Comparison Brigadier and Bukkit Commands
Registering commands
The old Bukkit way
In order to register Bukkit commands, you would define a class that extends BukkitCommand, and implements the execute(...) and tabComplete(...) methods. This might look like this:
public class BukkitPartyCommand extends BukkitCommand {
public BukkitPartyCommand(@NotNull String name, @NotNull String description, @NotNull String usageMessage, @NotNull List<String> aliases) {
super(name, description, usageMessage, aliases);
}
@Override
public boolean execute(@NotNull CommandSender sender, @NotNull String commandLabel, @NotNull String[] args) {
if (args.length == 0) {
sender.sendMessage("Please provide a player!");
return true;
}
final Player targetPlayer = Bukkit.getPlayer(args[0]);
if (targetPlayer == null) {
sender.sendMessage("Please provide a valid player!");
return true;
}
targetPlayer.sendMessage(sender.getName() + " started partying with you!");
sender.sendMessage("You are now partying with " + targetPlayer.getName() + "!");
return false;
}
@Override
public @NotNull List<String> tabComplete(@NotNull CommandSender sender, @NotNull String alias, @NotNull String[] args) throws IllegalArgumentException {
if (args.length == 1) {
return Bukkit.getOnlinePlayers().stream().map(Player::getName).toList();
}
return List.of();
}
}
After that, you can define your command like this:
this.getServer().getCommandMap().register(
this.getName().toLowerCase(),
new BukkitPartyCommand("bukkitparty", "Have a party", "/party <player>", List.of())
);
As you can see, you have to do a lot of manual checking in order to register a single, very simple command. But how does the Brigadier api do it?
The new Paper way
First, we need to retrieve a LiteralCommandNode<CommandSourceStack>
. That's a special Brigadier class that holds some sort of command tree.
In our case, it is the root of our command. We can do that by running Commands.literal(final String literal)
, which returns a
LiteralArgumentBuilder<CommandSourceStack>
, where we can define some arguments and executors. Once we are done, we can call
LiteralArgumentBuilder#build()
to retrieve our build LiteralCommandNode
, which we can then register. That sounds complicated at first,
but once you see it in action, it looks less terrifying:
public static LiteralCommandNode<CommandSourceStack> createCommand(final String commandName) {
return Commands.literal(commandName)
.then(Commands.argument("target", ArgumentTypes.player())
.executes(ctx -> {
final PlayerSelectorArgumentResolver playerSelector = ctx.getArgument("target", PlayerSelectorArgumentResolver.class);
final Player targetPlayer = playerSelector.resolve(ctx.getSource()).getFirst();
final CommandSender sender = ctx.getSource().getSender();
targetPlayer.sendMessage(sender.getName() + " started partying with you!");
sender.sendMessage("You are now partying with " + targetPlayer.getName() + "!");
return Command.SINGLE_SUCCESS;
}))
.build();
}
Each .then(...)
defines a new branch in our tree, which can either be a literal (Commands.literal(String)
) or an argument
(Commands.argument(String, ArgumentType<T>)
). Each branch may or may not define an .executes(Command)
executor. This is
where all the logic happens.
We will take a closer look at that in different pages, but for now, how do we register it? Paper uses a LifecycleEventManager system. In a nutshell, that is a way to register commands (or tags) that get loaded each time the server reloads its resources, like using /reload. Registering our command looks like this:
this.getLifecycleManager().registerEventHandler(LifecycleEvents.COMMANDS, commands -> {
commands.registrar().register(PaperPartyCommand.createCommand("paperparty"), "Have a nice party");
});
And we are done! As you can see here, both commands do the same thing: