Raise more informative error when converting tuples to Variable.

xarray/core/variable.py

@@ -76,11 +76,11 @@ def as_variable(obj, name=None):
     elif isinstance(obj, tuple):
         try:
             obj = Variable(*obj)
-        except TypeError:
+        except (TypeError, ValueError) as error:
             # use .format() instead of % because it handles tuples consistently
-            raise TypeError('tuples to convert into variables must be of the '
-                            'form (dims, data[, attrs, encoding]): '
-                            '{}'.format(obj))
+            raise error.__class__('Could not convert tuple of form '
+                                  '(dims, data[, attrs, encoding]): '
+                                  '{} to Variable.'.format(obj))
     elif utils.is_scalar(obj):
         obj = Variable([], obj)
     elif isinstance(obj, (pd.Index, IndexVariable)) and obj.name is not None:

xarray/tests/test_dataset.py

@@ -242,7 +242,7 @@ def test_constructor(self):
             Dataset({'a': x1, 'b': x2})
         with raises_regex(ValueError, "disallows such variables"):
             Dataset({'a': x1, 'x': z})
-        with raises_regex(TypeError, 'tuples to convert'):
+        with raises_regex(TypeError, 'tuple of form'):
             Dataset({'x': (1, 2, 3, 4, 5, 6, 7)})
         with raises_regex(ValueError, 'already exists as a scalar'):
             Dataset({'x': 0, 'y': ('x', [1, 2, 3])})

xarray/tests/test_variable.py

@@ -970,8 +970,10 @@ def test_as_variable(self):
                         expected_extra.attrs, expected_extra.encoding)
         assert_identical(expected_extra, as_variable(xarray_tuple))
 
-        with raises_regex(TypeError, 'tuples to convert'):
+        with raises_regex(TypeError, 'tuple of form'):
             as_variable(tuple(data))
+        with raises_regex(ValueError, 'tuple of form'):  # GH1016
+            as_variable(('five', 'six', 'seven'))
         with raises_regex(
                 TypeError, 'without an explicit list of dimensions'):
             as_variable(data)