Module-level fields are here! - Haxe - The Cross-platform Toolkit

Module-level fields are here!

It is now possible to define functions and variables directly in Haxe modules

Article by Dan Korostelev on 2020-05-26.

Comments

Module-level fields are here!

I am happy to announce that we've just merged a new Haxe feature: module-level functions and variables (module fields, in short)!

This means that Haxe now supports declaring static fields outside of classes, directly in modules next to type declarations:

typedef Greeting = {phrase:String};

function greet(greeting:Greeting) {
    trace(greeting.phrase);
}

final defaultGreeting:Greeting = {phrase: "Hello, world!"};

function main() {
    greet(defaultGreeting);
}

The code above is more or less equivalent to the good old static fields:

typedef Greeting = {phrase:String};

class MyModule {
    public static function greet(greeting:Greeting) {
        trace(greeting.phrase);
    }

    public static final defaultGreeting:Greeting = {phrase: "Hello, world!"};

    public static function main() {
        greet(defaultGreeting);
    }
}

In fact, the compiler internally transforms such module-level declarations into static fields of an implicitly created class, so they behave the same way.

Using module-level fields

From a usage standpoint, module-level fields are not very different from static fields. However, since they are declared at the module level, they share some rules with the other kind of module-level declarations: types. The most important to keep in mind being:

  • module-level fields are publicly available by default, but can be explicitly made private
  • importing a module will import all its fields (together with all its types)

Macros

We can also spot this duality if we take a look at the new macro structures:

In the untyped AST (haxe.macro.Expr) there is a new TypeDefKind variant:

TDField(kind:FieldType, ?access:Array<Access>)

So we can define module-level fields in macros using the same TypeDefinition structure, together with Context.defineType or Context.defineModule.

It may be argued that with the introduction of module-level fields, the naming is a bit confusing, since it's not only about type definitions anymore, but at this point it is not worth breaking existing macro code by changing this.

In the typed AST (haxe.macro.Type) there is a new ClassKind variant:

KModuleFields(module:String)

It is one of the possible values of the ClassType.kind field, indicating that the class in question is a container for module-level fields defined for this module. And the fields themselves are available in the statics array.

This implementation allowed for introducing this feature in the least invasive and breaking way and it automatically works on all targets!

Code generation

Speaking of targets, on the generator side, they can choose to "flatten" KModuleFields classes back into series of function/variable declarations. This is already implemented for the JavaScript target, so if our example module is actually called MyModule, the generated code for the module-level fields from the example above will be:

function MyModule_main() {
    MyModule_greet(MyModule_defaultGreeting);
}
function MyModule_greet(greeting) {
    console.log("src/MyModule.hx:4:",greeting.phrase);
}
var MyModule_defaultGreeting = { phrase : "Hello, world!"};

Use cases

We believe this feature will make it more pleasant to develop code in non-OOP paradigms. An obvious example would be functional programming, which focuses more on functions and passive data structures, but there is room for class-less functions in any code, no matter the style.

Utilities and static extensions

Almost any codebase has helper methods that don't belong to a data type and now they can be defined without redundant syntax. Same goes for static extensions!

Macros

Haxe macros are often written in simple procedural style, which makes them good candidates to become module-level functions.

Scripts

Did you know it is possible to use Haxe as a scripting language? It only makes sense: the Haxe compiler already has a built-in interpreter for running macros (eval), so why not use it for general scripting? That is exactly what the --run compiler argument is for! And with the addition of module-level functions it's much easier to write such scripts. Here's one:

function main() {
    Sys.println(Sys.args().join(" "));
}

Run it with haxe --run Script hello world and you'll get a hello world. Neat! 👌

One particularly active member of the Haxe and Flixel communities is already making use of this feature for scripting, check out setup-flixel for some inspiration!

Closing words

Module-level fields will be available in the next minor Haxe release (4.2), but of course if you're living on the edge and don't want to wait, please grab a nightly build and try it out (and don't forget to report any issues you find!)

Avatar for Dan Korostelev

By Dan Korostelev

Published 2020-05-26