Improve loop algorithms crosschecking for experiment

The current technique use had a flaw. With algorithm containing a different number of groups, it would truncate the to the shortest as explained in the zip function documentation.

Now itertools.zip_longest is used to ensure that the iteration is done on the longest input.

beat/core/experiment.py

@@ -44,6 +44,7 @@ Validation for experiments
 
 import os
 import collections
+import itertools
 import simplejson as json
 
 from . import utils
@@ -977,15 +978,21 @@ class Experiment(object):
                 evaluator_algorithm_name
             ].groups
 
-            groups_count = [
-                (
-                    len(algo_groups["inputs"]) + len(evaluator_algo_groups["inputs"]),
-                    len(algo_groups["outputs"]) + len(evaluator_algo_groups["outputs"]),
-                )
-                for algo_groups, evaluator_algo_groups in zip(
-                    algo_groups_list, evaluator_algo_groups_list
-                )
-            ]
+            groups_count = []
+            for algo_groups, evaluator_algo_groups in itertools.zip_longest(
+                algo_groups_list, evaluator_algo_groups_list
+            ):
+                inputs = 0
+                outputs = 0
+                if algo_groups:
+                    inputs = len(algo_groups["inputs"])
+                    outputs = len(algo_groups.get("outputs", []))
+
+                if evaluator_algo_groups:
+                    inputs += len(evaluator_algo_groups["inputs"])
+                    outputs += len(evaluator_algo_groups.get("outputs", []))
+
+                groups_count.append((inputs, outputs))
 
             if collections.Counter(chain_groups_count) != collections.Counter(
                 groups_count