目录
前言
前几天逛github发现了一个有趣的并发库-conc,其目标是:
- 更难出现goroutine泄漏
- 处理panic更友好
- 并发代码可读性高
从简介上看主要封装功能如下:
waitGrouppanics.CatcherrecoverpanicworkerstreamForEachmap
接下来就区分模块来介绍一下这个库;
WaitGroup的封装
sync.waitGroup
func main(){ var wg sync.WaitGroup for i:=0; i < 10; i++{ wg.Add(1) go func() { defer wg.Done() defer func() { // recover panic err := recover() if err != nil { fmt.Println(err) } } // do something handle() } } wg.Wait() }
conc
func main() { wg := conc.NewWaitGroup() for i := 0; i < 10; i++ { wg.Go(doSomething) } wg.Wait() } func doSomething() { fmt.Println("test") }
conc
type WaitGroup struct { wg sync.WaitGroup pc panics.Catcher }
Catcher
type Catcher struct { recovered atomic.Pointer[RecoveredPanic] }
recovered是原子指针类型,RecoveredPanic是捕获的recover封装,封装了堆栈等信息:
type RecoveredPanic struct { // The original value of the panic. Value any // The caller list as returned by runtime.Callers when the panic was // recovered. Can be used to produce a more detailed stack information with // runtime.CallersFrames. Callers []uintptr // The formatted stacktrace from the goroutine where the panic was recovered. // Easier to use than Callers. Stack []byte }
提供了Try方法执行方法,只会记录第一个panic的goroutine信息:
func (p *Catcher) Try(f func()) { defer p.tryRecover() f() } func (p *Catcher) tryRecover() { if val := recover(); val != nil { rp := NewRecoveredPanic(1, val) // 只会记录第一个panic的goroutine信息 p.recovered.CompareAndSwap(nil, &rp) } }
Repanic()
func (p *Catcher) Repanic() { if val := p.Recovered(); val != nil { panic(val) } } func (p *Catcher) Recovered() *RecoveredPanic { return p.recovered.Load() }
waitGroupWait()WaitAndRecover()
func (h *WaitGroup) Wait() { h.wg.Wait() // Propagate a panic if we caught one from a child goroutine. h.pc.Repanic() } func (h *WaitGroup) WaitAndRecover() *panics.RecoveredPanic { h.wg.Wait() // Return a recovered panic if we caught one from a child goroutine. return h.pc.Recovered() }
wait
waitAndRecover
waitGrouop
worker池
conc
- ContextPool:可以传递context的pool,若有goroutine发生错误可以cancel其他goroutine
- ErrorPool:通过参数可以控制只收集第一个error还是所有error
- ResultContextPool:若有goroutine发生错误会cancel其他goroutine并且收集错误
- RestultPool:收集work池中每个任务的执行结果,并不能保证顺序,保证顺序需要使用stream或者iter.map;
我们来看一个简单的例子:
import "github.com/sourcegraph/conc/pool" func ExampleContextPool_WithCancelOnError() { p := pool.New(). WithMaxGoroutines(4). WithContext(context.Background()). WithCancelOnError() for i := 0; i < 3; i++ { i := i p.Go(func(ctx context.Context) error { if i == 2 { return errors.New("I will cancel all other tasks!") } <-ctx.Done() return nil }) } err := p.Wait() fmt.Println(err) // Output: // I will cancel all other tasks! }
在创建pool时有如下方法可以调用:
p.WithMaxGoroutines()p.WithErrorsp.WithContext(ctx)p.WithFirstErrorp.WithCollectErrored
pool的基础结构如下:
type Pool struct { handle conc.WaitGroup limiter limiter tasks chan func() initOnce sync.Once }
limiter是控制器,用chan来控制goroutine的数量:
type limiter chan struct{} func (l limiter) limit() int { return cap(l) } func (l limiter) release() { if l != nil { <-l } }
pool的核心逻辑也比较简单,如果没有设置limiter,那么就看有没有空闲的worker,否则就创建一个新的worker,然后投递任务进去;
如果设置了limiter,达到了limiter worker数量上限,就把任务投递给空闲的worker,没有空闲就阻塞等着;
func (p *Pool) Go(f func()) { p.init() if p.limiter == nil { // 没有限制 select { case p.tasks <- f: // A goroutine was available to handle the task. default: // No goroutine was available to handle the task. // Spawn a new one and send it the task. p.handle.Go(p.worker) p.tasks <- f } } else { select { case p.limiter <- struct{}{}: // If we are below our limit, spawn a new worker rather // than waiting for one to become available. p.handle.Go(p.worker) // We know there is at least one worker running, so wait // for it to become available. This ensures we never spawn // more workers than the number of tasks. p.tasks <- f case p.tasks <- f: // A worker is available and has accepted the task. return } } }
这里work使用的是一个无缓冲的channel,这种复用方式很巧妙,如果goroutine执行很快避免创建过多的goroutine;
Stream
Stream
pool
func ExampleStream() { times := []int{20, 52, 16, 45, 4, 80} stream := stream2.New() for _, millis := range times { dur := time.Duration(millis) * time.Millisecond stream.Go(func() stream2.Callback { time.Sleep(dur) // This will print in the order the tasks were submitted return func() { fmt.Println(dur) } }) } stream.Wait() // Output: // 20ms // 52ms // 16ms // 45ms // 4ms // 80ms }
stream
type Stream struct { pool pool.Pool callbackerHandle conc.WaitGroup queue chan callbackCh initOnce sync.Once }
queue
type callbackCh chan func()
goroutine
func (s *Stream) Go(f Task) { s.init() // Get a channel from the cache. ch := getCh() // Queue the channel for the callbacker. s.queue <- ch // Submit the task for execution. s.pool.Go(func() { defer func() { // In the case of a panic from f, we don't want the callbacker to // starve waiting for a callback from this channel, so give it an // empty callback. if r := recover(); r != nil { ch <- func() {} panic(r) } }() // Run the task, sending its callback down this task's channel. callback := f() ch <- callback }) } var callbackChPool = sync.Pool{ New: func() any { return make(callbackCh, 1) }, } func getCh() callbackCh { return callbackChPool.Get().(callbackCh) } func putCh(ch callbackCh) { callbackChPool.Put(ch) }
ForEach和map
ForEach
conc库提供了ForEach方法可以优雅的并发处理切片,看一下官方的例子:
conc库使用泛型进行了封装,我们只需要关注handle代码即可,避免冗余代码,我们自己动手写一个例子:
func main() { input := []int{1, 2, 3, 4} iterator := iter.Iterator[int]{ MaxGoroutines: len(input) / 2, } iterator.ForEach(input, func(v *int) { if *v%2 != 0 { *v = -1 } }) fmt.Println(input) }
ForEach内部实现为Iterator结构及核心逻辑如下:
type Iterator[T any] struct { MaxGoroutines int } func (iter Iterator[T]) ForEachIdx(input []T, f func(int, *T)) { if iter.MaxGoroutines == 0 { // iter is a value receiver and is hence safe to mutate iter.MaxGoroutines = defaultMaxGoroutines() } numInput := len(input) if iter.MaxGoroutines > numInput { // No more concurrent tasks than the number of input items. iter.MaxGoroutines = numInput } var idx atomic.Int64 // 通过atomic控制仅创建一个闭包 task := func() { i := int(idx.Add(1) - 1) for ; i < numInput; i = int(idx.Add(1) - 1) { f(i, &input[i]) } } var wg conc.WaitGroup for i := 0; i < iter.MaxGoroutines; i++ { wg.Go(task) } wg.Wait() }
可以设置并发的goroutine数量,默认取的是GOMAXPROCS ,也可以自定义传参;
并发执行这块设计的很巧妙,仅创建了一个闭包,通过atomic控制idx,避免频繁触发GC;
map
conc库提供的map方法可以得到对切片中元素结果,官方例子:
使用map可以提高代码的可读性,并且减少了冗余代码,自己写个例子:
func main() { input := []int{1, 2, 3, 4} mapper := iter.Mapper[int, bool]{ MaxGoroutines: len(input) / 2, } results := mapper.Map(input, func(v *int) bool { return *v%2 == 0 }) fmt.Println(results) // Output: // [false true false true] }
map的实现也依赖于Iterator,也是调用的ForEachIdx方法,区别于ForEach是记录处理结果;
总结
花了小半天时间看了一下这个库,很多设计点值得我们学习,总结一下我学习到的知识点:
- conc.WatiGroup对Sync.WaitGroup进行了封装,对Add、Done、Recover进行了封装,提高了可读性,避免了冗余代码
- ForEach、Map方法可以更优雅的并发处理切片,代码简洁易读,在实现上Iterator中的并发处理使用atomic来控制只创建一个闭包,避免了GC性能问题
- pool是一个并发的协程队列,可以控制协程的数量,实现上也很巧妙,使用一个无缓冲的channel作为worker,如果goroutine执行速度快,避免了创建多个goroutine
- stream是一个保证顺序的并发协程队列,实现上也很巧妙,使用sync.Pool在提交goroutine时控制顺序,值得我们学习;