Architecture

In order to achieve extensibility, skaty strictly separates the rule set from the game state.

Every game is represented by a game state with an associated rule set. A game's state is modified by applying/undoing actions following the Command Pattern.

Game State

The GameState class contains all state of a specific game instance.

It acts as a base class containing attributes used in any variation of Skat. It can be extended to contain more attributes if needed by the rule set (see tutorial Creating Rule Sets and Actions).

For example ISkOGameState extends GameState to contain, among others, information about the current bidding phase, the highest bid and last bid.

State modification

The game state should never be directly mutated by the user. Instead, state changes are delegated to actions and the rule set.

For example:

from skaty.isko.actions import Pass
from skaty.isko.rules import ISkO
from skaty.isko.state import ISkOGameState

game = ISkOGameState.from_random_deal(ISkO(), 0)
action = Pass(2)

# This will add the action to game.action_history, make sure it is reversible and advance the state.
game.apply_action(action)

# The action can also be undone easily
# Remove it from game.action_history, reverse state changes exactly.
game.undo_action()

Rule Set

The rule set, along with actions, formsw the core of the game logic. It acts as a static, stateless engine that defines h the game flows.

By default, skaty provides an AbstractRuleSet with the necessary interfaces (e.g. is_valid_action, get_valid_actions, advance_state etc.).

For usage in interanl methods, rule sets need no know about the game state it is applied to. Therefore, generics are used.

Actions

Actions represent discrete choices made by the players (e.g. PlayCard, DeclareBid).

Actions are stateless and immutable. The previous state is pushed to the GameState.undo_memory stack.