box.contains check dtype and promote non-ndarrays (#2374)

* box.contains check dtype and promote non-ndarrays

Closes: #2357 and #2298

Instead of only casting list to ndarray, cast any class to ndarray (if possible) and emit a warning when casting. Also, check if the dtype of the input matches the dtype of the space.

* use import warnings

* blackify

* changs from code review

* fix wrapped space

Co-authored-by: Tristan Deleu <[email protected]>

* fix box bondaries

Co-authored-by: Tristan Deleu <[email protected]>

* TEST: add regression test.

* STY: black

Co-authored-by: Tristan Deleu <[email protected]>

gym/envs/toy_text/kellycoinflip.py

@@ -79,7 +79,7 @@ def step(self, action):
         return self._get_obs(), reward, done, {}
 
     def _get_obs(self):
-        return np.array([self.wealth]), self.rounds
+        return np.array([self.wealth], dtype=np.float32), self.rounds
 
     def reset(self):
         self.rounds = self.max_rounds
@@ -236,11 +236,11 @@ def step(self, action):
 
     def _get_obs(self):
         return (
-            np.array([float(self.wealth)]),
+            np.array([float(self.wealth)], dtype=np.float32),
             self.rounds_elapsed,
             self.wins,
             self.losses,
-            np.array([float(self.max_ever_wealth)]),
+            np.array([float(self.max_ever_wealth)], dtype=np.float32),
         )
 
     def reset(self):

gym/spaces/box.py

@@ -1,4 +1,5 @@
 import numpy as np
+import warnings
 
 from .space import Space
 from gym import logger
@@ -138,10 +139,15 @@ def sample(self):
         return sample.astype(self.dtype)
 
     def contains(self, x):
-        if isinstance(x, list):
-            x = np.array(x)  # Promote list to array for contains check
+        if not isinstance(x, np.ndarray):
+            warnings.warn("Casting input x to numpy array.")
+            x = np.asarray(x, dtype=self.dtype)
+
         return (
-            x.shape == self.shape and np.all(x >= self.low) and np.all(x <= self.high)
+            np.can_cast(x.dtype, self.dtype)
+            and x.shape == self.shape
+            and np.any(x >= self.low)
+            and np.any(x <= self.high)
         )
 
     def to_jsonable(self, sample_n):

gym/spaces/tests/test_spaces.py

@@ -178,3 +178,18 @@ def test_class_inequality(spaces):
 def test_bad_space_calls(space_fn):
     with pytest.raises(AssertionError):
         space_fn()
+
+
+def test_box_dtype_check():
+    # Related Issues:
+    # https://github.com/openai/gym/issues/2357
+    # https://github.com/openai/gym/issues/2298
+
+    space = Box(0, 2, tuple(), dtype=np.float32)
+
+    # casting will match the correct type
+    assert space.contains(0.5)
+
+    # float64 is not in float32 space
+    assert not space.contains(np.array(0.5))
+    assert not space.contains(np.array(1))

gym/wrappers/test_flatten_observation.py

@@ -19,12 +19,14 @@ def test_flatten_observation(env_id):
         space = spaces.Tuple(
             (spaces.Discrete(32), spaces.Discrete(11), spaces.Discrete(2))
         )
-        wrapped_space = spaces.Box(-np.inf, np.inf, [32 + 11 + 2], dtype=np.float32)
+        wrapped_space = spaces.Box(0, 1, [32 + 11 + 2], dtype=np.int64)
     elif env_id == "KellyCoinflip-v0":
         space = spaces.Tuple(
             (spaces.Box(0, 250.0, [1], dtype=np.float32), spaces.Discrete(300 + 1))
         )
-        wrapped_space = spaces.Box(-np.inf, np.inf, [1 + (300 + 1)], dtype=np.float32)
+        low = np.zeros((302,), dtype=np.float64)
+        high = np.array([250.0] + [1.0] * 301, dtype=np.float64)
+        wrapped_space = spaces.Box(low, high, [1 + (300 + 1)], dtype=np.float64)
 
     assert space.contains(obs)
     assert wrapped_space.contains(wrapped_obs)