take working

pandas/core/dtypes/common.py

@@ -151,8 +151,8 @@ def is_sparse(arr):
     >>> is_sparse(bsr_matrix([1, 2, 3]))
     False
     """
-
-    return isinstance(arr, (ABCSparseArray, ABCSparseSeries))
+    from pandas.core.sparse.array import SparseArray
+    return isinstance(arr, (SparseArray, ABCSparseSeries))
 
 
 def is_scipy_sparse(arr):
@@ -1705,6 +1705,8 @@ def is_extension_array_dtype(arr_or_dtype):
     array interface. In pandas, this includes:
 
     * Categorical
+    * Sparse
+    * Interval
 
     Third-party libraries may implement arrays or types satisfying
     this interface as well.
@@ -1714,6 +1716,11 @@ def is_extension_array_dtype(arr_or_dtype):
     if isinstance(arr_or_dtype, (ABCIndexClass, ABCSeries)):
         arr_or_dtype = arr_or_dtype._values
 
+    is_extension_array = isinstance(arr_or_dtype, ExtensionArray)
+
+    if is_extension_array:
+        return True
+
     try:
         arr_or_dtype = pandas_dtype(arr_or_dtype)
     except TypeError:
@@ -1992,11 +1999,6 @@ def pandas_dtype(dtype):
     TypeError if not a dtype
 
     """
-    # short-circuit
-    if isinstance(dtype, np.ndarray):
-        return dtype.dtype
-    elif isinstance(dtype, np.dtype):
-        return dtype
 
     # registered extension types
     result = registry.find(dtype)
@@ -2007,6 +2009,12 @@ def pandas_dtype(dtype):
     elif isinstance(dtype, ExtensionDtype):
         return dtype
 
+    # short-circuit
+    if isinstance(dtype, np.ndarray):
+        return dtype.dtype
+    elif isinstance(dtype, np.dtype):
+        return dtype
+
     # try a numpy dtype
     # raise a consistent TypeError if failed
     try:

pandas/core/internals/__init__.py

@@ -2998,7 +2998,7 @@ def concat_same_type(self, to_concat, placement=None):
             values, placement=placement or slice(0, len(values), 1))
 
 
-class SparseBlock(NonConsolidatableMixIn, Block):
+class SparseBlock(ExtensionBlock):
     """ implement as a list of sparse arrays of the same dtype """
     __slots__ = ()
     is_sparse = True
@@ -3032,9 +3032,6 @@ def fill_value(self):
     def fill_value(self, v):
         self.values.fill_value = v
 
-    def to_dense(self):
-        return self.values.to_dense().view()
-
     @property
     def sp_values(self):
         return self.values.sp_values
@@ -3172,7 +3169,9 @@ def get_block_type(values, dtype=None):
     dtype = dtype or values.dtype
     vtype = dtype.type
 
-    if is_extension_array_dtype(values):
+    if is_categorical(values):
+        cls = CategoricalBlock
+    elif is_extension_array_dtype(values):
         cls = ExtensionBlock
     elif issubclass(vtype, np.floating):
         cls = FloatBlock
@@ -3181,8 +3180,6 @@ def get_block_type(values, dtype=None):
         cls = TimeDeltaBlock
     elif issubclass(vtype, np.complexfloating):
         cls = ComplexBlock
-    elif is_categorical(values):
-        cls = CategoricalBlock
     elif issubclass(vtype, np.datetime64):
         assert not is_datetimetz(values)
         cls = DatetimeBlock

pandas/core/sparse/array.py

@@ -131,6 +131,7 @@ def _wrap_result(name, data, sparse_index, fill_value, dtype=None):
 
 
 class SparseArray(PandasObject, ExtensionArray):
+
     def __init__(self, data, sp_index=None, fill_value=np.nan, kind='block'):
 
         if sp_index is None:
@@ -148,7 +149,7 @@ def __init__(self, data, sp_index=None, fill_value=np.nan, kind='block'):
         self.fill_value = fill_value
 
     @classmethod
-    def _from_sequence(cls, scalars, copy=False):
+    def _from_sequence(cls, scalars, dtype=None, copy=False):
         return cls(scalars)
 
     @classmethod
@@ -248,9 +249,11 @@ def take(self, indices, allow_fill=False, fill_value=None):
         indices = np.asarray(indices, dtype=np.int32)
 
         if allow_fill:
-            return self._take_with_fill(indices, fill_value=fill_value)
+            result = self._take_with_fill(indices, fill_value=fill_value)
         else:
-            return self._take_without_fill(indices)
+            result = self._take_without_fill(indices)
+
+        return type(self)(result, fill_value=self.fill_value)
 
     def _take_with_fill(self, indices, fill_value=None):
         if fill_value is None:
@@ -271,12 +274,12 @@ def _take_with_fill(self, indices, fill_value=None):
             else:
                 raise IndexError('cannot do a non-empty take from an empty axes.')
 
-        # TODO: bounds check
         sp_indexer = self.sp_index.lookup_array(indices)
-        fillable = (indices < 0) | (sp_indexer < 0)
-
         taken = self.sp_values.take(sp_indexer)
-        taken[fillable] = fill_value
+        # Have to fill in two steps, since the user-passed fill value may be
+        # different from self.fill_value.
+        taken[sp_indexer < 0] = self.fill_value
+        taken[indices < 0] = fill_value
         return taken
 
     def _take_without_fill(self, indices):

pandas/tests/extension/sparse/test_sparse.py

@@ -92,6 +92,11 @@ class TestGetitem(base.BaseGetitemTests):
     def test_take_non_na_fill_value(self, data_missing):
         pass
 
+    def test_get(self, data):
+        s = pd.Series(data, index=[2 * i for i in range(len(data))])
+        assert np.isnan(s.get(4)) and np.isnan(s.iloc[2])
+        assert s.get(2) == s.iloc[1]
+
 
 class TestSetitem(base.BaseSetitemTests):
     pass