Handle set actlist in MDP.__init__ (#1259)

MDP.__init__ only assigned self.actlist for list and dict inputs; a set
fell through both branches, leaving the attribute unset so MDP.actions()
raised AttributeError (and value/policy iteration crashed). Add a set
branch that stores the actions as a list, keeping them indexable for
random.choice in policy_iteration. Mirror the fix in mdp4e.py and extend
the existing set-actlist test to exercise actions().

Author: dmeoli

mdp.py

@@ -42,6 +42,10 @@ def __init__(self, init, actlist, terminals, transitions=None, reward=None, stat
             # if actlist is a dict, different actions for each state
             self.actlist = actlist
 
+        elif isinstance(actlist, set):
+            # if actlist is a set, convert it to a list so actions are indexable
+            self.actlist = list(actlist)
+
         self.terminals = terminals
         self.transitions = transitions or {}
         if not self.transitions:

mdp4e.py

@@ -42,6 +42,10 @@ def __init__(self, init, actlist, terminals, transitions=None, reward=None, stat
             # if actlist is a dict, different actions for each state
             self.actlist = actlist
 
+        elif isinstance(actlist, set):
+            # if actlist is a set, convert it to a list so actions are indexable
+            self.actlist = list(actlist)
+
         self.terminals = terminals
         self.transitions = transitions or {}
         if not self.transitions:

tests/test_mdp.py

@@ -126,6 +126,10 @@ def test_transition_model():
     assert mdp.T("b", "plan2") == [(0.6, 'a'), (0.2, 'b'), (0.1, 'c'), (0.1, 'd')]
     assert mdp.T("c", "plan1") == [(0.3, 'a'), (0.5, 'b'), (0.1, 'c'), (0.1, 'd')]
 
+    # a set actlist must be accepted and yield indexable per-state actions
+    assert set(mdp.actions("a")) == {"plan1", "plan2", "plan3"}
+    assert mdp.actions("d") == [None]
+
 
 def test_pomdp_value_iteration():
     t_prob = [[[0.65, 0.35], [0.65, 0.35]], [[0.65, 0.35], [0.65, 0.35]], [[1.0, 0.0], [0.0, 1.0]]]