make debugging easier by saying which constraint is bad (facebook#2737)

Summary:
Pull Request resolved: facebook#2737

Would have made debugging this problem way easier https://fb.workplace.com/groups/ae.feedback/posts/8591041154263824/?comment_id=8591648994203040&reply_comment_id=8591748957526377

Generally if there are a bunch of constraints, it's helpful to let the user know which constraint we have a problem with.

Reviewed By: eonofrey

Differential Revision: D62154912

fbshipit-source-id: c6cb6bbfdfb77192aaa46999794a105efb8c94e8

ax/service/tests/test_instantiation_utils.py

@@ -70,6 +70,9 @@ def test_constraint_from_str(self) -> None:
             },
         )
 
+        with self.assertRaisesRegex(AssertionError, "Outcome constraint 'm1"):
+            InstantiationBase.outcome_constraint_from_str("m1 == 2*m2")
+
         self.assertEqual(three_val_constaint.bound, 3.0)
         with self.assertRaisesRegex(ValueError, "Parameter constraint should"):
             InstantiationBase.constraint_from_str(

ax/service/utils/instantiation.py

@@ -477,7 +477,8 @@ def outcome_constraint_from_str(
         """Parse string representation of an outcome constraint."""
         tokens = representation.split()
         assert len(tokens) == 3 and tokens[1] in COMPARISON_OPS, (
-            "Outcome constraint should be of form `metric_name >= x`, where x is a "
+            f"Outcome constraint '{representation}' should be of "
+            "form `metric_name >= x`, where x is a "
             "float bound and comparison operator is >= or <=."
         )
         op = COMPARISON_OPS[tokens[1]]
@@ -489,7 +490,9 @@ def outcome_constraint_from_str(
                 bound_repr = bound_repr[:-1]
             bound = float(bound_repr)
         except ValueError:
-            raise ValueError("Outcome constraint bound should be a float.")
+            raise ValueError(
+                f"Outcome constraint bound should be a float for '{representation}'."
+            )
         return OutcomeConstraint(
             cls._make_metric(
                 name=tokens[0],