Fix discrete CarRacing-v3

[Sebastian-Griesbach]

Description

CarRacing-v3 in discrete mode first converts the action to a float value before checking if the action is part of the discrete action-space. This causes an error to be raised.

Minimal example to reproduce this error:

import gymnasium as gym
env = gym.make("CarRacing-v3", continuous=False)
_ = env.reset()
action = env.action_space.sample()
_ = env.step(action)

This problem can be fixed simply by moving the float conversion to after the if self.continuous: check such that for the discrete case else: the action will still be an integer. And the InvalidAction error is not raised.

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• I have run the pre-commit checks with pre-commit run --all-files (see CONTRIBUTING.md instructions to set it up)
• [NA] I have commented my code, particularly in hard-to-understand areas
• [NA] I have made corresponding changes to the documentation
• My changes generate no new warnings
• [NA] I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[pseudo-rnd-thoughts]

Thanks for spotting this issue, could you add a test to check that this works, probably in tests/envs/test_env_implementations.py
I believe that Lunar Lander has a similar continuous=False, could you add this to the test. You just need to pass valid and invalid action and check that the error occurs

[Sebastian-Griesbach]

I added a test that checks if continuous actions fail in the discrete setting and discrete actions don't raise any error both for LunarLander-v3 and CarRacing-v3.

[pseudo-rnd-thoughts]

Thanks for making that change @Sebastian-Griesbach

[Kallinteris-Andreas]

@pseudo-rnd-thoughts For how long has this bug existed, do we need to list it in the environment version history

Gymnasium/gymnasium/envs/box2d/car_racing.py

Lines 192 to 196 in e732459

	## Version History
	- v2: Change truncation to termination when finishing the lap (1.0.0)
	- v1: Change track completion logic and add domain randomization (0.24.0)
	- v0: Original version

[pseudo-rnd-thoughts]

I dont think this changes environment behaviour