JMagic: the Gathering

A Java implementation for an environment which can play the game Magic: The Gathering.

Introduction

This code is licenced under MIT License. Check the attached LICENSE file.

This code was developed using Java 9 SDK. Please refer to the proper documentation.

Start Up

You can have a random game by executing the Runner class. Tweak its attributes to personalize your game instance. For instance:

• change the parameter seconds in HumanObserver constructor to slow or speed the game up
• modify Runner.N_CARDS from 40 to 80 to have all players starting the game with 80 cards on their decks
• remove AttachAction.class entry from the LooseOnIllegalActionAttempt instantiation in Runner.OBSERVERS to prevent all players from attaching cards

For more flexible behavior, check the constructors available at Game class.

Basic Mechanics

A Game represents a game of Magic. It contains the current state of the world (an instance of State), the players in the game and observers (instances of Observer), which are responsible for keeping the game consistent.

At every iteration, the Game object will retrieve who's the active Player in the current game state and ask them for an Action that can modify its State.

Players

Like in a real match, players are required to do ALL the work. That includes drawing, passing their turn and even computing the damage. These actions can be performed by simply instantiating an object of Action and returning it to the Game object, which will pass it onto its observers for validation and, finally, commit the action.

To play a game using your very own player, simply create a class extending from Player and implement the method Action YourPlayerClass#act(State state).

Basic Players

If you want to make your life a little bit easier when creating your own player, just extend one of the classes in jmagic.players package and override the Player#act(State state) method. With that, handle the game states you want and delegate the rest to the superior player. E.x.:

class MySimplePlayer extends RandomPlayer {
    @Override
    public Action act(State state) {
        if (state.step == TurnSteps.DECLARE_BLOCKERS
        && !(state.actionThatLedToThisState instanceof DeclareBlockersAction)) {
            Map<Creature, Creature> blockers;
            // Some fancy computation to declare blockers s.t. the number of
            // creatures killed is minimized...
            return new DeclareBlockersAction(blockers);
        }

        // Don't know how to handle this state. Delegate to `RandomPlayer`.
        return super.act(state);
    }
}

Complete Player

If you want to handle ALL cases, you can simply extend Player.

Suppose State current is the current state of the game, MyPlayer dylan is the current turn's player and it's his drawing phase. The Game will ask dylan for an Action by calling the dylan.act(current) method.

An implementation that would correctly draw a card from dylan's deck is this:

class MyCompletePlayer extends Player {
    @Override
    public Action act(State state) {
        if (state.turnsPlayerState().player.equals(this)
            && state.step == TurnSteps.DRAW
            && new HasNotAlreadyDrawnInThisTurn().isValid(state)) {
            return new DrawAction();
        }

        // Don't know what to do. Advance game.
        return AdvanceGameAction();
    }
}

Explanation: although we know for a fact that IT IS Dylan's drawing phase, Player#act(State state) is called in many other stages of the game. Dylan must, therefore, check if:

1. He is the active and turn's player (the player that has the upkeep)
2. This is the drawing phase
3. He didn't draw already.

If all of these conditions are true, he will return an instance of DrawAction, which will be read by the Game object and finally modify the game state. Otherwise, Dylan won't know what to do and will simply request the game to advance with AdvanceGameAction.

Note: although this implementation correctly draws from the deck, it does not cover many other important actions. Read the implementations in the players package for more insight.

Docs

This section describes the intertwining of the elements in this code.

Actions

An Action is the entity capable of modifying the match's current State. An action's implementation of the abstract method State Action#update(State state) will determine how the state is altered.

At every iteration of the Game, an action is requested from the active Player by calling Player#act(State state). The active Player therefore must implement this contract and act accordingly to the current state of the game, described by its parameter State state.

For more information on how actions work, check out the many examples in the actions package.

Observers

Player are not the only entity that can execute actions over a game state. Twice every iteration (before the active player's action and after it), the game will deliver the current state to each observer, which in turn will modify it at will.

Passing a List of Observers to a Game constructor is a way to add constrains to that game. For example:

Game game = new Game(players, playersCards, playerActTimeout, List.of(
        new LooseIfDrawingFromEmptyDeck(),
        new LooseOnActTimeout(3000)
));

This game will disqualify players that attempt to draw from an empty deck and the ones that failed to return with an Action in less than 3 seconds.

For more information on how observers work, check out the many examples in the observers package.

Validation

Not all actions are valid all the time. For instance, you cannot draw a card outside your own drawing phase unless you have a spell that explicitly allows you to do so. That's when ValidationRule becomes interesting: by sub-classing it and defining an implementation for ValidationRule#onValidate(State state), we can create an rule that checks whether or not the current state is valid. Furthermore, we can add to error messages to ValidationRule#errors to better inform users why that state is invalid.

Let's illustrate the concept with an example. Say we create a "rule that asserts that the game is in a given turn-step". The implementation is quite easy:

public class TurnsStepIs extends ValidationRule {

    private final TurnSteps step;

    public TurnsStepIs(TurnSteps step) {
        this.step = step;
    }

    @Override
    public void onValidate(State state) {
        if (state.step != this.step) {
            errors.add(String.format("Incorrect turn's step (expected: %s, actual: %s)",
                this.step.name(), state.step.name()));
        }
    }
}

We can now use TurnStepIs rule when defining our DrawAction:

public final class DrawAction extends Action {

    @Override
    public State update(State state) {
        // Logic to draw card...
    }

    @Override
    public ValidationRule validationRules() {
        return new TurnsStepIs(TurnSteps.DRAW),
    }
}

Before applying every action given by a player, the Game instance will use an observer to check if that action is valid (by simply checking action.validationRules().validate(state).isValid()). If it's not, the player will either automatically pass or be disqualified.

In the example above, DrawAction will only be valid during DRAW turn steps!

Finally, you can compose rules using a few connectives in infrastructure.validation.basic, such as And, Or and Not:

import static Connectives.*;

public final class DrawAction extends Action {

    // Rest of DrawAction's code...

    @Override
    public ValidationRule validationRules() {
        return And(
            new HasCardsInTheirDeck(),
            new TurnsStepIs(TurnSteps.DRAW),
            new ActiveAndTurnsPlayersAreTheSame(),
            new HasNotAlreadyDrawnInThisTurn());
    }
}

Notice that these connectives are merely sub-classes of ValidationRule, and the static methods in Connectives class are just aliases to the construction of a connective object.