Implement some improvements to random_ai and replay_loop

cgolubi1 commented 4 weeks ago

Add replay_loop --test-output, which nondestructively replay-tests whatever happens to be in the output directory right now
Add random_ai support for initializing a new game's data object in a more readable way, by leaning on existing diff methods
replay_loop should error out immediately if prep_replay_games fails

This partially addresses #2953; in particular it adds the flag to nondestructively test whatever is in output/ right now, and reduces the work needed to turn test output into a responderTest to much closer to zero.

I've tested the new flags manually, and played about 900 novel and replayed games, and i'm satisfied with the changes.

cgolubi1 commented 3 weeks ago

@blackshadowshade could you take a quick look at this this weekend? If it needs more scrutiny, i'm happy to wait, but i suspect you're going to rubber stamp it since it's replay-test-only, and it'd be handy for me to have the new flag in the codebase for testing other things.

blackshadowshade commented 3 weeks ago

Right, accepting these changes to random_ai and replay testing.

buttonmen-dev / buttonmen

Implement some improvements to random_ai and replay_loop #3002