):
+cpdef is_missing(ndarray data, object missing_value):
     """
-    Generic ismissing function that handles quirks with NaN.
+    Generic is_missing function that handles quirks with NaN.
     """
     if is_float(data) and isnan(missing_value):
         return isnan(data)
@@ -51,7 +51,7 @@ def masked_rankdata_2d(ndarray data,
             "Can't compute rankdata on array of dtype %r." % dtype_name
         )
 
-    cdef ndarray missing_locations = (~mask | ismissing(data, missing_value))
+    cdef ndarray missing_locations = (~mask | is_missing(data, missing_value))
 
     # Interpret the bytes of integral data as floats for sorting.
     data = data.copy().view(float64)

zipline/pipeline/filters/filter.py

@@ -14,7 +14,8 @@ from zipline.errors import (
     BadPercentileBounds,
     UnsupportedDataType,
 )
-from zipline.lib.rank import ismissing
+from zipline.lib.labelarray import LabelArray
+from zipline.lib.rank import is_missing
 from zipline.pipeline.mixins import (
     CustomTermMixin,
     LatestMixin,
@@ -230,19 +231,22 @@ class NullFilter(SingleInputMixin, Filter):
 
     Parameters
     ----------
-    factor : zipline.pipeline.Factor
+    factor : zipline.pipeline.Term
         The factor to compare against its missing_value.
     """
     window_length = 0
 
-    def __new__(cls, factor):
+    def __new__(cls, term):
         return super(NullFilter, cls).__new__(
             cls,
-            inputs=(factor,),
+            inputs=(term,),
         )
 
     def _compute(self, arrays, dates, assets, mask):
-        return ismissing(arrays[0], self.inputs[0].missing_value)
+        data = arrays[0]
+        if isinstance(data, LabelArray):
+            return data.is_missing()
+        return is_missing(arrays[0], self.inputs[0].missing_value)
 
 
 class PercentileFilter(SingleInputMixin, Filter):