1 % The Rust Tasks and Communication Guide
5 Rust provides safe concurrency through a combination
6 of lightweight, memory-isolated tasks and message passing.
7 This guide will describe the concurrency model in Rust, how it
8 relates to the Rust type system, and introduce
9 the fundamental library abstractions for constructing concurrent programs.
11 Rust tasks are not the same as traditional threads: rather,
12 they are considered _green threads_, lightweight units of execution that the Rust
13 runtime schedules cooperatively onto a small number of operating system threads.
14 On a multi-core system Rust tasks will be scheduled in parallel by default.
15 Because tasks are significantly
16 cheaper to create than traditional threads, Rust can create hundreds of
17 thousands of concurrent tasks on a typical 32-bit system.
18 In general, all Rust code executes inside a task, including the `main` function.
20 In order to make efficient use of memory Rust tasks have dynamically sized stacks.
21 A task begins its life with a small
22 amount of stack space (currently in the low thousands of bytes, depending on
23 platform), and acquires more stack as needed.
24 Unlike in languages such as C, a Rust task cannot accidentally write to
25 memory beyond the end of the stack, causing crashes or worse.
27 Tasks provide failure isolation and recovery. When a fatal error occurs in Rust
28 code as a result of an explicit call to `fail!()`, an assertion failure, or
29 another invalid operation, the runtime system destroys the entire
30 task. Unlike in languages such as Java and C++, there is no way to `catch` an
31 exception. Instead, tasks may monitor each other for failure.
33 Tasks use Rust's type system to provide strong memory safety guarantees. In
34 particular, the type system guarantees that tasks cannot share mutable state
35 with each other. Tasks communicate with each other by transferring _owned_
36 data through the global _exchange heap_.
38 ## A note about the libraries
40 While Rust's type system provides the building blocks needed for safe
41 and efficient tasks, all of the task functionality itself is implemented
42 in the standard and sync libraries, which are still under development
43 and do not always present a consistent or complete interface.
45 For your reference, these are the standard modules involved in Rust
46 concurrency at this writing:
48 * [`std::task`] - All code relating to tasks and task scheduling,
49 * [`std::comm`] - The message passing interface,
50 * [`sync::DuplexStream`] - An extension of `pipes::stream` that allows both sending and receiving,
51 * [`sync::Arc`] - The Arc (atomically reference counted) type, for safely sharing immutable data,
52 * [`sync::Semaphore`] - A counting, blocking, bounded-waiting semaphore,
53 * [`sync::Mutex`] - A blocking, bounded-waiting, mutual exclusion lock with an associated
54 FIFO condition variable,
55 * [`sync::RWLock`] - A blocking, no-starvation, reader-writer lock with an associated condvar,
56 * [`sync::Barrier`] - A barrier enables multiple tasks to synchronize the beginning
58 * [`sync::TaskPool`] - A task pool abstraction,
59 * [`sync::Future`] - A type encapsulating the result of a computation which may not be complete,
60 * [`sync::one`] - A "once initialization" primitive
61 * [`sync::mutex`] - A proper mutex implementation regardless of the "flavor of task" which is
64 [`std::task`]: std/task/index.html
65 [`std::comm`]: std/comm/index.html
66 [`sync::DuplexStream`]: sync/struct.DuplexStream.html
67 [`sync::Arc`]: sync/struct.Arc.html
68 [`sync::Semaphore`]: sync/raw/struct.Semaphore.html
69 [`sync::Mutex`]: sync/struct.Mutex.html
70 [`sync::RWLock`]: sync/struct.RWLock.html
71 [`sync::Barrier`]: sync/struct.Barrier.html
72 [`sync::TaskPool`]: sync/struct.TaskPool.html
73 [`sync::Future`]: sync/struct.Future.html
74 [`sync::one`]: sync/one/index.html
75 [`sync::mutex`]: sync/mutex/index.html
79 The programming interface for creating and managing tasks lives
80 in the `task` module of the `std` library, and is thus available to all
81 Rust code by default. At its simplest, creating a task is a matter of
82 calling the `spawn` function with a closure argument. `spawn` executes the
83 closure in the new task.
86 # use std::task::spawn;
88 // Print something profound in a different task using a named function
89 fn print_message() { println!("I am running in a different task!"); }
92 // Print something profound in a different task using a `proc` expression
93 // The `proc` expression evaluates to an (unnamed) owned closure.
94 // That closure will call `println!(...)` when the spawned task runs.
96 spawn(proc() println!("I am also running in a different task!") );
99 In Rust, there is nothing special about creating tasks: a task is not a
100 concept that appears in the language semantics. Instead, Rust's type system
101 provides all the tools necessary to implement safe concurrency: particularly,
102 _owned types_. The language leaves the implementation details to the standard
105 The `spawn` function has a very simple type signature: `fn spawn(f:
106 proc())`. Because it accepts only owned closures, and owned closures
107 contain only owned data, `spawn` can safely move the entire closure
108 and all its associated state into an entirely different task for
109 execution. Like any closure, the function passed to `spawn` may capture
110 an environment that it carries across tasks.
113 # use std::task::spawn;
114 # fn generate_task_number() -> int { 0 }
115 // Generate some state locally
116 let child_task_number = generate_task_number();
119 // Capture it in the remote task
120 println!("I am child number {}", child_task_number);
126 Now that we have spawned a new task, it would be nice if we could
127 communicate with it. Recall that Rust does not have shared mutable
128 state, so one task may not manipulate variables owned by another task.
129 Instead we use *pipes*.
131 A pipe is simply a pair of endpoints: one for sending messages and another for
132 receiving messages. Pipes are low-level communication building-blocks and so
133 come in a variety of forms, each one appropriate for a different use case. In
134 what follows, we cover the most commonly used varieties.
136 The simplest way to create a pipe is to use the `channel`
137 function to create a `(Sender, Receiver)` pair. In Rust parlance, a *sender*
138 is a sending endpoint of a pipe, and a *receiver* is the receiving
139 endpoint. Consider the following example of calculating two results
143 # use std::task::spawn;
145 let (tx, rx): (Sender<int>, Receiver<int>) = channel();
148 let result = some_expensive_computation();
152 some_other_expensive_computation();
153 let result = rx.recv();
154 # fn some_expensive_computation() -> int { 42 }
155 # fn some_other_expensive_computation() {}
158 Let's examine this example in detail. First, the `let` statement creates a
159 stream for sending and receiving integers (the left-hand side of the `let`,
160 `(tx, rx)`, is an example of a *destructuring let*: the pattern separates
161 a tuple into its component parts).
164 let (tx, rx): (Sender<int>, Receiver<int>) = channel();
167 The child task will use the sender to send data to the parent task,
168 which will wait to receive the data on the receiver. The next statement
169 spawns the child task.
172 # use std::task::spawn;
173 # fn some_expensive_computation() -> int { 42 }
174 # let (tx, rx) = channel();
176 let result = some_expensive_computation();
181 Notice that the creation of the task closure transfers `tx` to the child
182 task implicitly: the closure captures `tx` in its environment. Both `Sender`
183 and `Receiver` are sendable types and may be captured into tasks or otherwise
184 transferred between them. In the example, the child task runs an expensive
185 computation, then sends the result over the captured channel.
187 Finally, the parent continues with some other expensive
188 computation, then waits for the child's result to arrive on the
192 # fn some_other_expensive_computation() {}
193 # let (tx, rx) = channel::<int>();
195 some_other_expensive_computation();
196 let result = rx.recv();
199 The `Sender` and `Receiver` pair created by `channel` enables efficient
200 communication between a single sender and a single receiver, but multiple
201 senders cannot use a single `Sender` value, and multiple receivers cannot use a
202 single `Receiver` value. What if our example needed to compute multiple
203 results across a number of tasks? The following program is ill-typed:
206 # fn some_expensive_computation() -> int { 42 }
207 let (tx, rx) = channel();
210 tx.send(some_expensive_computation());
213 // ERROR! The previous spawn statement already owns the sender,
214 // so the compiler will not allow it to be captured again
216 tx.send(some_expensive_computation());
220 Instead we can clone the `tx`, which allows for multiple senders.
223 let (tx, rx) = channel();
225 for init_val in range(0u, 3) {
226 // Create a new channel handle to distribute to the child task
227 let child_tx = tx.clone();
229 child_tx.send(some_expensive_computation(init_val));
233 let result = rx.recv() + rx.recv() + rx.recv();
234 # fn some_expensive_computation(_i: uint) -> int { 42 }
237 Cloning a `Sender` produces a new handle to the same channel, allowing multiple
238 tasks to send data to a single receiver. It upgrades the channel internally in
239 order to allow this functionality, which means that channels that are not
240 cloned can avoid the overhead required to handle multiple senders. But this
241 fact has no bearing on the channel's usage: the upgrade is transparent.
243 Note that the above cloning example is somewhat contrived since
244 you could also simply use three `Sender` pairs, but it serves to
245 illustrate the point. For reference, written with multiple streams, it
246 might look like the example below.
249 # use std::task::spawn;
251 // Create a vector of ports, one for each child task
252 let rxs = Vec::from_fn(3, |init_val| {
253 let (tx, rx) = channel();
255 tx.send(some_expensive_computation(init_val));
260 // Wait on each port, accumulating the results
261 let result = rxs.iter().fold(0, |accum, rx| accum + rx.recv() );
262 # fn some_expensive_computation(_i: uint) -> int { 42 }
265 ## Backgrounding computations: Futures
266 With `sync::Future`, rust has a mechanism for requesting a computation and getting the result
269 The basic example below illustrates this.
275 # fn make_a_sandwich() {};
276 fn fib(n: u64) -> u64 {
277 // lengthy computation returning an uint
281 let mut delayed_fib = sync::Future::spawn(proc() fib(50));
283 println!("fib(50) = {:?}", delayed_fib.get())
287 The call to `future::spawn` returns immediately a `future` object regardless of how long it
288 takes to run `fib(50)`. You can then make yourself a sandwich while the computation of `fib` is
289 running. The result of the execution of the method is obtained by calling `get` on the future.
290 This call will block until the value is available (*i.e.* the computation is complete). Note that
291 the future needs to be mutable so that it can save the result for next time `get` is called.
293 Here is another example showing how futures allow you to background computations. The workload will
294 be distributed on the available cores.
298 fn partial_sum(start: uint) -> f64 {
299 let mut local_sum = 0f64;
300 for num in range(start*100000, (start+1)*100000) {
301 local_sum += (num as f64 + 1.0).powf(-2.0);
307 let mut futures = Vec::from_fn(1000, |ind| sync::Future::spawn( proc() { partial_sum(ind) }));
309 let mut final_res = 0f64;
310 for ft in futures.mut_iter() {
311 final_res += ft.get();
313 println!("π^2/6 is not far from : {}", final_res);
317 ## Sharing immutable data without copy: Arc
319 To share immutable data between tasks, a first approach would be to only use pipes as we have seen
320 previously. A copy of the data to share would then be made for each task. In some cases, this would
321 add up to a significant amount of wasted memory and would require copying the same data more than
324 To tackle this issue, one can use an Atomically Reference Counted wrapper (`Arc`) as implemented in
325 the `sync` library of Rust. With an Arc, the data will no longer be copied for each task. The Arc
326 acts as a reference to the shared data and only this reference is shared and cloned.
328 Here is a small example showing how to use Arcs. We wish to run concurrently several computations on
329 a single large vector of floats. Each task needs the full vector to perform its duty.
337 fn pnorm(nums: &[f64], p: uint) -> f64 {
338 nums.iter().fold(0.0, |a, b| a + b.powf(p as f64)).powf(1.0 / (p as f64))
342 let numbers = Vec::from_fn(1000000, |_| rand::random::<f64>());
343 let numbers_arc = Arc::new(numbers);
345 for num in range(1u, 10) {
346 let (tx, rx) = channel();
347 tx.send(numbers_arc.clone());
350 let local_arc : Arc<Vec<f64>> = rx.recv();
351 let task_numbers = &*local_arc;
352 println!("{}-norm = {}", num, pnorm(task_numbers.as_slice(), num));
358 The function `pnorm` performs a simple computation on the vector (it computes the sum of its items
359 at the power given as argument and takes the inverse power of this value). The Arc on the vector is
367 # let numbers = Vec::from_fn(1000000, |_| rand::random::<f64>());
368 let numbers_arc=Arc::new(numbers);
372 and a clone of it is sent to each task
379 # let numbers=Vec::from_fn(1000000, |_| rand::random::<f64>());
380 # let numbers_arc = Arc::new(numbers);
381 # let (tx, rx) = channel();
382 tx.send(numbers_arc.clone());
386 copying only the wrapper and not its contents.
388 Each task recovers the underlying data by
395 # let numbers=Vec::from_fn(1000000, |_| rand::random::<f64>());
396 # let numbers_arc=Arc::new(numbers);
397 # let (tx, rx) = channel();
398 # tx.send(numbers_arc.clone());
399 # let local_arc : Arc<Vec<f64>> = rx.recv();
400 let task_numbers = &*local_arc;
404 and can use it as if it were local.
406 The `arc` module also implements Arcs around mutable data that are not covered here.
408 # Handling task failure
410 Rust has a built-in mechanism for raising exceptions. The `fail!()` macro
411 (which can also be written with an error string as an argument: `fail!(
412 ~reason)`) and the `assert!` construct (which effectively calls `fail!()`
413 if a boolean expression is false) are both ways to raise exceptions. When a
414 task raises an exception the task unwinds its stack---running destructors and
415 freeing memory along the way---and then exits. Unlike exceptions in C++,
416 exceptions in Rust are unrecoverable within a single task: once a task fails,
417 there is no way to "catch" the exception.
419 While it isn't possible for a task to recover from failure, tasks may notify
420 each other of failure. The simplest way of handling task failure is with the
421 `try` function, which is similar to `spawn`, but immediately blocks waiting
422 for the child task to finish. `try` returns a value of type `Result<T,
423 ()>`. `Result` is an `enum` type with two variants: `Ok` and `Err`. In this
424 case, because the type arguments to `Result` are `int` and `()`, callers can
425 pattern-match on a result to check whether it's an `Ok` result with an `int`
426 field (representing a successful result) or an `Err` result (representing
427 termination with an error).
429 ~~~{.ignore .linked-failure}
431 # fn some_condition() -> bool { false }
432 # fn calculate_result() -> int { 0 }
433 let result: Result<int, ()> = task::try(proc() {
434 if some_condition() {
440 assert!(result.is_err());
443 Unlike `spawn`, the function spawned using `try` may return a value,
444 which `try` will dutifully propagate back to the caller in a [`Result`]
445 enum. If the child task terminates successfully, `try` will
446 return an `Ok` result; if the child task fails, `try` will return
449 [`Result`]: std/result/index.html
451 > *Note:* A failed task does not currently produce a useful error
452 > value (`try` always returns `Err(())`). In the
453 > future, it may be possible for tasks to intercept the value passed to
456 TODO: Need discussion of `future_result` in order to make failure
459 But not all failures are created equal. In some cases you might need to
460 abort the entire program (perhaps you're writing an assert which, if
461 it trips, indicates an unrecoverable logic error); in other cases you
462 might want to contain the failure at a certain boundary (perhaps a
463 small piece of input from the outside world, which you happen to be
464 processing in parallel, is malformed and its processing task can't
467 ## Creating a task with a bi-directional communication path
469 A very common thing to do is to spawn a child task where the parent
470 and child both need to exchange messages with each other. The
471 function `sync::comm::duplex` supports this pattern. We'll
472 look briefly at how to use it.
474 To see how `duplex` works, we will create a child task
475 that repeatedly receives a `uint` message, converts it to a string, and sends
476 the string in response. The child terminates when it receives `0`.
477 Here is the function that implements the child task:
482 fn stringifier(channel: &sync::DuplexStream<~str, uint>) {
485 value = channel.recv();
486 channel.send(value.to_str());
487 if value == 0 { break; }
493 The implementation of `DuplexStream` supports both sending and
494 receiving. The `stringifier` function takes a `DuplexStream` that can
495 send strings (the first type parameter) and receive `uint` messages
496 (the second type parameter). The body itself simply loops, reading
497 from the channel and then sending its response back. The actual
498 response itself is simply the stringified version of the received value,
499 `uint::to_str(value)`.
501 Here is the code for the parent task:
505 # use std::task::spawn;
506 # use sync::DuplexStream;
507 # fn stringifier(channel: &sync::DuplexStream<~str, uint>) {
508 # let mut value: uint;
510 # value = channel.recv();
511 # channel.send(value.to_str());
512 # if value == 0u { break; }
517 let (from_child, to_child) = sync::duplex();
520 stringifier(&to_child);
524 assert!(from_child.recv() == "22".to_owned());
529 assert!(from_child.recv() == "23".to_owned());
530 assert!(from_child.recv() == "0".to_owned());
535 The parent task first calls `DuplexStream` to create a pair of bidirectional
536 endpoints. It then uses `task::spawn` to create the child task, which captures
537 one end of the communication channel. As a result, both parent and child can
538 send and receive data to and from the other.