Skip to content

darmie/cabasa

Repository files navigation

Cābāsā

Cābāsā is a haxe framework for WebAssembly which includes a fast and secure WebAssembly VM (which is a haxe port of Life VM).

[Status: Not Ready For Use!]

Features

  • Fast - Includes a fast interpreter and an AOT compiler
  • Secure - Executed code is fully sandboxed, wasm modules do not have access to host resources unless explicitly given.
  • Portable - Does not rely on native dependencies and can redistribute wasm in AOT mode as normal haxe code.
  • Cross-Platform - Cābāsā aims to take full advantage of Haxe's cross platform capabilities, therefore wasm module can go anywhere Haxe goes.

Getting Started

Install package manager

npm i -g lix

Use latest Haxe version

lix use haxe 4.0.0-rc.5

Download dependencies

lix download

Executing WebAssembly Modules

Pass a wasm bytcode into a newly instantiated VM

import cabasa.exec.*;

...
var config:VMConfig = {...};
var resolver:ImportResolver = new MyImportResolver(); // handle function and global imports

var vm = new VM(code, config, resolver);

Look up an exported funtion from wasm code by its name, it will return its ID to be used to call the function later

var funcID = vm.getFunctionExport("main"); // could be the name of any exported function

Now call the function by its ID

var data = vm.run(funcID);
if(data.err != null){
    // do something with error
}
Sys.println('return value ${data.result}');

Implementing an Import Resolver

Assume a code compiled to wasm calls an virtual (or native) function __graphics_drawCirle(radius, color), we need to resolve for this function in the host application.

We create an ImportResolver class:

import cabasa.exec.*;

class MyImportResolver implements ImportResolver {
    ...
    public function resolveFunc(module:String, field:String):FunctionImport {
        return switch module {
            case 'env':{
                switch field:{
                    case '__graphics_drawCirle': function(vm:VM):I64 {
                            // get the local variables or function params
                            var radius:U32 = vm.getCurrentFrame().locals[0]; 
                            var color:U32 = vm.getCurrentFrame().locals[1]; 

                            // call the equivalent of the function in the host app
                            myhost.app.Graphics.drawCircle(cast radius, cast color); 

                            return 0;
                        };
                    }
                    default: throw 'cannot find field $field in module $module';
                }
            }
            default: throw 'module $module not found in host';
        }
    }
    // just like resolveFunc but returns the ID of the global export
    public function resolveGlobal(module:String, field:String):I64 {
        throw "not implemented for now"; 
    }
}

Now use this resolver in a VM instance

var vm = new VM(code, {...}, new MyImportResolver());

Run Test

#compile
haxe build.hxml

#run jar output
java -jar build/java/Test-Debug.jar

#run exe output
build/cs/bin/Test-Debug.exe

Why Cābāsā ?

I personally found other haxe embedded scripting interfaces quite limiting. CPPIA only works in C++ targets, HL only work in Hashlink VM just like Neko runs only in its VM. hscript is nice, it's crossplatform but it is a stripped down version of haxe, taking away useful features like types, OOP and modular imports (I honestly tried to fix this with hxComposer), it is at best just for expressions.

WebAssembly shows promise, apart from its much advertise use of redistributing native libraries to the web, it has drawn the attention of server side developers who feel the need to run self contained apps with all the advantages promised by WebAssembly. The WebAssembly specification describes a stack based immediate language with promise of a faster load time, smaller and portable binary format, and memory-safe execution environment.

Haxe + WebAssembly = ⚡ ⚡ , a combined force that will benefit all cross platform software engineers. With Haxe's cross platform abilities and wasm, software distribution accross platforms should be more fluid by reducing the need for platform specific glue code, which makes for a more portable software without compromise on quality and performance.

Dependencies

Supported Targets

  • C++
  • Java
  • C#

To-Do

  • Support runtime and AOT compilation for HashLink binary
  • Support runtime and AOT compilation for Neko binary
  • Support runtime and AOT compilation for CPPIA binary
  • Command Line Interface for running wasm module from terminal, disassemble wasm or compile wasm to Haxe AOT
  • Validate wasm binary
  • Extension framwork for distributing wasm modules as Haxe libraries
  • Examples

Releases

No releases published

Packages

No packages published

Languages