add test for SnakeGames (#106)

Author: findmyway

src/environments/3rd_party/snake.jl

@@ -1,3 +1,5 @@
+using .SnakeGames
+
 function SnakeGameEnv(; action_style = MINIMAL_ACTION_SET, kw...)
     game = SnakeGame(; kw...)
     n_snakes = length(game.snakes)
@@ -41,19 +43,20 @@ end
 
 RLBase.action_space(env::SnakeGameEnv) = 1:4
 RLBase.state(env::SnakeGameEnv) = env.game.board
+RLBase.state_space(env::SnakeGameEnv) = Space(fill(false..true, size(env.game.board)))
 RLBase.reward(env::SnakeGameEnv{<:Any,SINGLE_AGENT}) =
     length(env.game.snakes[]) - env.latest_snakes_length[]
 RLBase.reward(env::SnakeGameEnv) = length.(env.game.snakes) .- env.latest_snakes_length
 RLBase.is_terminated(env::SnakeGameEnv) = env.is_terminated
 
-RLBase.get_legal_actions(env::SnakeGameEnv{FULL_ACTION_SET,SINGLE_AGENT}) =
+RLBase.legal_action_space(env::SnakeGameEnv{FULL_ACTION_SET,SINGLE_AGENT}) =
     findall(!=(-env.latest_actions[]), SNAKE_GAME_ACTIONS)
-RLBase.get_legal_actions(env::SnakeGameEnv{FULL_ACTION_SET}) =
+RLBase.legal_action_space(env::SnakeGameEnv{FULL_ACTION_SET}) =
     [findall(!=(-a), SNAKE_GAME_ACTIONS) for a in env.latest_actions]
 
-RLBase.get_legal_actions_mask(env::SnakeGameEnv{FULL_ACTION_SET,SINGLE_AGENT}) =
+RLBase.legal_action_space_mask(env::SnakeGameEnv{FULL_ACTION_SET,SINGLE_AGENT}) =
     [a != -env.latest_actions[] for a in SNAKE_GAME_ACTIONS]
-RLBase.get_legal_actions_mask(env::SnakeGameEnv{FULL_ACTION_SET}) =
+RLBase.legal_action_space_mask(env::SnakeGameEnv{FULL_ACTION_SET}) =
     [[x != -a for x in SNAKE_GAME_ACTIONS] for a in env.latest_actions]
 
 function RLBase.reset!(env::SnakeGameEnv)

src/environments/wrappers/ActionTransformedEnv.jl

@@ -32,10 +32,11 @@ RLBase.state(env::ActionTransformedEnv, ss::RLBase.AbstractStateStyle) = state(e
 RLBase.state_space(env::ActionTransformedEnv, ss::RLBase.AbstractStateStyle) =
     state_space(env.env, ss)
 
-RLBase.action_space(env::ActionTransformedEnv) =
-    env.action_space_mapping(action_space(env.env))
-RLBase.legal_action_space(env::ActionTransformedEnv) =
-    env.action_space_mapping(legal_action_space(env.env))
+RLBase.action_space(env::ActionTransformedEnv, args...) =
+    env.action_space_mapping(action_space(env.env), args...)
+
+RLBase.legal_action_space(env::ActionTransformedEnv, args...) =
+    env.action_space_mapping(legal_action_space(env.env), args...)
 
 (env::ActionTransformedEnv)(action, args...; kwargs...) =
     env.env(env.action_mapping(action), args...; kwargs...)

src/environments/wrappers/DefaultStateStyle.jl

@@ -14,7 +14,7 @@ DefaultStateStyleEnv{S}(env::E) where {S,E} = DefaultStateStyleEnv{S,E}(env)
 RLBase.DefaultStateStyle(::DefaultStateStyleEnv{S}) where {S} = S
 
 for f in vcat(RLBase.ENV_API, RLBase.MULTI_AGENT_ENV_API)
-    if f ∉ (:DefaultStateStyle, :state, :state_space)
+    if f ∉ (:DefaultStateStyle, )
         @eval RLBase.$f(x::DefaultStateStyleEnv, args...; kwargs...) =
             $f(x.env, args...; kwargs...)
     end

src/environments/wrappers/StateOverriddenEnv.jl

@@ -4,6 +4,10 @@ export StateOverriddenEnv
     StateOverriddenEnv(f, env)
 
 Apply `f` to override `state(env)`.
+
+!!! note
+    If the meaning of state space is changed after apply `f`, one should
+    manually redefine the `RLBase.state_space(env::YourSpecificEnv)`.
 """
 struct StateOverriddenEnv{F,E<:AbstractEnv} <: AbstractEnvWrapper
     env::E
@@ -15,7 +19,7 @@ StateOverriddenEnv(f) = env -> StateOverriddenEnv(f, env)
 (env::StateOverriddenEnv)(args...; kwargs...) = env.env(args...; kwargs...)
 
 for f in vcat(RLBase.ENV_API, RLBase.MULTI_AGENT_ENV_API)
-    if f ∉ (:state, :state_space)
+    if f ∉ (:state, )
         @eval RLBase.$f(x::StateOverriddenEnv, args...; kwargs...) =
             $f(x.env, args...; kwargs...)
     end
@@ -26,5 +30,6 @@ RLBase.state(env::StateOverriddenEnv, args...; kwargs...) =
 
 RLBase.state(env::StateOverriddenEnv, ss::RLBase.AbstractStateStyle) =
     env.f(state(env.env, ss))
+
 RLBase.state_space(env::StateOverriddenEnv, ss::RLBase.AbstractStateStyle) =
-    state_space(env.env, ss)
+    state_space(env.env, ss)

test/Project.toml

@@ -4,6 +4,7 @@ OpenSpiel = "ceb70bd2-fe3f-44f0-b81f-41608acaf2f2"
 PyCall = "438e738f-606a-5dbb-bf0a-cddfbfd45ab0"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 ReinforcementLearningBase = "e575027e-6cd6-5018-9292-cdc6200d2b44"
+SnakeGames = "34dccd9f-48d6-4445-aa0f-8c2e373b5429"
 StableRNGs = "860ef19b-820b-49d6-a774-d7a799459cd3"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

test/environments/3rd_party/snake_game.jl

@@ -0,0 +1,4 @@
+@testset "snake game env" begin
+    env = SnakeGameEnv()
+    RLBase.test_runnable!(env)
+end

test/runtests.jl

@@ -4,7 +4,7 @@ using ReinforcementLearningEnvironments
 using ArcadeLearningEnvironment
 using PyCall
 using OpenSpiel
-# using SnakeGames
+using SnakeGames
 using Random
 using StableRNGs
 using Statistics