ECMAScript 5.1(+) implementation in Go.
Goja is an implementation of ECMAScript 5.1 in pure Go with emphasis on standard compliance and performance.
This project was largely inspired by otto.
Minimum required Go version is 1.14.
- Full ECMAScript 5.1 support (including regex and strict mode).
- Passes nearly all tc39 tests tagged with es5id. The goal is to pass all of them. Note, the last working commit is https://github.com/tc39/test262/commit/1ba3a7c4a93fc93b3d0d7e4146f59934a896837d. The next commit made use of template strings which goja does not support.
- Capable of running Babel, Typescript compiler and pretty much anything written in ES5.
- Sourcemaps.
- Some ES6 functionality, still work in progress, see https://github.com/cleancluster/goja/milestone/1?closed=1
Although it's faster than many scripting language implementations in Go I have seen (for example it's 6-7 times faster than otto on average) it is not a replacement for V8 or SpiderMonkey or any other general-purpose JavaScript engine. You can find some benchmarks here.
It greatly depends on your usage scenario. If most of the work is done in javascript (for example crypto or any other heavy calculations) you are definitely better off with V8.
If you need a scripting language that drives an engine written in Go so that you need to make frequent calls between Go and javascript passing complex data structures then the cgo overhead may outweigh the benefits of having a faster javascript engine.
Because it's written in pure Go there are no cgo dependencies, it's very easy to build and it should run on any platform supported by Go.
It gives you a much better control over execution environment so can be useful for research.
No. An instance of goja.Runtime can only be used by a single goroutine at a time. You can create as many instances of Runtime as you like but it's not possible to pass object values between runtimes.
setTimeout() assumes concurrent execution of code which requires an execution environment, for example an event loop similar to nodejs or a browser. There is a separate project aimed at providing some NodeJS functionality, and it includes an event loop.
Some ES6 functionality has been implemented. So far this is mostly built-ins, not syntax enhancements. See https://github.com/cleancluster/goja/milestone/1 for more details.
The ongoing work is done in the es6 branch which is merged into master when appropriate. Every commit in this branch represents a relatively stable state (i.e. it compiles and passes all enabled tc39 tests), however because the version of tc39 tests I use is quite old, it may be not as well tested as the ES5.1 functionality. Because ES6 is a superset of ES5.1 it should not break your existing code.
I will be adding features in their dependency order and as quickly as my time allows. Please do not ask for ETAs. Features that are open in the milestone are either in progress or will be worked on next.
Before submitting a pull request please make sure that:
- You followed ECMA standard as close as possible. If adding a new feature make sure you've read the specification, do not just base it on a couple of examples that work fine.
- Your change does not have a significant negative impact on performance (unless it's a bugfix and it's unavoidable)
- It passes all relevant tc39 tests.
- There should be no breaking changes in the API, however it may be extended.
- Some of the AnnexB functionality is missing.
vm := goja.New()
v, err := vm.RunString("2 + 2")
if err != nil {
panic(err)
}
if num := v.Export().(int64); num != 4 {
panic(num)
}
Any Go value can be passed to JS using Runtime.ToValue() method. See the method's documentation for more details.
A JS value can be exported into its default Go representation using Value.Export() method.
Alternatively it can be exported into a specific Go variable using Runtime.ExportTo() method.
Within a single export operation the same Object will be represented by the same Go value (either the same map, slice or a pointer to the same struct). This includes circular objects and makes it possible to export them.
There are 2 approaches:
- Using AssertFunction():
vm := New()
_, err := vm.RunString(`
function sum(a, b) {
return a+b;
}
`)
if err != nil {
panic(err)
}
sum, ok := AssertFunction(vm.Get("sum"))
if !ok {
panic("Not a function")
}
res, err := sum(Undefined(), vm.ToValue(40), vm.ToValue(2))
if err != nil {
panic(err)
}
fmt.Println(res)
// Output: 42
- Using Runtime.ExportTo():
const SCRIPT = `
function f(param) {
return +param + 2;
}
`
vm := New()
_, err := vm.RunString(SCRIPT)
if err != nil {
panic(err)
}
var fn func(string) string
err = vm.ExportTo(vm.Get("f"), &fn)
if err != nil {
panic(err)
}
fmt.Println(fn("40")) // note, _this_ value in the function will be undefined.
// Output: 42
The first one is more low level and allows specifying this value, whereas the second one makes the function look like a normal Go function.
By default, the names are passed through as is which means they are capitalised. This does not match the standard JavaScript naming convention, so if you need to make your JS code look more natural or if you are dealing with a 3rd party library, you can use a FieldNameMapper:
vm := New()
vm.SetFieldNameMapper(TagFieldNameMapper("json", true))
type S struct {
Field int `json:"field"`
}
vm.Set("s", S{Field: 42})
res, _ := vm.RunString(`s.field`) // without the mapper it would have been s.Field
fmt.Println(res.Export())
// Output: 42
There are two standard mappers: TagFieldNameMapper and UncapFieldNameMapper, or you can use your own implementation.
In order to implement a constructor function in Go:
func MyObject(call goja.ConstructorCall) *Object {
// call.This contains the newly created object as per http://www.ecma-international.org/ecma-262/5.1/index.html#sec-13.2.2
// call.Arguments contain arguments passed to the function
call.This.Set("method", method)
//...
// If return value is a non-nil *Object, it will be used instead of call.This
// This way it is possible to return a Go struct or a map converted
// into goja.Value using runtime.ToValue(), however in this case
// instanceof will not work as expected.
return nil
}
runtime.Set("MyObject", MyObject)
Then it can be used in JS as follows:
var o = new MyObject(arg);
var o1 = MyObject(arg); // same thing
o instanceof MyObject && o1 instanceof MyObject; // true
Goja uses the embedded Go regexp library where possible, otherwise it falls back to regexp2.
Any exception thrown in JavaScript is returned as an error of type *Exception. It is possible to extract the value thrown by using the Value() method:
vm := New()
_, err := vm.RunString(`
throw("Test");
`)
if jserr, ok := err.(*Exception); ok {
if jserr.Value().Export() != "Test" {
panic("wrong value")
}
} else {
panic("wrong type")
}
If a native Go function panics with a Value, it is thrown as a Javascript exception (and therefore can be caught):
var vm *Runtime
func Test() {
panic(vm.ToValue("Error"))
}
vm = New()
vm.Set("Test", Test)
_, err := vm.RunString(`
try {
Test();
} catch(e) {
if (e !== "Error") {
throw e;
}
}
`)
if err != nil {
panic(err)
}
func TestInterrupt(t *testing.T) {
const SCRIPT = `
var i = 0;
for (;;) {
i++;
}
`
vm := New()
time.AfterFunc(200 * time.Millisecond, func() {
vm.Interrupt("halt")
})
_, err := vm.RunString(SCRIPT)
if err == nil {
t.Fatal("Err is nil")
}
// err is of type *InterruptError and its Value() method returns whatever has been passed to vm.Interrupt()
}
There is a separate project aimed at providing some of the NodeJS functionality.