Implement ko for baduk

[benjaminpjones]

Implement ko for baduk

[benjaminpjones]

It's a bit complicated in that some types of variants (like parallel) may not want to have ko evaluated every move. I am now leaning toward a Ko-detector object that will throw if Ko is detected. This fits in with the throw-on-illegal-move paradigm we work with currently, and it's composable.

example:

const ko_detector = new KoDetector();
ko_detector.push({ a: 1 })
ko_detector.push({ a: 2 })
ko_detector.push({ a: 1 }) // KoError: board state repeated

[JonKo314]

Game state repetitions will be a problem in many variants. In parallel go they just look different. Instead of going back and forth like in a baduk ko, there might be no progress at all. Examples:

• Collisions that result in passes
• All placed stones are captured immediately

But we can always extend the base class with some shared repetition detection and inherit from that extension wherever it is convenient (or find other ways to reuse code).

[benjaminpjones]

Yep, I agree that ko will be needed in many variants. Feel free to open an issue for the ones that still need it! (parallel and maybe @merowin 's graph variant?)

I've tried to make KoDetector is relatively reusable:

• it is basically a 2-liner (one line for initialization and another for pushing state)
• can be placed anywhere
• each variant has full control over which subset of the state is relevant.