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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
//! [](https://crates.io/crates/safina-threadpool) //! [](http://www.apache.org/licenses/LICENSE-2.0) //! [](https://github.com/rust-secure-code/safety-dance/) //! [](https://gitlab.com/leonhard-llc/safina-rs/-/pipelines) //! //! This is a safe Rust thread pool library. //! //! You can use it alone or with [`safina`](https://crates.io/crates/safina), //! a safe async runtime. //! //! # Features //! - `forbid(unsafe_code)` //! - Depends only on `std` //! - Good test coverage (100%) //! //! # Limitations //! - Allocates memory //! - Not optimized //! //! # Documentation //! <https://docs.rs/safina-threadpool> //! //! # Examples //! ```rust //! # type ProcessResult = (); //! # fn process_data(data: (), sender: std::sync::mpsc::Sender<ProcessResult>) -> ProcessResult { //! # sender.send(()).unwrap(); //! # } //! # fn f() { //! # let data_source = vec![(),()]; //! let pool = //! safina_threadpool::ThreadPool::new("worker", 2); //! let receiver = { //! let (sender, receiver) = //! std::sync::mpsc::channel(); //! for data in data_source { //! let sender_clone = sender.clone(); //! pool.schedule( //! move || process_data(data, sender_clone)); //! } //! receiver //! }; //! let results: Vec<ProcessResult> = //! receiver.iter().collect(); //! // ... //! # } //! ``` //! //! # Alternatives //! - [`blocking`](https://crates.io/crates/blocking) //! - Popular //! - A little `unsafe` code //! - [`threadpool`](https://crates.io/crates/threadpool) //! - Popular //! - Well maintained //! - Dependencies have `unsafe` code //! - [`futures-executor`](https://crates.io/crates/futures-executor) //! - Very popular //! - Full of `unsafe` //! - [`scoped_threadpool`](https://crates.io/crates/scoped_threadpool) //! - Popular //! - Contains `unsafe` code //! - [`scheduled-thread-pool`](https://crates.io/crates/scheduled-thread-pool) //! - Used by a popular connection pool library //! - Dependencies have `unsafe` code //! - [`workerpool`](https://crates.io/crates/workerpool) //! - Dependencies have `unsafe` code //! - [`threads_pool`](https://crates.io/crates/threads_pool) //! - Full of `unsafe` //! - [`thread-pool`](https://crates.io/crates/thread-pool) //! - Old //! - Dependencies have `unsafe` code //! - [`tasque`](https://crates.io/crates/tasque) //! - Dependencies have `unsafe` code //! - [`fast-threadpool`](https://crates.io/crates/fast-threadpool) //! - Dependencies have `unsafe` code //! - [`blocking-permit`](https://crates.io/crates/blocking-permit) //! - Full of `unsafe` //! - [`rayon-core`](https://crates.io/crates/rayon-core) //! - Full of `unsafe` //! //! # Changelog //! - v0.1.2 - Add another example //! - v0.1.1 - Simplified internals and improved documentation. //! - v0.1.0 - First release //! //! # TO DO //! - DONE - Add `schedule` and `try_schedule` //! - DONE - Add tests //! - DONE - Add docs //! - DONE - Publish on crates.io //! - Add a stress test //! - Add a benchmark. See benchmarks in <https://crates.io/crates/executors> //! - Add a way for a job to schedule another job on the same thread, with stealing. //! //! # Release Process //! 1. Edit `Cargo.toml` and bump version number. //! 1. Run `./release.sh` #![forbid(unsafe_code)] use core::fmt::{Display, Formatter}; use core::sync::atomic::{AtomicUsize, Ordering}; use core::time::Duration; use std::error::Error; use std::sync::mpsc::{Receiver, RecvTimeoutError, SyncSender, TrySendError}; use std::sync::{Arc, Mutex}; struct AtomicCounter { next_value: AtomicUsize, } impl AtomicCounter { pub fn new() -> Self { Self { next_value: AtomicUsize::new(0), } } pub fn next(&self) -> usize { self.next_value.fetch_add(1, Ordering::AcqRel) } } /// Returned by [`try_schedule`](struct.ThreadPool.html#method.try_schedule) /// when the queue is full. /// This can happen when the program schedules many closures at one time. /// It can also happen when closures panic their threads. The pool's /// throughput goes down when it must create new threads. #[derive(Debug)] pub struct QueueFull {} impl Display for QueueFull { fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), std::fmt::Error> { std::fmt::Debug::fmt(self, f) } } impl Error for QueueFull {} struct Inner { name: &'static str, next_name_num: AtomicCounter, size: usize, receiver: Mutex<Receiver<Box<dyn FnOnce() + Send>>>, } impl Inner { pub fn num_live_threads(self: &Arc<Inner>) -> usize { Arc::strong_count(self) - 1 } fn work(self: &Arc<Inner>) { loop { let recv_result = self .receiver .lock() .unwrap() .recv_timeout(Duration::from_millis(500)); self.start_threads(); match recv_result { Ok(f) => f(), Err(RecvTimeoutError::Timeout) => {} // ThreadPool was dropped. Err(RecvTimeoutError::Disconnected) => return, }; self.start_threads(); } } fn start_thread(self: &Arc<Inner>) { let self_clone = self.clone(); if self.num_live_threads() <= self.size { std::thread::Builder::new() .name(format!("{}{}", self.name, self.next_name_num.next())) .spawn(move || self_clone.work()) .unwrap(); } } fn start_threads(self: &Arc<Inner>) { while self.num_live_threads() < self.size { self.start_thread(); } } } /// A collection of threads and a queue for jobs (`FnOnce` structs) they execute. /// /// Threads die when they execute a job that panics. /// If one thread survives, it will recreate all the threads. /// The next call to [`schedule`](#method.schedule) or [`try_schedule`](#method.try_schedule) /// also recreates threads. /// /// If your threadpool load is bursty and you want to automatically recover /// from an all-threads-panicked state, you could use /// [`safina_timer`](https://crates.io/crates/safina-timer) to periodically call /// [`schedule`](#method.schedule) or [`try_schedule`](#method.try_schedule). /// /// # Example /// ```rust /// # type ProcessResult = (); /// # fn process_data(data: (), sender: std::sync::mpsc::Sender<ProcessResult>) -> ProcessResult { /// # sender.send(()).unwrap(); /// # } /// # fn f() { /// # let data_source = vec![(),()]; /// let pool = /// safina_threadpool::ThreadPool::new("worker", 2); /// let receiver = { /// let (sender, receiver) = /// std::sync::mpsc::channel(); /// for data in data_source { /// let sender_clone = sender.clone(); /// pool.schedule( /// move || process_data(data, sender_clone)); /// } /// receiver /// }; /// let results: Vec<ProcessResult> = /// receiver.iter().collect(); /// // ... /// # } /// ``` /// /// ```rust /// # use core::time::Duration; /// # use std::sync::Arc; /// let pool = /// Arc::new(safina_threadpool::ThreadPool::new("worker", 2)); /// let executor = safina_executor::Executor::default(); /// safina_timer::start_timer_thread(); /// let pool_clone = pool.clone(); /// executor.spawn(async move { /// loop { /// safina_timer::sleep_for(Duration::from_millis(500)).await; /// pool_clone.schedule(|| {}); /// } /// }); /// # assert_eq!(2, pool.num_live_threads()); /// # for _ in 0..2 { /// # pool.schedule(|| { /// # std::thread::sleep(Duration::from_millis(100)); /// # panic!("ignore this panic") /// # }); /// # } /// # std::thread::sleep(Duration::from_millis(200)); /// # assert_eq!(0, pool.num_live_threads()); /// # std::thread::sleep(Duration::from_millis(500)); /// # assert_eq!(2, pool.num_live_threads()); /// ``` pub struct ThreadPool { inner: Arc<Inner>, sender: SyncSender<Box<dyn FnOnce() + Send>>, } impl ThreadPool { /// Creates a new thread pool containing `size` threads. /// The threads all start immediately. /// /// Threads are named with `name` with a number. /// For example, `ThreadPool::new("worker", 2)` /// creates threads named "worker-1" and "worker-2". /// If one of those threads panics, the pool creates "worker-3". /// /// Panics if `name` is empty or `size` is zero. /// /// After the `ThreadPool` struct drops, the threads continue processing /// jobs and stop when the queue is empty. #[must_use] pub fn new(name: &'static str, size: usize) -> Self { if name.is_empty() { panic!("ThreadPool::new called with empty name") } if size < 1 { panic!("ThreadPool::new called with invalid size value: {:?}", size) } // Use a channel with bounded size. // If the channel was unbounded, the process could OOM when throughput goes down. let (sender, receiver) = std::sync::mpsc::sync_channel(size * 200); let pool = ThreadPool { inner: Arc::new(Inner { name, next_name_num: AtomicCounter::new(), size, receiver: Mutex::new(receiver), }), sender, }; pool.inner.start_threads(); pool } /// Returns the number of threads in the pool. #[must_use] pub fn size(&self) -> usize { self.inner.size } /// Returns the number of threads currently alive. #[must_use] pub fn num_live_threads(&self) -> usize { self.inner.num_live_threads() } /// Adds a job to the queue. The next idle thread will execute it. /// Jobs are started in FIFO order. /// /// Blocks when the queue is full. /// See [`try_schedule`](#method.try_schedule). /// /// Recreates any threads that panicked. /// /// Puts `f` in a [`Box`](https://doc.rust-lang.org/stable/std/boxed/struct.Box.html) before /// adding it to the queue. pub fn schedule(&self, f: impl FnOnce() + Send + 'static) { // If all workers panicked and the queue is full, adding to the queue will // block forever. So we start threads first. self.inner.start_threads(); self.sender.send(Box::new(f)).unwrap(); } /// Adds a job to the queue. The next idle thread will execute it. /// Jobs are started in FIFO order. /// /// Recreates any threads that panicked. /// /// Puts `f` in a [`Box`](https://doc.rust-lang.org/stable/std/boxed/struct.Box.html) before /// adding it to the queue. /// /// # Errors /// Returns `Err(QueueFull)` when the queue is full. pub fn try_schedule(&self, f: impl FnOnce() + Send + 'static) -> Result<(), QueueFull> { let result = match self.sender.try_send(Box::new(f)) { Ok(_) => Ok(()), Err(TrySendError::Disconnected(_)) => unreachable!(), Err(TrySendError::Full(_)) => Err(QueueFull {}), }; self.inner.start_threads(); result } } #[cfg(test)] mod tests { use super::*; use core::fmt::Debug; use core::ops::Range; use core::time::Duration; use std::time::Instant; fn assert_in_range<T: PartialOrd + Debug>(range: Range<T>, value: &T) { if range.is_empty() { panic!("invalid range {:?}", range) } // println!( // "measured concurrency value {:?}, expected range {:?}", // value, range, // ); if !range.contains(value) { panic!( "measured concurrency value {:?} out of range {:?}", value, range, ); } } pub fn assert_elapsed(before: Instant, range_ms: Range<u64>) { if range_ms.is_empty() { panic!("invalid range {:?}", range_ms) } let elapsed = before.elapsed(); let duration_range = Duration::from_millis(range_ms.start)..Duration::from_millis(range_ms.end); if !duration_range.contains(&elapsed) { panic!("{:?} elapsed, out of range {:?}", elapsed, duration_range); } } fn measure_concurrency(pool: &ThreadPool, num_jobs: usize) -> f32 { const WAIT_DURATION: Duration = Duration::from_millis(100); let before = Instant::now(); let receiver = { let (sender, receiver) = std::sync::mpsc::channel(); for _ in 0..num_jobs { let sender_clone = sender.clone(); pool.schedule(move || { std::thread::sleep(WAIT_DURATION); sender_clone.send(()).unwrap(); }); } receiver }; for _ in 0..num_jobs { receiver.recv_timeout(Duration::from_millis(500)).unwrap(); } let elapsed = before.elapsed(); elapsed.as_secs_f32() / WAIT_DURATION.as_secs_f32() } fn sleep(ms: u64) { std::thread::sleep(Duration::from_millis(ms)); } #[test] fn atomic_counter() { let counter = Arc::new(AtomicCounter::new()); assert_eq!(0, counter.next()); assert_eq!(1, counter.next()); assert_eq!(2, counter.next()); } #[test] fn atomic_counter_many_readers() { let receiver = { let counter = Arc::new(AtomicCounter::new()); let (sender, receiver) = std::sync::mpsc::channel(); for _ in 0..10 { let counter_clone = counter.clone(); let sender_clone = sender.clone(); std::thread::spawn(move || { for _ in 0..10 { sender_clone.send(counter_clone.next()).unwrap(); } }); } receiver }; let mut values: Vec<usize> = receiver.iter().collect(); values.sort_unstable(); assert_eq!((0_usize..100).collect::<Vec<usize>>(), values); } #[test] fn queue_full_display() { assert_eq!("QueueFull", format!("{}", QueueFull {})); } #[test] fn empty_name() { if std::panic::catch_unwind(|| ThreadPool::new("", 1)).is_ok() { panic!("expected panic") } } #[test] fn zero_size() { if std::panic::catch_unwind(|| ThreadPool::new("pool1", 0)).is_ok() { panic!("expected panic") } } #[test] fn test_size() { let pool = ThreadPool::new("pool1", 3); assert_eq!(3, pool.size()); } #[test] fn should_name_threads_consecutively() { let pool = ThreadPool::new("poolA", 2); pool.schedule(move || { sleep(200); }); pool.schedule(move || panic!("ignore this panic")); sleep(100); let (sender, receiver) = std::sync::mpsc::channel(); pool.schedule(move || { sender .send(std::thread::current().name().unwrap().to_string()) .unwrap(); }); assert_eq!( "poolA2", receiver.recv_timeout(Duration::from_millis(500)).unwrap() ); } #[test] fn test_num_live_threads() { let pool = ThreadPool::new("pool1", 3); sleep(100); assert_eq!(3, pool.num_live_threads()); pool.schedule(move || { sleep(100); panic!("ignore this panic") }); pool.schedule(move || { sleep(100); panic!("ignore this panic") }); pool.schedule(move || { sleep(100); panic!("ignore this panic") }); sleep(200); assert_eq!(0, pool.num_live_threads()); pool.schedule(move || {}); assert_eq!(3, pool.num_live_threads()); } #[test] fn schedule_should_run_the_fn() { let pool = ThreadPool::new("pool1", 1); let before = Instant::now(); let (sender, receiver) = std::sync::mpsc::channel(); pool.schedule(move || { sender.send(()).unwrap(); }); receiver.recv_timeout(Duration::from_millis(500)).unwrap(); assert_elapsed(before, 0..100); } #[test] fn schedule_should_start_a_thread_if_none() { let pool = ThreadPool::new("pool1", 3); sleep(100); pool.schedule(move || { sleep(100); panic!("ignore this panic") }); pool.schedule(move || { sleep(100); panic!("ignore this panic") }); pool.schedule(move || { sleep(100); panic!("ignore this panic") }); sleep(200); assert_eq!(0, pool.num_live_threads()); pool.schedule(|| {}); assert_eq!(3, pool.num_live_threads()); } #[test] fn try_schedule_should_run_the_fn() { let pool = ThreadPool::new("pool1", 1); let before = Instant::now(); let (sender, receiver) = std::sync::mpsc::channel(); pool.try_schedule(move || { sender.send(()).unwrap(); }) .unwrap(); receiver.recv_timeout(Duration::from_millis(500)).unwrap(); assert_elapsed(before, 0..100); } #[test] fn try_schedule_queue_full() { let pool = ThreadPool::new("pool1", 1); let before = Instant::now(); while Instant::now() - before < Duration::from_millis(500) { if pool .try_schedule(move || panic!("ignore this panic")) .is_err() { //println!("try_schedule got {:?}", e); // Sometimes a thread's panic message is interspersed with the // test runner result, confusing the IDE. We sleep here to // let the threads run and finish before we return. sleep(100); return; } } panic!("timeout"); } #[test] fn check_concurrency1() { let pool = ThreadPool::new("pool1", 1); assert_in_range(1.0..1.99, &measure_concurrency(&pool, 1)); assert_in_range(2.0..2.99, &measure_concurrency(&pool, 2)); } #[test] fn check_concurrency2() { let pool = ThreadPool::new("pool1", 2); assert_in_range(1.0..1.99, &measure_concurrency(&pool, 1)); assert_in_range(1.0..1.99, &measure_concurrency(&pool, 2)); assert_in_range(2.0..2.99, &measure_concurrency(&pool, 3)); assert_in_range(2.0..2.99, &measure_concurrency(&pool, 4)); } #[test] fn check_concurrency5() { let pool = ThreadPool::new("pool1", 5); assert_in_range(1.0..1.99, &measure_concurrency(&pool, 5)); assert_in_range(2.0..2.99, &measure_concurrency(&pool, 6)); } #[test] fn should_respawn_when_idle() { let pool = ThreadPool::new("pool1", 2); sleep(100); pool.schedule(move || panic!("ignore this panic")); sleep(100); assert_eq!(1, pool.num_live_threads()); sleep(500); assert_eq!(2, pool.num_live_threads()); assert_in_range(1.0..1.99, &measure_concurrency(&pool, 2)); } #[test] fn should_respawn_after_recv() { let pool = ThreadPool::new("pool1", 2); sleep(100); pool.schedule(move || panic!("ignore this panic")); sleep(100); assert_eq!(1, pool.num_live_threads()); pool.schedule(move || sleep(200)); sleep(100); assert_eq!(2, pool.num_live_threads()); } #[test] fn should_respawn_after_executing_job() { let pool = ThreadPool::new("pool1", 2); pool.schedule(move || sleep(200)); pool.schedule(move || panic!("ignore this panic")); sleep(100); assert_eq!(1, pool.num_live_threads()); sleep(200); assert_eq!(2, pool.num_live_threads()); } // TODO(mleonhard) Test that threads stop after `ThreadPool` drops. }