Interaction with V8

Creating V8 Platform
From src/main.rs, we would find these 2 lines of code:
src/main.rs
let snapshot = snapshot::deno_snapshot();
let isolate = isolate::Isolate::new(snapshot, state, ops::dispatch);
This is where the V8 isolate is created. Going into the definition of Isolate::new in src/isolate.rs:
src/isolate.rs
pub fn new(
    snapshot: libdeno::deno_buf,
    state: Arc<IsolateState>,
    dispatch: Dispatch,
  ) -> Self {
    DENO_INIT.call_once(|| {
      unsafe { libdeno::deno_init() };
    });
    let config = libdeno::deno_config {
      will_snapshot: 0,
      load_snapshot: snapshot,
      shared: libdeno::deno_buf::empty(),
      recv_cb: pre_dispatch, // callback to invoke when Rust receives a message
    };
    let libdeno_isolate = unsafe { libdeno::deno_new(config) };
    // This channel handles sending async messages back to the runtime.
    let (tx, rx) = mpsc::channel::<(i32, Buf)>();
​
    Self {
      libdeno_isolate,
      dispatch,
      rx,
      tx,
      ntasks: Cell::new(0),
      timeout_due: Cell::new(None),
      state,
    }
  }
Here, we would find 2 calls on libdeno: deno_init and deno_new. These two functions, unlike many other functions used, are defined from libdeno/api.cc, and are made available thanks to FFI and libc. Their Rust interface is provided through src/libdeno.rs.
Go to libdeno/api.cc:
libdeno/api.cc
void deno_init() {
  auto* p = v8::platform::CreateDefaultPlatform();
  v8::V8::InitializePlatform(p);
  v8::V8::Initialize();
}
​
Deno* deno_new(deno_config config) {
  // ... code omitted
  deno::DenoIsolate* d = new deno::DenoIsolate(config);
  // ... code omitted
  v8::Isolate* isolate = v8::Isolate::New(params);
  // ... code omitted
​
  v8::Locker locker(isolate);
  v8::Isolate::Scope isolate_scope(isolate);
  {
    v8::HandleScope handle_scope(isolate);
    auto context =
        v8::Context::New(isolate, nullptr, v8::MaybeLocal<v8::ObjectTemplate>(),
                         v8::MaybeLocal<v8::Value>(),
                         v8::DeserializeInternalFieldsCallback(
                             deno::DeserializeInternalFields, nullptr));
    if (!config.load_snapshot.data_ptr) {
      // If no snapshot is provided, we initialize the context with empty
      // main source code and source maps.
      deno::InitializeContext(isolate, context);
    }
    d->context_.Reset(isolate, context);
  }
​
  return reinterpret_cast<Deno*>(d);
}
From these two functions, we would find that deno_init is used to initialize the V8 platform, while deno_new is used to create a new isolated VM instance on this platform. (A few V8 embedding APIs are used here. To learn more about V8 embedding, check out https://v8.dev/docs/embed for concepts, and https://denolib.github.io/v8-docs/ for API Reference.)
Adding Bindings
There are 2 important functions/constructors used in deno_new that might not be immediately clear: DenoIsolate and InitializeContext. It turns out DenoIsolate serves more or less as a collection of Isolate information. Instead, InitializeContext is the more interesting one. (It seems to be invoked here only when there is no snapshot provided. However, you'll also find the function being used in deno_new_snapshotter in libdeno/api.cc to create a new snapshot, so it is always an inevitable step):
libdeno/binding.cc
void InitializeContext(v8::Isolate* isolate, v8::Local<v8::Context> context) {
  v8::HandleScope handle_scope(isolate);
  v8::Context::Scope context_scope(context);
​
  auto global = context->Global();
​
  auto deno_val = v8::Object::New(isolate);
  CHECK(global->Set(context, deno::v8_str("libdeno"), deno_val).FromJust());
​
  auto print_tmpl = v8::FunctionTemplate::New(isolate, Print);
  auto print_val = print_tmpl->GetFunction(context).ToLocalChecked();
  CHECK(deno_val->Set(context, deno::v8_str("print"), print_val).FromJust());
​
  auto recv_tmpl = v8::FunctionTemplate::New(isolate, Recv);
  auto recv_val = recv_tmpl->GetFunction(context).ToLocalChecked();
  CHECK(deno_val->Set(context, deno::v8_str("recv"), recv_val).FromJust());
​
  auto send_tmpl = v8::FunctionTemplate::New(isolate, Send);
  auto send_val = send_tmpl->GetFunction(context).ToLocalChecked();
  CHECK(deno_val->Set(context, deno::v8_str("send"), send_val).FromJust());
​
  CHECK(deno_val->SetAccessor(context, deno::v8_str("shared"), Shared)
            .FromJust());
}
If you are familiar with the libdeno API on the TypeScript end, the above names wrapped in deno::v8_str might sound familiar to you. In fact, this is where the very few C++ bindings onto JavaScript are attached: we get the global object of current Context (a V8 execution environment that allows separate code to run) and add some extra properties onto it, from C++.
Based on the code above, whenever you call libdeno.send(...) from TypeScript (you'll find such usage in sendInternal of js/dispatch.ts), you are actually calling into a C++ function called Send. Similar things happens to libdeno.print.
Another example: console.log(...) is defined as
js/console.ts
  log = (...args: any[]): void => {
    this.printFunc(stringifyArgs(args));
  };
This printFunc is a private field in class Console, and its initial value is provided through
js/globals.ts
window.console = new consoleTypes.Console(libdeno.print);
Okay, we find libdeno.print here. From the above bindings code, we know that calling to libdeno.print from TypeScript is equivalent to calling Print in libdeno/binding.cc, inside of which we would discover
void Print(const v8::FunctionCallbackInfo<v8::Value>& args) {
  // ... code omitted
  v8::String::Utf8Value str(isolate, args[0]);
  bool is_err =
      args.Length() >= 2 ? args[1]->BooleanValue(context).ToChecked() : false;
  FILE* file = is_err ? stderr : stdout;
  fwrite(*str, sizeof(**str), str.length(), file);
  fprintf(file, "\n");
  fflush(file);
}
Nothing more surprising than some familiar standard C print formula. Therefore, calling console.log is in fact just indirectly calling fwrite/fprintf!
Check out the source code of Print, Send and Recv in libdeno/binding.cc to understand what is happening behind the scene.
Executing Code on V8
From src/main.rs, immediately following isolate creation, we find
src/main.rs
isolate
      .execute("denoMain();")
      .unwrap_or_else(print_err_and_exit);
From the previous section, we know denoMain is a TypeScript side function. isolate.execute is used to run the code. Let's extract its definition from src/isolate.rs:
src/isolate.rs
  /// Same as execute2() but the filename defaults to "<anonymous>".
  pub fn execute(&self, js_source: &str) -> Result<(), JSError> {
    self.execute2("<anonymous>", js_source)
  }
​
  /// Executes the provided JavaScript source code. The js_filename argument is
  /// provided only for debugging purposes.
  pub fn execute2(
    &self,
    js_filename: &str,
    js_source: &str,
  ) -> Result<(), JSError> {
    let filename = CString::new(js_filename).unwrap();
    let source = CString::new(js_source).unwrap();
    let r = unsafe {
      libdeno::deno_execute(
        self.libdeno_isolate,
        self.as_raw_ptr(),
        filename.as_ptr(),
        source.as_ptr(),
      )
    };
    if r == 0 {
      let js_error = self.last_exception().unwrap();
      return Err(js_error);
    }
    Ok(())
  }
The only interesting function we care in this section is libdeno::deno_execute. We can find its actual definition in libdeno/api.cc again:
libdeno/api.cc
int deno_execute(Deno* d_, void* user_data, const char* js_filename,
                 const char* js_source) {
  auto* d = unwrap(d_);
  // ... code omitted
  auto context = d->context_.Get(d->isolate_);
  CHECK(!context.IsEmpty());
  return deno::Execute(context, js_filename, js_source) ? 1 : 0;
}
Eventually, let's find deno::Execute:
bool Execute(v8::Local<v8::Context> context, const char* js_filename,
             const char* js_source) {
  // ... code omitted
  auto source = v8_str(js_source);
  return ExecuteV8StringSource(context, js_filename, source);
}
​
bool ExecuteV8StringSource(v8::Local<v8::Context> context,
                           const char* js_filename,
                           v8::Local<v8::String> source) {
  // ... code omitted
​
  v8::TryCatch try_catch(isolate);
​
  auto name = v8_str(js_filename);
​
  v8::ScriptOrigin origin(name);
​
  auto script = v8::Script::Compile(context, source, &origin);
​
  if (script.IsEmpty()) {
    DCHECK(try_catch.HasCaught());
    HandleException(context, try_catch.Exception());
    return false;
  }
​
  auto result = script.ToLocalChecked()->Run(context);
​
  if (result.IsEmpty()) {
    DCHECK(try_catch.HasCaught());
    HandleException(context, try_catch.Exception());
    return false;
  }
​
  return true;
}
As we see here, Execute is eventually submitting the code to v8::Script::Compile, which compiles the JavaScript code and call Run on it. Any exceptions, compile time or runtime, are further processed through HandleException:
libdeno/binding.cc
void HandleException(v8::Local<v8::Context> context,
                     v8::Local<v8::Value> exception) {
  v8::Isolate* isolate = context->GetIsolate();
  DenoIsolate* d = FromIsolate(isolate);
  std::string json_str = EncodeExceptionAsJSON(context, exception);
  CHECK(d != nullptr);
  d->last_exception_ = json_str;
}
It sets d->last_exception_ to be an error message formatted in JSON, which was read in deno_last_exception:
libdeno/api.cc
const char* deno_last_exception(Deno* d_) {
  auto* d = unwrap(d_);
  if (d->last_exception_.length() > 0) {
    return d->last_exception_.c_str();
  } else {
    return nullptr;
  }
}
which is used in Isolate::last_exception from Rust:
src/isolate.rs
pub fn last_exception(&self) -> Option<JSError> {
    let ptr = unsafe { libdeno::deno_last_exception(self.libdeno_isolate) };
    // ... code omitted
  }
and is checked inside of Isolate::execute to decide if error handling is necessary.
Whew! That's a long trip across TypeScript, C/C++ and Rust. However, it should be very clear to you how Deno is interacting with V8 now. Nothing magical.
Advanced - Previous
Process Lifecycle
Next - Advanced
DENO_DIR, Code Fetch and Cache
Last updated -1