sui_indexer_alt_framework/task.rs
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
// Copyright (c) Mysten Labs, Inc.
// SPDX-License-Identifier: Apache-2.0
use std::{future::Future, panic, pin::pin};
use futures::stream::{Stream, StreamExt};
use tokio::task::JoinSet;
/// Extension trait introducing `try_for_each_spawned` to all streams.
pub trait TrySpawnStreamExt: Stream {
/// Attempts to run this stream to completion, executing the provided asynchronous closure on
/// each element from the stream as elements become available.
///
/// This is similar to [StreamExt::for_each_concurrent], but it may take advantage of any
/// parallelism available in the underlying runtime, because each unit of work is spawned as
/// its own tokio task.
///
/// The first argument is an optional limit on the number of tasks to spawn concurrently.
/// Values of `0` and `None` are interpreted as no limit, and any other value will result in no
/// more than that many tasks being spawned at one time.
///
/// ## Safety
///
/// This function will panic if any of its futures panics, will return early with success if
/// the runtime it is running on is cancelled, and will return early with an error propagated
/// from any worker that produces an error.
fn try_for_each_spawned<Fut, F, E>(
self,
limit: impl Into<Option<usize>>,
f: F,
) -> impl Future<Output = Result<(), E>>
where
Fut: Future<Output = Result<(), E>> + Send + 'static,
F: FnMut(Self::Item) -> Fut,
E: Send + 'static;
}
impl<S: Stream + Sized + 'static> TrySpawnStreamExt for S {
async fn try_for_each_spawned<Fut, F, E>(
self,
limit: impl Into<Option<usize>>,
mut f: F,
) -> Result<(), E>
where
Fut: Future<Output = Result<(), E>> + Send + 'static,
F: FnMut(Self::Item) -> Fut,
E: Send + 'static,
{
// Maximum number of tasks to spawn concurrently.
let limit = match limit.into() {
Some(0) | None => usize::MAX,
Some(n) => n,
};
// Number of permits to spawn tasks left.
let mut permits = limit;
// Handles for already spawned tasks.
let mut join_set = JoinSet::new();
// Whether the worker pool has stopped accepting new items and is draining.
let mut draining = false;
// Error that occurred in one of the workers, to be propagated to the called on exit.
let mut error = None;
let mut self_ = pin!(self);
loop {
tokio::select! {
next = self_.next(), if !draining && permits > 0 => {
if let Some(item) = next {
permits -= 1;
join_set.spawn(f(item));
} else {
// If the stream is empty, signal that the worker pool is going to
// start draining now, so that once we get all our permits back, we
// know we can wind down the pool.
draining = true;
}
}
Some(res) = join_set.join_next() => {
match res {
Ok(Err(e)) if error.is_none() => {
error = Some(e);
permits += 1;
draining = true;
}
Ok(_) => permits += 1,
// Worker panicked, propagate the panic.
Err(e) if e.is_panic() => {
panic::resume_unwind(e.into_panic())
}
// Worker was cancelled -- this can only happen if its join handle was
// cancelled (not possible because that was created in this function),
// or the runtime it was running in was wound down, in which case,
// prepare the worker pool to drain.
Err(e) => {
assert!(e.is_cancelled());
permits += 1;
draining = true;
}
}
}
else => {
// Not accepting any more items from the stream, and all our workers are
// idle, so we stop.
if permits == limit && draining {
break;
}
}
}
}
if let Some(e) = error {
Err(e)
} else {
Ok(())
}
}
}
#[cfg(test)]
mod tests {
use std::{
sync::{
atomic::{AtomicUsize, Ordering},
Arc, Mutex,
},
time::Duration,
};
use futures::stream;
use super::*;
#[tokio::test]
async fn explicit_sequential_iteration() {
let actual = Arc::new(Mutex::new(vec![]));
let result = stream::iter(0..20)
.try_for_each_spawned(1, |i| {
let actual = actual.clone();
async move {
tokio::time::sleep(Duration::from_millis(20 - i)).await;
actual.lock().unwrap().push(i);
Ok::<(), ()>(())
}
})
.await;
assert!(result.is_ok());
let actual = Arc::try_unwrap(actual).unwrap().into_inner().unwrap();
let expect: Vec<_> = (0..20).collect();
assert_eq!(expect, actual);
}
#[tokio::test]
async fn concurrent_iteration() {
let actual = Arc::new(AtomicUsize::new(0));
let result = stream::iter(0..100)
.try_for_each_spawned(16, |i| {
let actual = actual.clone();
async move {
actual.fetch_add(i, Ordering::Relaxed);
Ok::<(), ()>(())
}
})
.await;
assert!(result.is_ok());
let actual = Arc::try_unwrap(actual).unwrap().into_inner();
let expect = 99 * 100 / 2;
assert_eq!(expect, actual);
}
#[tokio::test]
async fn implicit_unlimited_iteration() {
let actual = Arc::new(AtomicUsize::new(0));
let result = stream::iter(0..100)
.try_for_each_spawned(None, |i| {
let actual = actual.clone();
async move {
actual.fetch_add(i, Ordering::Relaxed);
Ok::<(), ()>(())
}
})
.await;
assert!(result.is_ok());
let actual = Arc::try_unwrap(actual).unwrap().into_inner();
let expect = 99 * 100 / 2;
assert_eq!(expect, actual);
}
#[tokio::test]
async fn explicit_unlimited_iteration() {
let actual = Arc::new(AtomicUsize::new(0));
let result = stream::iter(0..100)
.try_for_each_spawned(0, |i| {
let actual = actual.clone();
async move {
actual.fetch_add(i, Ordering::Relaxed);
Ok::<(), ()>(())
}
})
.await;
assert!(result.is_ok());
let actual = Arc::try_unwrap(actual).unwrap().into_inner();
let expect = 99 * 100 / 2;
assert_eq!(expect, actual);
}
#[tokio::test]
async fn max_concurrency() {
#[derive(Default, Debug)]
struct Jobs {
max: AtomicUsize,
curr: AtomicUsize,
}
let jobs = Arc::new(Jobs::default());
let result = stream::iter(0..32)
.try_for_each_spawned(4, |_| {
let jobs = jobs.clone();
async move {
jobs.curr.fetch_add(1, Ordering::Relaxed);
tokio::time::sleep(Duration::from_millis(100)).await;
let prev = jobs.curr.fetch_sub(1, Ordering::Relaxed);
jobs.max.fetch_max(prev, Ordering::Relaxed);
Ok::<(), ()>(())
}
})
.await;
assert!(result.is_ok());
let Jobs { max, curr } = Arc::try_unwrap(jobs).unwrap();
assert_eq!(curr.into_inner(), 0);
assert!(max.into_inner() <= 4);
}
#[tokio::test]
async fn error_propagation() {
let actual = Arc::new(Mutex::new(vec![]));
let result = stream::iter(0..100)
.try_for_each_spawned(None, |i| {
let actual = actual.clone();
async move {
if i < 42 {
actual.lock().unwrap().push(i);
Ok(())
} else {
Err(())
}
}
})
.await;
assert!(result.is_err());
let actual = Arc::try_unwrap(actual).unwrap().into_inner().unwrap();
let expect: Vec<_> = (0..42).collect();
assert_eq!(expect, actual);
}
#[tokio::test]
#[should_panic]
async fn panic_propagation() {
let _ = stream::iter(0..100)
.try_for_each_spawned(None, |i| async move {
assert!(i < 42);
Ok::<(), ()>(())
})
.await;
}
}