Sokoban for Java (using Swing) tailored for casual playing but especially for creating custom Sokoban agents. Fully playable but truly meant for programmers for the development of Sokoban artificial players.

Supports load of .sok packages from sokobano.de; many .sok packages downloaded within Sokoban4J/levels/sokobano.de folder. Thank you for such a wonderful collection!

Art used under CC-BY-SA-4.0 downloaded from OpenGameArt created by 1001.com; thank you!

LICENSED UNDER CC-BY-SA-4.0 Please retain URL to the Sokoban4J in your work.

What are benefits over using JSoko? The answer is simplicity; this project does not contain any technical commplexities (no applet, no need for the installer, no extra framework overhead in agent implementation, much more smaller code base).

1. Four different representation of the game state - see Board for OOP representation used by the simulator and then BoardCompact (use CMove and CPush to update), BoardSlim (use SMove and SPush to update) and BoardCompressed (use MMove and MPush to update) that should be used for state space search. Use StateCompressed or StateMinimal for the representation of no-good states.

2. HumanAgent and ArtificialAgent stubs; ArtificialAgent is using own thread for thinking (does not stuck GUI). See DFSAgent from Sokoban4J-Agents project for the example of artificial player.

3. Character is animated; human may use arrows or WSAD to control the avatar.

4. Possible to run headless simulations (same result as visualized, only faster).

5. Use "Sokoban" static methods for quick startups of your code (both for humans and artificial agents).

6. Simple Tree-DFS agent that can solve the simplest levels.

7. You can define up-to 6 different kinds of "boxes" (yellow, blue, gray, purple, red and black) and their specific target places; brown place (the brown dot) is "generic spot for any kind of box".

8. Mavenized (repo and dependency at the end of the page); uses some of my other stuff, but that can be easily cut off if you're considering branching.

9. Tested with Java 1.8, compilable with 1.6 as well; jars in Maven are compiled using Java 1.6.

10. Using pure Swing, no other 3rd-party libs = no complexities...

11. Load of .sok level packages from sokobano.de fully supported; many packages from https://sokobano.de/en/levels.php were downloaded into Sokoban4J/levels/sokobano.de (Thank you for such a wonderful collection!)

12. Large levels automatically scales down to fit the screen in order to be playable by humans.

13. Level play time may be limited (in milliseconds).

14. Possibility to assess Sokoban4J Java agents using Sokoban4J-Tournament project; tournament is executing separate JVMs per agent/level so it is possible to give an agent specific amount of memory for solving the levels.

15. Find introductory tips for creating Sokoban artificial player in this report (courtesy of Pavel Klavík).

Sokoban4J -> main project containing the simulator and visualizer of the game

Sokoban4J-Agents -> example artificial agents for Sokoban4J, so far only Tree DFS in here

Sokoban4J-Playground -> meant for easy hop-on the train of Sokoban agent development; fool around in here rather than Sokoban4J-Agents that should serve for you as backup ;)

Sokoban4J-Tournament -> meant for batch-assessing of your Java agents from console.

Compile Sokoban4J project (from within Sokoban4J directory), requires Maven:

Windows (from cmd; assuming you have mvn.bat on path):

mvn package

Linux (from bash, assuming you have mvn on path):

mvn package

<repository>
    <id>amis-artifactory</id>
    <name>AMIS Artifactory</name>
    <url>https://diana.ms.mff.cuni.cz:8081/artifactory/repo</url>
</repository>

<dependency>
    <groupId>cz.sokoban4j</groupId>
    <artifactId>sokoban4j</artifactId>
    <version>0.1.0-SNAPSHOT</version>
</dependency>

<dependency>
    <groupId>cz.sokoban4j</groupId>
    <artifactId>sokoban4j-tournament</artifactId>
    <version>0.1.0-SNAPSHOT</version>
</dependency>