UCI

UCI is an acronym for Universal Chess Interface. It is a standard for communication that competes with XBoard/Winboard. http://wbec-ridderkerk.nl/html/UCIProtocol.html makes communication a little easier, but it seems there's a nearly religious debate on which is better.

Here are some benefits to using UCI:

• It works with Chessbase 14
• Simple to use
• Fewer bugs in code
• Built from scratch rather than ad-hoc
• Flexible time controls
• Additional search information can be displayed

and to be fair, some downsides:

• Stateless
• Delegates some important decisions to the GUI
• Difficult to add new features on top of it
• Sends the whole move list each turn

Overall, UCI seems to be fairly popular and is worth considering for your next engine.

Why

People shouldn't waste their time implementing protocols that really should be libraries. With this, you can include it in your project, build your engine on top of it, and be able to focus on what matters - beating Stockfish.

Yeah, but why Rust?

Originally I had written part of this library in C++. However, considering how prone people are to errors in C++, I decided to choose a language that provides a little more safety to the user. Currently Rust is beating Go in the Computer Language Benchmark Game, so it was the winner. People who write chess engines care about performance. If you're not writing C/C++, Rust seems like the next best thing.

Will You Implement Winboard?

Yeah, probably.

Why GPL 3.0?

I have noticed that there are no good free solutions so that anyone can build a chess engine. Since UCI is a common format, and something all engines (should) implement, making this chunk of it free makes total sense.

Usage

UCI is designed to complement your chess engine rather than be a one-stop shop for engine development.

As such, you will be responsible for maintaining the game loop. What UCI provides is a series of functions that will help you send data to, and receive data from the GUI.

First Steps

Your engine will first need to wait on STDIN for a command uci from the GUI. This indicates that your engine should switch to UCI mode. You can use commands::UCI to make sure the command text you are waiting for is correct.

Next, you'll need to create a copy of Engine by calling Engine::new. Once this is setup, you can call Engine::identify to send identification information to the GUI.

Once identification is done, you need to send your configuration options. This is dependent on your engine. Refer to the UCI standard for the available options. Here is an example of an option configuration:

  let o1 = EngineOption {
    name: constants::HASH,
    option_type: EngineOptionType::Spin,
    option_data: [(EngineOptionDataType::DefaultVal, EngineOptionData::Int(1)),
     (EngineOptionDataType::Min, EngineOptionData::Int(1)),
     (EngineOptionDataType::Max, EngineOptionData::Int(128))
    ].iter().cloned().collect();
  };

As nice as it would be to have fully typechecked options you will need to be careful a little here. The available options your engine uses must be passed to Engine::new as an array of Options::EngineOption<T> from options.rs. Every Options::EngineOption<T> has a name, which you can use any of the Options::* constants to represent, a type, which can be an Options::EngineOptionType, and a hashmap of Options::EngineOptionDataType to Options::EngineOptionDataValue<T>. Refer to the standard for more information on option configurations. The T type parameters represents the type of that particular Options::EngineOption<T> and allows us to at least fix the type in the HashMap to a single type. You can still hurt yourself, but doing this makes it much harder. Keep this list of options close, the GUI may send back information after call Engine::send_available_options to reconfigure default settings.

You must configure this before calling Engine::new so the rest of this guide assumes you've done that already.

To send your configuration options simply call Engine::send_available_options. Once this finishes uciok will also be sent, indicating to the GUI your engine is ready to roll. At this point you need to set up two threads, one to calculate with your engine, and one to read STDIN.

EXAMPLE HERE WITH LOOPS IN THREADS AND BLOCKING AND WHATEVER

Notice how in our STDIN thread we are calling Engine::parse and the handling the output using a match statement depending on what kind of token it was. You are responsible for obeying the commands from this thread, the UCI library just makes it convenient to work with. Here is an example of handling a parsed command:

EXAMPLE OF HANDLING A PARSED COMMAND AND SENDING BACK SOMETHING

At this point the engine will send a series of setoption commands to you in your STDIN thread. We can use the Engine::parse function to get these:

EXAMPLE HERE

Once the options are extracted you will be responsible for configuring your engine with the values. The GUI will send an isready command which you will be responsible for replying to by using Engine::ready.

TODO: MORE STUFF WITH EXAMPLES THINGS LIKE SENDINB BEST MOVE AFTER CALCULATING, ETC RECEIVING SETOPTION COMMANDS AND PROCESSING THEM

Other Options

TODO: Talk about the additional helpers available in the UCI library and what-not.

• Copy protection checking
• Registration checking

** TODO: Put an example engine under /examples that does nothing but talks to the GUI and receives commands. ** Document it here and mention it can be used for guidance.

Other Libraries in the works

In the future more libraries will be available to help chess engine developers get started. I will be writing a library to give some standard implementations of Zobrist Hashing, among other useful tools that are commonly re-implemented a thousand times for every engine.