Example project on embedding a JavaScript engine in a rust project.
What we want is to embed a script engine which can receive events which occur in our rust app and add functionality by using javascript or typescript.
The projects we'll be using to achieve this are the following
- quickjs_runtime the quickjs wrapper library
- hirofa_utils ScriptEngine abstraction layer and utilities
- GreenCopperRuntime adds more features to JavaScript engine like fetch and module loaders
- typescript_utils Typescript transpiler (uses SWC)
You can read more about the quickjs_runtime project inner workings here https://hirofa.github.io/quickjs_es_runtime/quickjs_runtime/index.html
Our sample application will be a simple actix based webapp with a simple request handler
Async support will be provided by using tokio
[package]
name = "my_app"
version = "0.1.0"
authors = ["Andries Hiemstra <[email protected]>"]
edition = "2018"
[dependencies]
tokio = "1"
actix-web = "4.0.0-rc.3"
log = "0.4"
use actix_web::{web, App, HttpRequest, HttpResponse, HttpServer};
async fn index(_req: HttpRequest) -> HttpResponse {
HttpResponse::Ok().body("hello there")
}
#[actix_web::main]
async fn main() -> std::io::Result<()> {
HttpServer::new(|| App::new().service(web::resource("/").to(index)))
.bind(("0.0.0.0", 8070))?
.run()
.await
}
We add the following dependencies to our Cargo.toml
[dependencies]
...
quickjs_runtime = "0.7.1"
hirofa_utils = "0.4"
lazy_static = "1.4.0"
...
And we init a runtime as a lazy static in our main.rs
...
use lazy_static::lazy_static;
use quickjs_runtime::builder::QuickJsRuntimeBuilder;
use quickjs_runtime::facades::QuickJsRuntimeFacade;
lazy_static! {
static ref SCRIPT_RT: QuickJsRuntimeFacade = init_quickjs();
}
fn init_quickjs() -> QuickJsRuntimeFacade {
let rt = QuickJsRuntimeBuilder::new().build();
// init code here
return rt;
}
...
In order to give our script something to communicate with we'll define a Proxy class
This proxy class will be used to dispatch events and add functions for our script file to invoke.
A bit of terminology: when working with the thread-safe objects you work with facades (JsRuntimeFacade, JsValueFacade) when working directly with Script objects ion the worker thread of the script engine you'll be working with Adapters(JsRuntimeAdapter, JsRealmAdapter, JsValueAdapter). These adapter cannot be moved out of the worker thread. To pass values from and to teh worker thread you need to convert them to JsValueFacades using JsRealmAdapter.to_js_value_facade()
and JsRealmAdapter.from_js_value_facade()
in main.rs
we add the following:
...
use hirofa_utils::js_utils::adapters::proxies::JsProxy;
use hirofa_utils::js_utils::adapters::{JsRealmAdapter, JsValueAdapter};
use hirofa_utils::js_utils::facades::JsRuntimeFacade;
use hirofa_utils::js_utils::JsError;
...
fn init_quickjs() -> QuickJsRuntimeFacade {
...
// to install out proxy we add a job to the RuntimeFacade
// we won't use multiple realms so we pass None as realm_name, this will make the runtime use the main realm (or context)
rt.js_loop_realm_sync(None, |_rt, realm| {
init_proxy(realm)?;
let res: Result<(), JsError> = Ok(());
res
})
.ok()
.expect("init proxy failed");
...
}
fn init_proxy<R: JsRealmAdapter>(realm: &R) -> Result<(), JsError> {
let proxy = JsProxy::new(&["com", "mycompany"], "MyApp")
// out proxy wil have a single static method printSomething
.add_static_method("printSomething", |_rt, realm: &R, args| {
// if first arg is a string, log that string
if args[0].js_is_string() {
println!("script printed: {}", args[0].js_to_str()?)
}
// return undefined
realm.js_undefined_create()
})
// setting the static_event_target to true means we can dispatch events and and add listeners from script
// by calling com.mycompany.MyApp.addEventListener()
.set_static_event_target(true);
// we set add_global_var to true so there will be a com.mycompany.MyApp usable in script
realm.js_proxy_install(proxy, true)?;
Ok(())
}
...
Now in order to actually get something to work we need to dispatch events to our script engine and respond to them from javascript.
in in main.rs
we add a function to dispatch an event and call it for every request
...
async fn do_dispatch() {
// for every request we add a job to the script engine and await until it is done
SCRIPT_RT
.js_loop_realm(None, |_rt, realm| {
// dispatch the request event to our proxy class, our event obj wil ne null for now
let event_obj = realm
.js_null_create()
.ok()
.expect("could not create event obj");
match realm.js_proxy_dispatch_static_event(
&["com", "mycompany"],
"MyApp",
"request",
&event_obj,
) {
Ok(_vetoed) => {
//
}
Err(err) => {
log::error!("could not dispatch event: {}", err);
}
}
})
.await;
}
async fn index(_req: HttpRequest) -> HttpResponse {
do_dispatch().await;
HttpResponse::Ok().body("hello there")
}
...
Now in order to respond to that event we add the following to our main.js
file:
com.mycompany.MyApp.addEventListener("request", (evt) => {
com.mycompany.MyApp.printSomething("Just letting you know javascript received your event loud and clear!");
});
and we run main.js
when starting actix in main.rs
...
#[actix_web::main]
async fn main() -> std::io::Result<()> {
SCRIPT_RT
.js_eval(None, Script::new("file://main.js", include_str!("main.js")))
.await
.ok()
.expect("main.js failed");
...
Now when we run our app and point our browser to http://localhost:8070/
the output in our console will be as follows:
Finished dev [unoptimized + debuginfo] target(s) in 0.08s
Running `target/debug/my_app`
script printed: Just letting you know javascript received your event loud and clear!
Eureka!
To get some more possibilities in our newly created script-enabled application we're going to add support for fetch and console.log
In order to keep quickjs_runtime as clean as possible I decided to add these things to a separate project called GreenCopper you know... because copper connects everything and when it rusts it turns green :)
Lets add GreCo to our Cargo.toml
The "com" feature includes fetch, the "features" feature includes the console, the "db" features includes mysql query utils
[dependencies]
...
green_copper_runtime = { git = 'https://github.com/HiRoFa/GreenCopperRuntime', branch="main", features = ["com", "features", "db"], default-features=false}
...
And then in main.rs
we add the GreCo features to our builder.
fn init_quickjs() -> QuickJsRuntimeFacade {
let mut builder = QuickJsRuntimeBuilder::new();
builder = green_copper_runtime::init_greco_rt(builder);
let rt = builder.build();
...
The fetch api should work the same as fetch as defined at MDN, it is currently a pretty minimal implementation but simple calls should work.
The console enables use of things like console.log
and console.debug
it uses the log
crate so you should be able to log things to log files by using .e.g simple-logging
in main.rs
...
#[actix_web::main]
async fn main() -> std::io::Result<()> {
#[cfg(not(debug_assertions))]
{
simple_logging::log_to_file("myapp.log", LevelFilter::Info)?;
}
#[cfg(debug_assertions)]
{
simple_logging::log_to_file("myapp.log", LevelFilter::Trace)?;
}
and in main.js
...
console.log("logging from javascript");
...
will result in this line in myapp.log
[00:00:02.263] (7fef18f37640) INFO JS_REALM:[__main__]: logging from javascript
quickjs_runtime has support for script preprocessors. in the code above you unknowingly added a cpp style preprocessor enabling you to do things like this
...
#IFDEF DEBUG
console.trace("logging some mighty large strings" + " which you don't want to construct in release mode");
#ENDIF
...
And that same principle allows us to add a typescript transpiler, this functionality is still in beta and therefore in a repo of its own at https://github.com/HiRoFa/typescript_utils
in order to enable this we need to add a dependency to our Cargo.toml
[dependencies]
...
typescript_utils = {git="https://github.com/HiRoFa/typescript_utils"}
...
and add it to our builder in main.rs
...
use typescript_utils::{TargetVersion, TypeScriptPreProcessor};
...
fn init_quickjs() -> QuickJsRuntimeFacade {
let tspp = TypeScriptPreProcessor::new(TargetVersion::Es2020, false, false);
let mut builder = QuickJsRuntimeBuilder::new().script_pre_processor(tspp);
The nice thing about using typescript is that we can define in script what our proxy looks like making it easier to develop scripts.
Now we can replace our main.js
with main.ts
don't forget to alter the init code which runs main.js
in main.rs
...
SCRIPT_RT
.js_eval(None, Script::new("file://main.ts", include_str!("main.ts")))
.await
.ok()
.expect("main.ts failed");
...
our main.ts looks like this
type MyApp = EventTarget & {
printSomething: (thing: string) => void
};
const myApp: MyApp = com.mycompany.MyApp;
com.mycompany.MyApp.addEventListener("request", (evt) => {
myApp.printSomething("Just letting you know javascript received your event loud and clear!");
console.log("logging from javascript");
});
Last but not least we're going to enable your scripters to include modules from the filesystem as well as online.
To achieve this we need to add two ModuleLoaders to our builder
...
use green_copper_runtime::moduleloaders::{FileSystemModuleLoader, HttpModuleLoader};
...
fn init_quickjs() -> QuickJsRuntimeFacade {
let tspp = TypeScriptPreProcessor::new(TargetVersion::Es2020, false, false);
// a filesystem module loader which loads modules form ./modules
let fsml = FileSystemModuleLoader::new("./modules");
// a http module loader which is allowed to load modules via https only from github.com
let html = HttpModuleLoader::new()
.secure_only()
.allow_domain("https://github.com");
let mut builder = QuickJsRuntimeBuilder::new()
.script_pre_processor(tspp)
.js_script_module_loader(fsml)
.js_script_module_loader(html);
builder = green_copper_runtime::init_greco_rt(builder);
let rt = builder.build();
...
now in main.ts
we can load modules like this
import {calc} from 'ModuleA.ts';
console.log("7*8=%s", calc(7, 8));
...
This does however mean we need to run main.ts as a module and not just as a script file so in main.rs
we alter our init code
...
SCRIPT_RT
.js_eval_module(None, Script::new("file://main.ts", include_str!("main.ts")))
.await
.ok()
.expect("main.ts failed");
...
I hope this gives you an idea of how to implement a javascript or typescript engine in your rust project. When going forward you may want to look at doing things async and awaiting the resulting promises instaed of firing sync events.
Have fun!