make arg check private and add note about init

opt_einsum/path_random.py

@@ -163,7 +163,7 @@ def setup(self, inputs, output, size_dict):
         raise NotImplementedError
 
     def __call__(self, inputs, output, size_dict, memory_limit):
-        self.check_args_against_first_call(inputs, output, size_dict)
+        self._check_args_against_first_call(inputs, output, size_dict)
 
         # start a timer?
         if self.max_time is not None:

opt_einsum/paths.py

@@ -44,12 +44,13 @@ def __call__(self, inputs, output, size_dict, memory_limit=None):
     where ``path`` is a list of int-tuples specifiying a contraction order.
     """
 
-    def check_args_against_first_call(self, inputs, output, size_dict):
+    def _check_args_against_first_call(self, inputs, output, size_dict):
         """Utility that stateful optimizers can use to ensure they are not
         called with different contractions across separate runs.
         """
         args = (inputs, output, size_dict)
         if not hasattr(self, '_first_call_args'):
+            # simply set the attribute as currently there is no global PathOptimizer init
             self._first_call_args = args
         elif args != self._first_call_args:
             raise ValueError("The arguments specifiying the contraction that this path optimizer "
@@ -348,7 +349,7 @@ def __call__(self, inputs, output, size_dict, memory_limit=None):
         >>> optimal(isets, oset, idx_sizes, 5000)
         [(0, 2), (0, 1)]
         """
-        self.check_args_against_first_call(inputs, output, size_dict)
+        self._check_args_against_first_call(inputs, output, size_dict)
 
         inputs = tuple(map(frozenset, inputs))
         output = frozenset(output)