ENH: add repr to MultiDimensionalDataset

Author: Joe Jevnik

tests/pipeline/test_multidimensional_dataset.py

@@ -21,6 +21,31 @@
 
 
 class TestMultiDimensionalDataSet(ZiplineTestCase):
+    def test_repr(self):
+        class MD1(MultiDimensionalDataSet):
+            extra_dims = [('dim_0', [])]
+
+        expected_repr = (
+            "<MultiDimensionalDataSet: 'MD1', extra_dims=['dim_0']>"
+        )
+        assert_equal(repr(MD1), expected_repr)
+
+        class MD2(MultiDimensionalDataSet):
+            extra_dims = [('dim_0', []), ('dim_1', [])]
+
+        expected_repr = (
+            "<MultiDimensionalDataSet: 'MD2', extra_dims=['dim_0', 'dim_1']>"
+        )
+        assert_equal(repr(MD2), expected_repr)
+
+        class MD3(MultiDimensionalDataSet):
+            extra_dims = [('dim_1', []), ('dim_0', [])]
+
+        expected_repr = (
+            "<MultiDimensionalDataSet: 'MD3', extra_dims=['dim_1', 'dim_0']>"
+        )
+        assert_equal(repr(MD3), expected_repr)
+
     def test_cache(self):
         class MD1(MultiDimensionalDataSet):
             extra_dims = [('dim_0', ['a', 'b', 'c'])]
@@ -270,12 +295,14 @@ class Child(Parent):
         def make_expected_msg(ds, attr):
             return dedent(
                 """\
-                Attempted to access column from a MultiDimensionalDataSet.
-                You must first slice the dataset along the extra dimensions like:
+                Attempted to access column {c} from multi-dimensional dataset {d}:
+
+                To work with multi-dimensional datasets, you must first choose a
+                slice using the ``slice`` method:
 
-                    %s.slice(...).%s
-                """,  # noqa
-            ) % (ds, attr)
+                    {d}.slice(...).{c}
+                """.format(c=attr, d=ds),  # noqa
+            )
 
         expected_msg = make_expected_msg('Parent', 'column_0')
         with assert_raises_str(AttributeError, expected_msg):

zipline/pipeline/data/dataset.py

@@ -521,14 +521,17 @@ def __init__(self, dataset_name, column_name):
         self.column_name = column_name
 
     def __str__(self):
+        # NOTE: when ``aggregate`` is added, remember to update this message
         return dedent(
             """\
-            Attempted to access column from a MultiDimensionalDataSet.
-            You must first slice the dataset along the extra dimensions like:
+            Attempted to access column {c} from multi-dimensional dataset {d}:
 
-                %s.slice(...).%s
-            """
-        ) % (self.dataset_name, self.column_name)
+            To work with multi-dimensional datasets, you must first choose a
+            slice using the ``slice`` method:
+
+                {d}.slice(...).{c}
+            """.format(c=self.column_name, d=self.dataset_name)
+        )
 
 
 class _MultiDimensionalDataSetColumn(object):
@@ -596,6 +599,12 @@ class BaseSlice(self._SliceType):
         self._slice_cache = {}
         return self
 
+    def __repr__(self):
+        return '<MultiDimensionalDataSet: %r, extra_dims=%r>' % (
+            self.__name__,
+            list(self.extra_dims),
+        )
+
 
 _base = with_metaclass(
     MultiDimensionalDataSetMeta,
@@ -672,12 +681,12 @@ class SomeDataSet(MultiDimensionalDataSet):
 
     _SliceType = MultiDimensionalDataSetSlice
 
-    @abc.abstractproperty
-    def extra_dims(self):
-        raise NotImplementedError(
-            'extra_dims must be specified as a sequence of tuples of dimension'
-            ' name and the set of values on the given dimension',
-        )
+    @type.__call__
+    class extra_dims(object):
+        __isabstractmethod__ = True
+
+        def __get__(self, instance, owner):
+            return []
 
     @classmethod
     def _canonical_key(cls, args, kwargs):