Disregard washer rules for now

bddl/knowledge_base/models.py

@@ -357,23 +357,29 @@ def transition_subgraph(self):
         return nx.relabel_nodes(G, lambda x: (x.name if isinstance(x, Synset) else f'recipe: {x.name}'), copy=True)
 
     def is_produceable_from(self, synsets):
-        # If it's already available, then we're good.
-        if self in synsets:
-            return True, set()
-
-        # Otherwise, are there any recipes that I can use to obtain it?
-        recipe_alternatives = set()
-        for recipe in self.produced_by_transition_rules:
-            producabilities_and_recipe_sets = [ingredient.is_produceable_from(synsets) for ingredient in recipe.input_synsets]
-            producabilities, recipe_sets = zip(*producabilities_and_recipe_sets)
-            if all(producabilities):
-                recipe_alternatives.add(recipe)
-                recipe_alternatives.update(ingredient_recipe for recipe_set in recipe_sets for ingredient_recipe in recipe_set)
+        def _is_produceable_from(_self, _synsets, _seen):
+            if _self in _seen:
+                return False, set()
+
+            # If it's already available, then we're good.
+            if _self in _synsets:
+                return True, set()
+
+            # Otherwise, are there any recipes that I can use to obtain it?
+            recipe_alternatives = set()
+            for recipe in _self.produced_by_transition_rules:
+                producabilities_and_recipe_sets = [_is_produceable_from(ingredient, _synsets, _seen | {self}) for ingredient in recipe.input_synsets]
+                producabilities, recipe_sets = zip(*producabilities_and_recipe_sets)
+                if all(producabilities):
+                    recipe_alternatives.add(recipe)
+                    recipe_alternatives.update(ingredient_recipe for recipe_set in recipe_sets for ingredient_recipe in recipe_set)
+
+            if not recipe_alternatives:
+                return False, set()
 
-        if not recipe_alternatives:
-            return False, set()
+            return True, recipe_alternatives
 
-        return True, recipe_alternatives
+        return _is_produceable_from(self, synsets, set())
 
 
 @dataclass(eq=False, order=False)

bddl/knowledge_base/processing.py

@@ -309,16 +309,24 @@ def create_transitions(self):
         json_paths = glob.glob(str(GENERATED_DATA_DIR / "transition_map/tm_jsons/*.json"))
         transitions = []
         for jp in json_paths:
+            if "washer_" in jp:
+                continue
             with open(jp) as f:
                 transitions.extend(json.load(f))
 
         # Create the transition objects
         for transition_data in self.tqdm(transitions):
-            transition = TransitionRule.create(name=transition_data["rule_name"])
-            for synset_name in transition_data["input_objects"].keys():
+            rule_name = transition_data["rule_name"]
+            transition = TransitionRule.create(name=rule_name)
+            inputs = set(transition_data["input_objects"].keys())
+            assert inputs, f"Transition {transition.name} has no inputs!"
+            outputs = set(transition_data["output_objects"].keys())
+            assert outputs, f"Transition {transition.name} has no outputs!"
+            assert inputs & outputs == set(), f"Inputs and outputs of {transition.name} overlap!"
+            for synset_name in inputs:
                 synset = Synset.get(name=synset_name)
                 transition.input_synsets.add(synset)
-            for synset_name in transition_data["output_objects"].keys():
+            for synset_name in outputs:
                 synset = Synset.get(name=synset_name)
                 transition.output_synsets.add(synset)