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
//! [![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 rstarts panicked threads
//! - `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);
//! 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
//! - v0.1.3 - 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
//! - 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)]

#[cfg(test)]
mod tests;

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".
    ///
    /// After the `ThreadPool` struct drops, the threads continue processing
    /// jobs and stop when the queue is empty.
    ///
    /// # Panics
    /// Panics if `name` is empty or `size` is zero.
    #[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.
    #[allow(clippy::missing_panics_doc)]
    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
    }
}