Skip to content


Switch branches/tags


Failed to load latest commit information.
Latest commit message
Commit time



Write command line tools with ease Haxe.


To illustrate the usage, let's look at the follow quick mock-up of the Haxe command line.

static function main() {
	Cli.process(Sys.args(), new MockupHaxe()).handle(Cli.exit);

class MockupHaxe {
	public var js:String;
	public var lib:Array<String>;
	public var main:String;
	public var defines:Array<String>;
	public var help:Bool;
	public var helpDefines:Bool;
	public function new() {}
	public function run(rest:Rest<String>) {
		Sys.println('js: $js');
		Sys.println('lib: $lib');
		Sys.println('main: $main');
		Sys.println('defines: $defines');
		Sys.println('help: $help');
		Sys.println('helpDefines: $helpDefines');
		Sys.println('rest: $rest');

And then run with:

./ -js bin/run.js -D release -D mobile -lib tink_core -lib tink_json Main

Gives you:

js: bin/run.js
lib: [tink_core,tink_json]
main: null
defines: [release,mobile]
help: null
helpDefines: null
rest: [Main]

or: ./ --help-defines

js: null
lib: null
main: null
defines: null
help: null
helpDefines: true
rest: []

Check out the examples folder for the complete code.


Every public var in the class will be treated as a cli flag. (To override this behavior, tag a field with @:flag(false))

For example public var flag:String will be set to value <x> by the cli switch --flag <x>. Also, the framework will also recognize the first letter of the flag name as alias. So in this case -f <x> will do the same.

You can use metadata data to govern the flag name (@:flag('my-custom-flag-name')) and alias (@:alias('a')). Note that you can only use a single character for alias.

The reason for a single-char restriction for alias is that you can use a condensed alias format, for example: -abcdefg is actually equivalent to -a -b -c -d -e -f -g.

If you specify a flag name starting with a single dash (e.g. @:flag('-flag')), it will be respected but then alias support will be automatically disabled. You can also use @:alias(false) to manually disable alias, which works on both field level and class level.


Public methods tagged with @:command will be treated as a runnable command.

For example @:command public function run() {} can be triggered by ./yourscript run. In case you want to run a function under your binary name, you can tag a function with @:defaultCommand, then you will be able to run the function with ./yourscript.

By default the framework infers the command name from the method name, you can provide a parameter to the metadata like @:command('my-cmd') to change that.

Also, if you tag a public var with @:command, it will be treated as a sub-command. For instance:

public var sub:AnotherCommand;

In this case, when the program is called with ./yourscript sub -a, the default command in AnotherCommand will run, with the argument -a

Data Types

By default, Bool, Int, Float and String are supported. Furthermore, you can use any abstract that is castable from String (i.e. from String or @:from public static function ofString(v:String)) as the data type.

For example, to use a Map:

abstract CustomMap(StringMap<Int>) from StringMap<Int> to StringMap<Int> {
	public static function fromString(v:String):CustomMap {
		var map = new StringMap<Int>();
		for(i in v.split(',')) switch i.split('=') {
			case [key, value]: map.set(key, (value:tink.Stringly));
			default: throw 'Invalid format';
		return map;

Also, note that if a flag is of type Bool, the flag will be set to true without considering the "switch argument", For example, --force is used instead of --force true to set force = true. In fact in the latter case, the true string is considered as a Rest argument.

Rest argument is a list of strings which are not consumed by the flag parser. You can capture it in a command with Rest<String>. For example:

public function run(rest:Rest<String>) {}

Note that at most one Rest argument may appear in the list.

User Input

Besides a non-interaction tool as described above. One can also build an interactive tool, by utilizing the Prompt interface.

It is basically:

interface Prompt {
	function prompt(type:PromptType):Promise<String>;

enum PromptType {
	MultipleChoices(prompt:String, choices:Array<String>);

Basically you ask for an input from the user, and then you will get a promised result. It is just an interface so you can basically implement any mechanism of input, from simple text input to "GUI" input with arrow movements, etc.

For now there is a SimplePrompt which basically read from the stdin and take in anything the user gives. And in case of multiple choice prompt, RetryPrompt will make sure the user are choosing from the provided list of choices, and fail after certain number of retries.

First, you set a Prompt instance in Cli.process(args, command, ?prompt) (defaults to a SimplePrompt instance if omitted) and then you can later obtain it back from a command's function like so:

public function run(prompt:tink.cli.Prompt) {
	prompt.prompt('Input your name: ').handle(...);


	var doc = Cli.getDoc(myCommand);


Write command line tools in Haxe








No packages published