1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
//! A lightweight actor model inspired framework to build //! customizable components with message-based intercommunications. #![warn(missing_docs, missing_debug_implementations, rust_2018_idioms)] mod error; use std::{ any::type_name, fmt::{self, Debug}, panic::catch_unwind, sync::{Arc, Weak}, thread, time::Duration, }; use { async_io::block_on, flume::{bounded, unbounded, Receiver, RecvTimeoutError, Sender, TrySendError}, futures_lite::future::{or, pending}, once_cell::sync::Lazy, }; pub use crate::error::Error; /// An async executor. pub type Executor<'a> = async_executor::Executor<'a>; /// A default executor. It runs on per-core threads and is fair /// in terms of task priorities. pub static DEFAULT_EXECUTOR: Lazy<Executor<'static>> = Lazy::new(|| { let num_threads = num_cpus::get(); for n in 1..=num_threads { thread::Builder::new() .name(format!("appliance-{}", n)) .spawn(|| loop { catch_unwind(|| block_on(DEFAULT_EXECUTOR.run(pending::<()>()))).ok(); }) .expect("cannot spawn an appliance executor thread"); } Executor::new() }); /// `Message` must be implemented for any type which is intended for /// sending to appliances. /// /// # Example /// ``` /// # use std::time::Duration; /// # use appliance::{Appliance, ApplianceHandle, Handler, Message}; /// type Counter = Appliance<'static, usize>; /// /// struct Ping; /// /// impl Message for Ping { type Result = usize; } /// /// impl Handler<Ping> for Counter { /// fn handle(&mut self, _msg: &Ping) -> usize { /// *self.state() += 1; /// *self.state() /// } /// } /// /// fn do_ping(handle: ApplianceHandle<Counter>) { /// match handle.send_and_wait_with_timeout(Ping, Duration::from_secs(10)) { /// Ok(cnt) => println!("Appliance was pinged successfully {} times", *cnt), /// Err(err) => panic!("Ping to appliance has failed: {}", err), /// } /// } /// ``` pub trait Message: Send { /// The type of replies generated by handling this message. type Result: Send; } /// A trait which must be implemented for all appliances which are intended to receive /// messages of type `M`. One appliance can handle multiple message types. /// /// Handler's logic is strongly encouraged to include only fast (non-blocking) and synchronous /// mutations of the appliance state. Otherwise, the appliance's event loop may get slow, and /// hence flood the internal buffer causing message sending denials. /// /// # Example /// ``` /// # use std::time::Duration; /// # use appliance::{Appliance, ApplianceHandle, DEFAULT_EXECUTOR, Handler, Message}; /// type Counter = Appliance<'static, usize>; /// /// struct Ping; /// /// impl Message for Ping { type Result = usize; } /// /// impl Handler<Ping> for Counter { /// fn handle(&mut self, _msg: &Ping) -> usize { /// *self.state() += 1; /// *self.state() /// } /// } /// /// struct Reset; /// /// impl Message for Reset { type Result = (); } /// /// impl Handler<Reset> for Counter { /// fn handle(&mut self, _msg: &Reset) { /// *self.state() = 0; /// } /// } /// /// const BUF_SIZE: usize = 10; /// /// fn main() -> Result<(), appliance::Error> { /// let handle = Appliance::new_bounded(&DEFAULT_EXECUTOR, 0, BUF_SIZE); /// assert_eq!(*handle.send_and_wait_sync(Ping)?, 1); /// assert_eq!(*handle.send_and_wait_sync(Ping)?, 2); /// handle.send(Reset)?; /// assert_eq!(*handle.send_and_wait_sync(Ping)?, 1); /// Ok(()) /// } /// ``` pub trait Handler<M: Message> { /// Handle the incoming message fn handle(&mut self, msg: M) -> M::Result; } /// A dual trait for `Handler`. For any type of messages `M` and any type of handlers `H`, /// if `impl Handle<M> for H`, then `impl HandledBy<H> for M`. I.e. we can either ask /// "which messages can be handled by this appliance" or "which actors can handle this message", /// and the answers to these questions are dual. The trait `HandledBy` answers the second /// question. /// /// Normally one should always implement `Handle<M>`, unless for some reason it is impossible /// to do. The dual `HandledBy` impl is then provided automatically. /// /// Generic methods, on the other hand, should use the trait constraint `T: HandledBy<H>`, since /// the set of types for which `T: HandledBy<H>` is strictly larger than those for which /// `H: Handler<T>`. An example where the client would need to implement `HandledBy` is if they /// want to add custom messages for a library-provided handler type. pub trait HandledBy<H: ?Sized>: Message { /// Handle the given message with the provided handler. /// /// The return type is wrapped in order to remove generic parameters from this function. /// The actual result value can be recovered with `ResultWrapper::downcast` if the type /// of the result is known. fn handle_by(self, handler: &mut H) -> Self::Result; } impl<H, M: Message> HandledBy<H> for M where H: Handler<M>, { fn handle_by(self, handler: &mut H) -> Self::Result { handler.handle(self) } } struct InnerMessage<'a, H: ?Sized + 'a> { handle_message: Box<dyn FnOnce(&mut H) + Send + 'a>, } impl<'a, H: ?Sized + 'a> InnerMessage<'a, H> { fn new<M>(message: M, reply_channel: Option<Sender<M::Result>>) -> Self where M: HandledBy<H> + 'a, { InnerMessage { handle_message: Box::new(move |handler| { let result = message.handle_by(handler); if let Some(rc) = &reply_channel { rc.send(result).ok(); } }), } } } type MessageSender<'a, H> = Sender<InnerMessage<'a, H>>; /// A stateful entity that only allows to interact with via handling messages. /// /// The appliance itself is not directly available. Messages must be sent to it /// using its handle which is returned by the `Appliance::new_bounded` and /// `Appliance::new_unbounded` methods, and can also be obtained from `&Appliance` /// using `Appliance::handle` method. Note that the latter route is generally /// available only for message handler `Handler` implementations. pub struct Appliance<'s, S> { state: S, handle: Weak<MessageSender<'s, Self>>, } impl<'s, S: Debug + 's> Debug for Appliance<'s, S> { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "Appliance({:?})", self.state) } } impl<'s, S> Appliance<'s, S> where S: Send + 's, { /// Creates a new appliance with a bounded message buffer, /// a state, and a handler. pub fn new_bounded( executor: &'s Executor<'s>, state: S, size: usize, ) -> ApplianceHandle<'s, Self> { Self::run(executor, state, size) } /// Creates a new appliance with an unbounded message buffer, /// a state, and a handler. It's not recommended to use this /// version of appliance in production just like any other /// memory unbounded contruct. pub fn new_unbounded(executor: &'s Executor<'s>, state: S) -> ApplianceHandle<'s, Self> { Self::run(executor, state, None) } /// Creates a new appliance with the given state, message handler, /// and buffer size, if any. fn run( executor: &'s Executor<'s>, state: S, size: impl Into<Option<usize>>, ) -> ApplianceHandle<'s, Self> { let (in_, out_) = if let Some(mbs) = size.into() { bounded(mbs) } else { unbounded() }; let handle = ApplianceHandle { inner: Arc::new(in_), }; let mut appliance = Appliance { state, handle: Arc::downgrade(&handle.inner), }; executor .spawn(async move { appliance.handle_messages(out_).await }) .detach(); handle } /// Returns a handle object of the appliance. /// /// A handle is a cloneable object which allows to send messages to the appliance. /// /// This function will return `None` if all appliance handles were already dropped. /// In this case the appliance must shutdown. pub fn handle(&'s self) -> Option<ApplianceHandle<'s, Self>> { self.handle.upgrade().map(|inner| ApplianceHandle { inner }) } /// The mutable inner state of the appliance. /// /// Note that this function requires mutable access to the appliance itself (not its /// handle), but the appliance object is never returned by the API. The only place where /// the appliance can be accessed is the implementation of `Handle` and `HandledBy` traits /// for the message types, which is thus also the only place where one can (and should) mutate its state. pub fn state(&mut self) -> &mut S { &mut self.state } async fn handle_messages(&mut self, out_: Receiver<InnerMessage<'_, Self>>) { while let Ok(InnerMessage { handle_message }) = out_.recv_async().await { handle_message(self); } } } /// Appliance handle is a cloneable object which allows to send messages to the appliance. /// /// Once all handles to the appliance are dropped, the appliance will terminate its event loop /// and be destroyed. pub struct ApplianceHandle<'a, A: ?Sized> { /// The incoming channel which is used to send messages to the appliance. /// /// We are forced to stupidly wrap `Sender` in an `Arc` even though it already is a /// wrapped `Arc`. We need the extra `Arc` so that we can pass a weak reference to it into /// the `Appliance` object, but unfortunately `flume::Sender` doesn't provide weak references /// in the API. /// /// Make sure that the inner `Sender` is never leaked outside of the containing `Arc`. If that /// happens, the appliance will stay alive after all handles are dropped, which violates the /// API contract. inner: Arc<MessageSender<'a, A>>, } impl<'a, A: ?Sized + 'a> Debug for ApplianceHandle<'a, A> { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "ApplianceHandle<{}>(..)", type_name::<A>()) } } impl<'a, A: ?Sized + 'a> Clone for ApplianceHandle<'a, A> { fn clone(&self) -> Self { ApplianceHandle { inner: self.inner.clone(), } } } impl<'a, A: ?Sized + 'a> ApplianceHandle<'a, A> { /// Sends a message to the current appliance without /// waiting for the message to be handled. pub fn send_sync<M>(&self, message: M) -> Result<(), Error> where M: HandledBy<A> + 'a, { match self.inner.try_send(InnerMessage::new(message, None)) { Ok(_) => Ok(()), Err(TrySendError::Full(_)) => Err(Error::FullBuffer), Err(TrySendError::Disconnected(_)) => Err(Error::UnexpectedFailure), } } /// Does conceptually the same thing as `send_sync` but gets intended /// to be used in async context. pub async fn send_async<M>(&self, message: M) -> Result<(), Error> where M: HandledBy<A> + 'a, { self.inner .send_async(InnerMessage::new(message, None)) .await .map_err(|_| Error::UnexpectedFailure) } /// Sends a message to the current appliance and waits /// forever, if `timeout` is None, or for only given time /// for the message to be handled. /// This synchronous blocking method is a fit for callers /// who don't use async execution and must be assured that /// the message has been handled. /// Note, it is supposed to be used less often than `send` /// as it may suffer a significant performance hit due to /// synchronization with the handling loop. pub fn send_and_wait_sync<M>( &self, message: M, timeout: Option<Duration>, ) -> Result<M::Result, Error> where M: HandledBy<A> + 'a, { let (s, r) = bounded(1); match self.inner.try_send(InnerMessage::new(message, Some(s))) { Err(TrySendError::Full(_)) => return Err(Error::FullBuffer), Err(TrySendError::Disconnected(_)) => return Err(Error::UnexpectedFailure), _ => {} } if let Some(timeout) = timeout { match r.recv_timeout(timeout) { Ok(r) => Ok(r), Err(RecvTimeoutError::Timeout) => Err(Error::Timeout), Err(RecvTimeoutError::Disconnected) => Err(Error::UnexpectedFailure), } } else { r.recv().map_err(|_| Error::UnexpectedFailure) } } /// Does conceptually the same thing as `send_and_wait_sync` /// but gets intended to be used in async context. This method /// is well suited for waiting for a result. pub async fn send_and_wait_async<M>( &self, message: M, timeout: Option<Duration>, ) -> Result<M::Result, Error> where M: HandledBy<A> + 'a, { let (send, recv) = bounded(1); if let Err(_) = self .inner .send_async(InnerMessage::new(message, Some(send))) .await { return Err(Error::UnexpectedFailure); } if let Some(timeout) = timeout { let f1 = async { recv.recv_async() .await .map_err(|_| Error::UnexpectedFailure) }; let f2 = async { async_io::Timer::after(timeout).await; Err(Error::Timeout) }; or(f1, f2).await } else { recv.recv_async() .await .map_err(|_| Error::UnexpectedFailure) } } }