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
//! [![crates.io version](https://img.shields.io/crates/v/safina-threadpool.svg)](https://crates.io/crates/safina-threadpool)
//! [![license: Apache 2.0](https://gitlab.com/leonhard-llc/safina-rs/-/raw/main/license-apache-2.0.svg)](http://www.apache.org/licenses/LICENSE-2.0)
//! [![unsafe forbidden](https://gitlab.com/leonhard-llc/safina-rs/-/raw/main/unsafe-forbidden-success.svg)](https://github.com/rust-secure-code/safety-dance/)
//! [![pipeline status](https://gitlab.com/leonhard-llc/safina-rs/badges/main/pipeline.svg)](https://gitlab.com/leonhard-llc/safina-rs/-/pipelines)
//!
//! A threadpool.
//!
//! You can use it alone or with [`safina`](https://crates.io/crates/safina),
//! a safe async runtime.
//!
//! # Features
//! - Add closures or `FnOnce` to the pool
//! - One of the pool's threads will execute it
//! - Automatically restarts panicked threads
//! - Handles hitting process thread limit.
//! - `forbid(unsafe_code)`
//! - Depends only on `std`
//! - 100% test coverage
//!
//! # 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).unwrap();
//! 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
//! - v0.2.0
//! - `ThreadPool::new` to return `Result`.
//! - `ThreadPool::try_schedule` to return an error when it fails to restart panicked threads.
//! - `ThreadPool::schedule` to handle failure starting replacement threads.
//! - v0.1.4 - Stop threads on drop.
//! - v0.1.3 - Support stable Rust! Needs 1.51+.
//! - v0.1.2 - Add another example
//! - v0.1.1 - Simplified internals and improved documentation.
//! - v0.1.0 - First release
//!
//! # TO DO
//! - Log a warning when all threads panicked.
//! - Update test coverage.
//! - Add a public `respawn_threads` function.
//! - 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::{Debug, 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)
}
}
#[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);
}
fn err_eq(a: &std::io::Error, b: &std::io::Error) -> bool {
a.kind() == b.kind() && format!("{}", a) == format!("{}", b)
}
#[derive(Debug)]
pub enum StartThreadsError {
/// The pool has no threads and `std::thread::Builder::spawn` returned the included error.
NoThreads(std::io::Error),
/// The pool has at least one thread and `std::thread::Builder::spawn` returned the included error.
Respawn(std::io::Error),
}
impl Display for StartThreadsError {
fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), std::fmt::Error> {
match self {
StartThreadsError::NoThreads(e) => write!(
f,
"ThreadPool workers all panicked, failed starting replacement threads: {}",
e
),
StartThreadsError::Respawn(e) => {
write!(
f,
"ThreadPool failed starting threads to replace panicked threads: {}",
e
)
}
}
}
}
impl Error for StartThreadsError {}
impl PartialEq for StartThreadsError {
fn eq(&self, other: &Self) -> bool {
match (self, other) {
(StartThreadsError::NoThreads(a), StartThreadsError::NoThreads(b))
| (StartThreadsError::Respawn(a), StartThreadsError::Respawn(b)) => err_eq(a, b),
_ => false,
}
}
}
impl Eq for StartThreadsError {}
#[derive(Debug)]
pub enum NewThreadPoolError {
Parameter(String),
/// `std::thread::Builder::spawn` returned the included error.
Spawn(std::io::Error),
}
impl Display for NewThreadPoolError {
fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), std::fmt::Error> {
match self {
NewThreadPoolError::Parameter(s) => write!(f, "{}", s),
NewThreadPoolError::Spawn(e) => {
write!(f, "ThreadPool failed starting threads: {}", e)
}
}
}
}
impl Error for NewThreadPoolError {}
impl PartialEq for NewThreadPoolError {
fn eq(&self, other: &Self) -> bool {
match (self, other) {
(NewThreadPoolError::Parameter(a), NewThreadPoolError::Parameter(b)) => a == b,
(NewThreadPoolError::Spawn(a), NewThreadPoolError::Spawn(b)) => err_eq(a, b),
_ => false,
}
}
}
impl Eq for NewThreadPoolError {}
impl From<StartThreadsError> for NewThreadPoolError {
fn from(err: StartThreadsError) -> Self {
match err {
StartThreadsError::NoThreads(e) | StartThreadsError::Respawn(e) => {
NewThreadPoolError::Spawn(e)
}
}
}
}
#[derive(Debug)]
pub enum TryScheduleError {
QueueFull,
/// The pool has no threads and `std::thread::Builder::spawn` returned the included error.
NoThreads(std::io::Error),
/// The pool has at least one thread and `std::thread::Builder::spawn` returned the included error.
Respawn(std::io::Error),
}
impl Display for TryScheduleError {
fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), std::fmt::Error> {
match self {
TryScheduleError::QueueFull => write!(f, "ThreadPool queue is full"),
TryScheduleError::NoThreads(e) => write!(
f,
"ThreadPool workers all panicked, failed starting replacement threads: {}",
e
),
TryScheduleError::Respawn(e) => {
write!(
f,
"ThreadPool failed starting threads to replace panicked threads: {}",
e
)
}
}
}
}
impl Error for TryScheduleError {}
impl PartialEq for TryScheduleError {
fn eq(&self, other: &Self) -> bool {
match (self, other) {
(TryScheduleError::QueueFull, TryScheduleError::QueueFull) => true,
(TryScheduleError::NoThreads(a), TryScheduleError::NoThreads(b))
| (TryScheduleError::Respawn(a), TryScheduleError::Respawn(b)) => err_eq(a, b),
_ => false,
}
}
}
impl Eq for TryScheduleError {}
impl From<StartThreadsError> for TryScheduleError {
fn from(err: StartThreadsError) -> Self {
match err {
StartThreadsError::NoThreads(e) => TryScheduleError::NoThreads(e),
StartThreadsError::Respawn(e) => TryScheduleError::Respawn(e),
}
}
}
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));
match recv_result {
Ok(f) => {
let _ignored = self.start_threads();
f();
}
Err(RecvTimeoutError::Timeout) => {}
// ThreadPool was dropped.
Err(RecvTimeoutError::Disconnected) => return,
};
let _ignored = self.start_threads();
}
}
fn start_thread(self: &Arc<Inner>) -> Result<(), StartThreadsError> {
let self_clone = self.clone();
let num_live_threads = self.num_live_threads() - 1;
if num_live_threads < self.size {
std::thread::Builder::new()
.name(format!("{}{}", self.name, self.next_name_num.next()))
.spawn(move || self_clone.work())
.map_err(|e| {
if num_live_threads == 0 {
StartThreadsError::NoThreads(e)
} else {
StartThreadsError::Respawn(e)
}
})?;
}
Ok(())
}
fn start_threads(self: &Arc<Inner>) -> Result<(), StartThreadsError> {
while self.num_live_threads() < self.size {
self.start_thread()?;
}
Ok(())
}
}
/// A collection of threads and a queue for jobs (`FnOnce` structs) they execute.
///
/// Threads stop 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).
///
/// After drop, threads stop as they become idle.
///
/// # 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).unwrap();
/// 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).unwrap());
/// 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".
///
/// After the `ThreadPool` struct drops, the threads continue processing
/// jobs and stop when the queue is empty.
///
/// # Errors
/// Returns an error when `name` is empty, `size` is zero, or it fails to start the threads.
pub fn new(name: &'static str, size: usize) -> Result<Self, NewThreadPoolError> {
if name.is_empty() {
return Err(NewThreadPoolError::Parameter(
"ThreadPool::new called with empty name".to_string(),
));
}
if size < 1 {
return Err(NewThreadPoolError::Parameter(format!(
"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()?;
Ok(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 or no threads are running.
/// See [`try_schedule`](#method.try_schedule).
///
/// Recreates any threads that panicked.
/// Retries on failure to start a new thread.
///
/// Puts `f` in a [`Box`](https://doc.rust-lang.org/stable/std/boxed/struct.Box.html) before
/// adding it to the queue.
#[allow(clippy::missing_panics_doc)]
pub fn schedule<F: FnOnce() + Send + 'static>(&self, f: F) {
let mut opt_box_f: Option<Box<dyn FnOnce() + Send + 'static>> = Some(Box::new(f));
loop {
match self.inner.start_threads() {
Ok(()) | Err(StartThreadsError::Respawn(_)) => {
// At least one thread is running.
}
Err(StartThreadsError::NoThreads(_)) => {
std::thread::sleep(Duration::from_millis(10));
continue;
}
}
opt_box_f = match self.sender.try_send(opt_box_f.take().unwrap()) {
Ok(()) => return,
Err(TrySendError::Disconnected(_)) => unreachable!(),
Err(TrySendError::Full(box_f)) => Some(box_f),
};
std::thread::sleep(Duration::from_millis(10));
}
}
/// Adds a job to the queue and then starts threads to replace any panicked threads.
/// The next idle thread will execute the job.
/// Starts jobs in FIFO order.
///
/// Puts `f` in a [`Box`](https://doc.rust-lang.org/stable/std/boxed/struct.Box.html) before
/// adding it to the queue.
///
/// # Errors
/// Returns an error when the queue is full or it fails to start a thread.
/// If the return value is not `TryScheduleError::QueueFull` then it added the job to the queue.
#[allow(clippy::missing_panics_doc)]
pub fn try_schedule(&self, f: impl FnOnce() + Send + 'static) -> Result<(), TryScheduleError> {
match self.sender.try_send(Box::new(f)) {
Ok(_) => {}
Err(TrySendError::Disconnected(_)) => unreachable!(),
Err(TrySendError::Full(_)) => return Err(TryScheduleError::QueueFull),
};
self.inner.start_threads().map_err(std::convert::Into::into)
}
}
impl Debug for ThreadPool {
fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), core::fmt::Error> {
write!(
f,
"ThreadPool{{{:?},size={:?}}}",
self.inner.name, self.inner.size
)
}
}
#[cfg(test)]
#[test]
fn threads_stop_after_pool_drops() {
let pool = ThreadPool::new("pool1", 3).unwrap();
let inner = pool.inner.clone();
std::thread::sleep(Duration::from_millis(100));
for _ in 0..3 {
pool.schedule(move || std::thread::sleep(Duration::from_millis(10)));
}
std::thread::sleep(Duration::from_millis(100));
drop(pool);
std::thread::sleep(Duration::from_millis(100));
assert_eq!(0, inner.num_live_threads());
}