Gen 8 support

[szymonWojdat]

As discussed in #20

Copying over a checklist by @hsahovic that he mentioned here in order to keep track of what's left

Things to update:

• src/data/pokedex.json
• src/data/moves.json
• src/environment/Battle.battle_parse_message
• src/environment/Battle.battle_parse_request
• src/player/Player.choose_random_move

Things to add:

• src/environment/Battle.can_dynamax
• src/environment/Pokemon._dynamax

Things to add if necessary:

• src/environment/Battle.can_gigantamax
• src/environment/Pokemon._gigantamax
• src/environment/Pokemon.is_dynamaxed
• src/environment/Pokemon.is_gigantamaxed

[szymonWojdat]

Alright, now that I think about it, this information (can_dynamax) actually belongs in the Battle state because it tells whether it is possible to dynamax or not, while Pokemon instance only knows if it's currently dynamaxed or not. When it comes to available_moves I think I'll just keep it simple for now. I was considering adding it to Battle state to have it dynamically change between max moves, regular moves and both depending on the dynamax state of the Battle (which would also make Player.choose_random_move() a bit simpler. But for now I think it's going to be the best to do it in a similar way to max moves/megas and maybe optimize later on. Of course I won't be making such changes without discussing them with you first.

[szymonWojdat]

I just managed to rerun tests locally, adjusted a few methods, nothing too big, just pushed out the changes. A couple of issues:

• Move.can_z_move doesn't work as expected as moves.json no longer contains zMovePower for all damage moves, my guess is that PS has switched to some sort of dynamic calculation of the z-move power. The only place where it's used in code is at Pokemon.available_z_moves(). I assume the easiest workaround would be to modify Move.can_z_move so that it checks if given moves appears on a small list of moves that cannot z-move (test_can_z_move() implies as if Metronome could not z-move but it actually can, it just doesn't have any extra effect, moves that can't z-moves would be eg. Struggle)
• for the same reason as mentioned above, test_z_move_power is failing as well. I'll try taking a look at their codebase to see what's the formula that they're using

[hsahovic]

@szymonWojdat thanks for bringing that up, I'll look into it :)

[szymonWojdat]

Just added:

• Battle.opponent_can_dynamax - as discussed
• Battle.dynamax_turns_left - as a way of determining how many turns of dynamax are left for player, I figured this would be crucial information for any agent as it wouldn't have a way of figuring out how many turns of dynamax are left otherwise based on the current Battle state
• Battle.opponent_dynamax_turns_left - same as above but for the opponent

I have tested both locally manually. Aside from writing tests for the new functionalities and fixing the failing ones, I believe we're done?

[hsahovic]

We'll have to update docs too and be sure that z-moves still work as expected. Other than that, yes :)

[szymonWojdat]

Just added a simple test for dynamax. Do you mind explaining what do you mean by "be sure that z-moves work as expected"? Were you refering to the fact that those were gone for a while and then moved back in? If so, then those commits aren't even in the tree anymore, so I'd expect them to work normally. I can also test those manually against my local PS instance or write a simple unit test if necessary.

[hsahovic]

I was referring to this:

• Move.can_z_move doesn't work as expected as moves.json no longer contains zMovePower for all damage moves, my guess is that PS has switched to some sort of dynamic calculation of the z-move power. The only place where it's used in code is at Pokemon.available_z_moves(). I assume the easiest workaround would be to modify Move.can_z_move so that it checks if given moves appears on a small list of moves that cannot z-move (test_can_z_move() implies as if Metronome could not z-move but it actually can, it just doesn't have any extra effect, moves that can't z-moves would be eg. Struggle)
• for the same reason as mentioned above, test_z_move_power is failing as well. I'll try taking a look at their codebase to see what's the formula that they're using

[szymonWojdat]

Thanks! I fixed both of those. Also added two pyre-ignores.

Now integration tests are failing. test_max_concurrent_battle_works is timing out. Do you mind taking a look? I guess we'll also need to add an integration test for gen8randombattle

[hsahovic]

@szymonWojdat Yes, sure! I just need to finish a couple of uni-related things; I'll get back into poke-env by the end of the week