Address invalid lengths

Author: jschendel

pandas/_libs/interval.pyx

@@ -59,7 +59,12 @@ cdef class IntervalMixin(object):
     @property
     def length(self):
         """Return the length of the Interval"""
-        return self.right - self.left
+        try:
+            return self.right - self.left
+        except TypeError:
+            # length not defined for some types, e.g. string
+            msg = 'cannot compute length between {left} and {right}'
+            raise TypeError(msg.format(left=self.left, right=self.right))
 
 
 cdef _interval_like(other):

pandas/core/indexes/interval.py

@@ -605,7 +605,13 @@ def length(self):
         Return an Index with entries denoting the length of each Interval in
         the IntervalIndex
         """
-        return self.right - self.left
+        try:
+            return self.right - self.left
+        except TypeError:
+            # length not defined for some types, e.g. string
+            msg = ('IntervalIndex contains Intervals without defined length, '
+                   'e.g. Intervals with string endpoints')
+            raise TypeError(msg)
 
     def __len__(self):
         return len(self.left)

pandas/tests/indexes/test_interval.py

@@ -300,6 +300,17 @@ def test_length(self, closed, breaks):
         expected = Index(iv.length if notna(iv) else iv for iv in index)
         tm.assert_index_equal(result, expected)
 
+    @pytest.mark.parametrize('breaks', [
+        list('abcdefgh'),
+        lzip(range(10), range(1, 11)),
+        [['A', 'B'], ['a', 'b'], ['c', 'd'], ['e', 'f']],
+        [Interval(0, 1), Interval(1, 2), Interval(3, 4), Interval(4, 5)]])
+    def test_length_errors(self, closed, breaks):
+        index = IntervalIndex.from_breaks(breaks)
+        msg = 'IntervalIndex contains Intervals without defined length'
+        with tm.assert_raises_regex(TypeError, msg):
+            index.length
+
     def test_with_nans(self, closed):
         index = self.create_index(closed=closed)
         assert not index.hasnans

pandas/tests/scalar/test_interval.py

@@ -97,6 +97,19 @@ def test_length_timestamp(self, tz, left, right, expected):
         expected = Timedelta(expected)
         assert result == expected
 
+    @pytest.mark.parametrize('left, right', [
+        ('a', 'z'),
+        (('a', 'b'), ('c', 'd')),
+        (list('AB'), list('ab')),
+        (Interval(0, 1), Interval(1, 2))])
+    def test_length_errors(self, left, right):
+        # GH 18789
+        iv = Interval(left, right)
+        msg = 'cannot compute length between {left} and {right}'.format(
+            left=left, right=right)
+        with tm.assert_raises_regex(TypeError, msg):
+            iv.length
+
     def test_math_add(self, interval):
         expected = Interval(1, 2)
         actual = interval + 1