[Proposal] add utils.check_environments_match

[Kallinteris-Andreas]

Proposal

Add:

def check_environments_match(
    env_a: gym.Env,
    env_b: gym.Env,
    num_steps: int,
    seed: int = None,
    skip_obs: bool = False,
    skip_rew: bool = False,
    skip_terminal: bool = False,
    skip_truncated: bool = False,
    skip_info: bool = False,
):
    """Checks if the environments `env_a` & `env_b` are identical.

    Args:
        num_steps: number of timesteps to test for, setting to 0 tests only resetting.
        skip_obs: If `True` it does not check for equivalence of the observation.
        skip_rew: If `True` it does not check for equivalence of the observation.
        skip_terminal: If `True` it does not check for equivalence of the observation.
        skip_truncated: If `True` it does not check for equivalence of the observation.
        skip_info: If `True` it does not check for equivalence of the observation.
    """
    assert env_a.action_space == env_b.action_space
    assert skip_obs or env_b.observation_space == env_b.observation_space

    obs_a, info_a = env_a.reset(seed)
    obs_b, info_b = env_b.reset(seed)
    assert skip_obs or data_equivalence(obs_a, obs_b)
    assert skip_info or data_equivalence(info_a, info_b)
    env_a.action_space.seed(seed)
    for _ in range(num_steps):
        action = env_a.action_space.sample()
        obs_a, rew_a, terminal_a, truncated_a, info_a = env_a.step(action)
        obs_b, rew_b, terminal_b, truncated_b, info_b = env_b.step(action)
        assert skip_obs or data_equivalence(obs_a, obs_b)
        assert skip_rew or data_equivalence(rew_a, rew_b)
        assert skip_terminal or terminal_a == terminal_b
        assert skip_truncated or truncated_a == truncated_b
        assert skip_info or data_equivalence(info_a, info_b)

        if terminal_a or truncated_a:
            obs_a, info_a = env_a.reset(seed)
            obs_b, info_b = env_b.reset(seed)
            assert skip_obs or data_equivalence(obs_a, obs_b)
            assert skip_info or data_equivalence(info_a, info_b)

Note: pettingzoo has

https://github.com/Farama-Foundation/PettingZoo/blob/969c079164e6fd84b4878f26debfeff644aa2786/pettingzoo/test/seed_test.py#L16

Motivation

It could enable:

env_v4 = gym.make('HalfCheetah-v4')
env_v5 = gym.make("HalfCheetah-v5")
check_environments_match(env_v4, env_v5, 1000,

Or for any new environment revision.

Pitch

No response

Alternatives

I could simply keep it in test_mujoco_v5.py

Additional context

No response

Checklist

• I have checked that there is no similar issue in the repo

[pseudo-rnd-thoughts]

I think this is a good idea, thanks for the proposal.
At least internally, this would be helpful to assess if a version change is required