irondash_message_channel_derive

Crates.ioirondash_message_channel_derive
lib.rsirondash_message_channel_derive
version0.1.1
sourcesrc
created_at2022-11-15 18:16:34.889744
updated_at2022-11-15 18:16:34.889744
descriptionDerive proc-macros for Irondash
homepage
repository
max_upload_size
id715893
size52,007
Matej Knopp (knopp)

documentation

README

irondash_message_channel

Rust-dart bridge similar to Flutter's platform channel.

This package allows calling Rust code from Dart and vice versa using pattern similar to Flutter's platform channel.

  • Easy to use convenient API (Dart side mimics platform channel API).
  • High performance
    • Zero copy for binary data when calling Dart from Rust
    • Exactly one copy of binary data when calling Rust from Dart
  • Rust macros for automatic serialization and deserialization (similar to Serde but optimized for zero copy)
  • No code generation
  • Thread affinity - Rust channel counterpart is bound to thread on which the channel was created. You can have channels on platform thread or on any background thread as long as it's running a RunLoop.
  • Finalize handlers - Rust side can get notified when Dart object is garbage collected.
  • Async support

Usage

Initial setup

Because Rust code needs access to Dart FFI api some setup is required.

/// initialize context for Native library.
MessageChannelContext _initNativeContext() {
    final dylib = defaultTargetPlatform == TargetPlatform.android
        ? DynamicLibrary.open("libmyexample.so")
        : (defaultTargetPlatform == TargetPlatform.windows
            ? DynamicLibrary.open("myexample.dll")
            : DynamicLibrary.process());

    // This function will be called by MessageChannel with opaque FFI
    // initialization data. From it you should call
    // `irondash_init_message_channel_context` and do any other initialization,
    // i.e. register rust method channel handlers.
    final function =
        dylib.lookup<NativeFunction<MessageChannelContextInitFunction>>(
            "my_example_init_message_channel_context");
    return MessageChannelContext.forInitFunction(function);
}

final nativeContext = _initNativeContext();

// Now you can create method channels

final _channel =
    NativeMethodChannel('my_method_channel', context: nativeContext);

_channel.setMethodCallHandler(...);

Rust side:

use irondash_message_channel::*;

#[no_mangle]
pub extern "C" fn my_example_init_message_channel_context(data: *mut c_void) -> FunctionResult {
    irondash_init_message_channel_context(data)
}

Simple usage

After the setup, you can use the Dart NativeMethodChannel similar to Flutter's PlatformChannel:


final _channel = NativeMethodChannel('my_method_channel', context: nativeContext);

_channel.setMessageHandler((call) async {
    if (call.method == 'myMethod') {
        return 'myResult';
    }
    return null;
});

final res = await _channel.invokeMethod('someMethod', 'someArg');

On Rust side, you can implement the MethodHandler trait for non-async version, or AsyncMethodHandler if you want to use async/await:

use irondash_message_channel::*;

struct MyHandler {}

impl MethodHandler for MyHandler {
    fn on_method_call(&self, call: MethodCall, reply: MethodCallReply) {
        match call.method.as_str() {
            "getMeaningOfUniverse" => {
                reply.send_ok(42);
            }
            _ => reply.send_error(
                "invalid_method".into(),
                Some(format!("Unknown Method: {}", call.method)),
                Value::Null,
            ),
        }
    }
}

fn init() {
    let handler = MyHandler {}.register("my_method_channel");
    // make sure handler is not dropped, otherwise it can't handle method calls.
}

Or async version:

use irondash_message_channel::*;

struct MyHandler {}

#[async_trait(?Send)]
impl AsyncMethodHandler for MyHandler {
    async fn on_method_call(&self, call: MethodCall) -> PlatformResult {
        match call.method.as_str() {
            "getMeaningOfUniverse" => {
                Ok(42.into())
            }
            _ => Err(PlatformError {
                code: "invalid_method".into(),
                message: Some(format!("Unknown Method: {}", call.method)),
                detail: Value::Null,
            })),
        }
    }
}

fn init() {
    let handler = MyHandler {}.register("my_method_channel");
    // make sure handler is not dropped, otherwise it can't handle method calls.
}

Calling Dart from Rust

use irondash_message_channel::*;

struct MyHandler {
    invoker: Late<AsyncMethodInvoker>,
}

#[async_trait(?Send)]
impl AsyncMethodHandler for MyHandler {
    // This will be called right after method channel registration.
    // You can use invoker to call Dart methods handlers.
    fn assign_invoker(&self, invoker: AsyncMethodInvoker) {
        self.invoker.set(invoker);
    }

    // ...
}

Note that to use Invoker you need to know target isolateId. You can get it from MethodCall structure while handling method calls in Rust. You can also get notified when isolate is destroyed:

impl MethodHandler for MyHandler {
    /// Called when isolate is about to be destroyed.
    fn on_isolate_destroyed(&self, _isolate: IsolateId) {}
    // ...

To see message channel in action look at the example project.

Threading consideration

MethodHandler and AsyncMethodHandler are bound to thread on which they were created. The thread must be running a RunLoop. This is implicitely true for platform thread. To use channels on background threads, you need to create a RunLoop and run it yourself.

MethodInvoker is Send. It can be passed between threads and the response to method call will be received on same thread as the request was sent. Again, the thread must have a RunLoop running.

Converting to and from Value

Value is represents all types that can be sent between Rust and Dart. To simplify serialization and deserialization on Rust side, irondash_message_channel provides IntoValue and TryFromValue proc macros, that generate TryInto<YourStruct> and From<YourStruct> traits for Value. This is an optional feature:

[dependencies]
irondash_message_channel = { version = "0.1.0", features = ["derive"] }
#[derive(TryFromValue, IntoValue)]
struct AdditionRequest {
    a: f64,
    b: f64,
}

#[derive(IntoValue)]
struct AdditionResponse {
    result: f64,
    request: AdditionRequest,
}

let value: Value = get_value_from_somewhere();
let request: AdditionRequest = value.try_into()?;
let response: Value = AdditionResponse {
    result: request.a + request.b,
    request,
}.into();

More advanced mapping options are also supported, for example:

#[derive(IntoValue, TryFromValue)]
#[irondash(tag = "t", content = "c")]
#[irondash(rename_all = "UPPERCASE")]
enum Enum3CustomTagContent {
    Abc,
    #[irondash(rename = "_Def")]
    Def,
    SingleValue(i64),
    #[irondash(rename = "_DoubleValue")]
    DoubleValue(f64, f64),
    Xyz {
        x: i64,
        s: String,
        z1: Option<i64>,
        #[irondash(skip_if_empty)]
        z2: Option<i64>,
        z3: Option<f64>,
    },
}

Unlike serde, .into() and try_into() consume the original value, making it possible for zero-copy serialization and deserializaton.

Commit count: 0

cargo fmt