goroutine和channel应用——处理队列

1 最简单的worker处理队列方式

package main

import "time"

type Job int

func worker(jobChan <-chan Job) {
    for job := range jobChan {
        // 顺序执行，缺点：如果处理过程中有等待或阻塞，会影响整个队列
        Process(job)

        // 并发执行，如果处理过程中有等待或阻塞，不会影响其他的job，
        // 缺点：并发处理job的goroutine数量不可控，每来一个新job就会启动一个goroutine，不建议这样处理。
        // 通常做法是开启有限个worker的goroutine来并行处理队列的job，而不是在process里并发执行。
        //go Process(job)
    }
}

func Process(job Job) {
    if job == 3 { // 等于3的这个job阻塞1s
        time.Sleep(time.Second)
    }
    println("job", job)
}

func main() {
    // make a channel with a capacity of 10.
    jobChan := make(chan Job, 10)

    // start the worker
    go worker(jobChan)

    // enqueue a job
    for i := 0; i < 20; i++ {
        jobChan <- Job(i)
    }

    time.Sleep(2 * time.Second)
}

/* 当次执行的结果如下：

(1) 顺序执行，在处理job 3时阻塞了一秒，其他job要等job 3处理完毕后再处理往下执行。
job 0
job 1
job 2
job 3
job 4
job 5
job 6
job 7
job 8
job 9
job 10
job 11
job 12
job 13
job 14
job 15
job 16
job 17
job 18
job 19

------------------------------------------------------

(2) 并发process执行，在处理job 3时阻塞了一秒，不影响其他的job处理。
job 2
job 10
job 0
job 6
job 4
job 5
job 9
job 7
job 1
job 8
job 13
job 11
job 12
job 19
job 17
job 18
job 14
job 15
job 16
job 3
*/

2 使用worker池处理队列

package main

import (
    "fmt"
    "time"
)

type Job int

func worker(i int, jobChan <-chan Job) {
    for job := range jobChan {
        Process(i, job)
    }
}

func Process(i int, job Job) {
    if job == 3 {
        time.Sleep(time.Second)
    }
    fmt.Printf("worker %2d process job %d\n", i, job)
}

func workPool(workerNum int, jobChan chan Job) {
    for i := 0; i < workerNum; i++ {
        go func(i int) {
            worker(i, jobChan)
        }(i)
    }
}

func main() {
    jobChan := make(chan Job, 10)

    // 启动多个worker池并发处理队列的job，多个worker去抢队列job来处，只有空闲的worker才能从队列中获取job
    workPool(5, jobChan)

    for i := 0; i < 20; i++ {
        jobChan <- Job(i)
    }

    time.Sleep(2 * time.Second)
}

/*当次执行结果如下：

可以看到worker 2处理job 3时阻塞一秒，worker 2在阻塞过程中没有去抢占队列新的job来处理。
worker  4 process job 0
worker  3 process job 4
worker  0 process job 2
worker  3 process job 6
worker  3 process job 8
worker  3 process job 9
worker  3 process job 10
worker  3 process job 11
worker  3 process job 12
worker  3 process job 13
worker  3 process job 14
worker  3 process job 15
worker  3 process job 16
worker  3 process job 17
worker  3 process job 18
worker  1 process job 1
worker  4 process job 5
worker  0 process job 7
worker  3 process job 19
worker  2 process job 3
*/

3 等待worker处理所有队列的job

package main

import (
    "fmt"
    "math/rand"
    "sync"
    "time"
)

type Job int

func worker(n int, jobChan <-chan Job, wg *sync.WaitGroup) {
    for job := range jobChan {
        Process(n, job, wg)
    }
}

func Process(n int, job Job, wg *sync.WaitGroup) {
    defer wg.Done()

    // 加上随机延时，方便查看打印效果
    size := randTime()
    time.Sleep(size * time.Millisecond)

    fmt.Printf("worker %2d process job %d, time %dms\n", n, job, size)
}

// 随机时间100~500
func randTime() time.Duration {
    rand.Seed(time.Now().UnixNano())
    return time.Duration(rand.Intn(400) + 100)
}

func WaitTimeout(wg *sync.WaitGroup, timeout time.Duration) bool {
    ch := make(chan struct{})

    go func() {
        wg.Wait()
        close(ch)
    }()

    select {
    case <-ch:
        return true
    case <-time.After(timeout):
        return false
    }
}

func workPool(workerNum int, jobChan chan Job, wg *sync.WaitGroup) {
    for i := 0; i < workerNum; i++ {
        go func(i int) {
            worker(i, jobChan, wg)
        }(i)
    }
}

func main() {
    wg := &sync.WaitGroup{}
    jobChan := make(chan Job, 10)

    workPool(5, jobChan, wg)

    for i := 0; i < 20; i++ {
        wg.Add(1)
        jobChan <- Job(i)
    }

    t := time.Now()

    // 等待所有job处理完毕才退出，不足：不管是顺序执行还是并发process方式处理job，如果其中有一个job阻塞了，会一直等待下去
    wg.Wait()

    // 带超时等待，如果超时了，直接忽略等待
    //ok := WaitTimeout(wg, 300*time.Millisecond)
    //if !ok {
    //  fmt.Printf("\n warning, process job timeout \n")
    //}

    fmt.Printf("\n handle queue time %v\n", time.Now().Sub(t))
}

/* 当次执行结果如下：

(1) 不带超时的wait
worker  2 process job 2, time 147ms
worker  0 process job 0, time 147ms
worker  1 process job 1, time 147ms
worker  3 process job 4, time 147ms
worker  4 process job 3, time 147ms
worker  1 process job 7, time 282ms
worker  4 process job 9, time 282ms
worker  0 process job 6, time 282ms
worker  3 process job 8, time 282ms
worker  2 process job 5, time 399ms
worker  1 process job 10, time 149ms
worker  0 process job 12, time 149ms
worker  3 process job 13, time 149ms
worker  4 process job 11, time 149ms
worker  2 process job 14, time 145ms
worker  0 process job 16, time 265ms
worker  1 process job 15, time 265ms
worker  3 process job 17, time 265ms
worker  4 process job 18, time 265ms
worker  2 process job 19, time 234ms

 handle queue time 786.8609ms

------------------------------------------------

(2) 有超时的wait
worker  3 process job 3, time 257ms
worker  1 process job 1, time 272ms
worker  0 process job 0, time 272ms
worker  4 process job 4, time 272ms
worker  2 process job 2, time 272ms
worker  4 process job 8, time 102ms
worker  3 process job 5, time 224ms

 warning, process job timeout

 handle queue time 300.3939ms
*/

4 使用context或channel停止worker

有两种方式停止处理队列中的worker，分别是context和channel，多级函数传递或复杂点的控制建议使用context。

4.1 使用context停止worker

package main

import (
    "fmt"
    "math/rand"
    "time"

    "golang.org/x/net/context"
)

type Job int

func workPool(workerNum int, jobChan chan Job, ctx context.Context) {
    for i := 0; i < workerNum; i++ {
        go func(i int) {
            worker(i, jobChan, ctx)
        }(i)
    }
}

// 通过context取消未完成的job
func worker(n int, jobChan <-chan Job, ctx context.Context) {
    for {
        select {
        case job := <-jobChan:
            Process(n, job)

        case <-ctx.Done():
            fmt.Printf("cancel worker %d\n", n)
            return
        }
    }
}

func Process(n int, job Job) {
    size := randTime()
    time.Sleep(size * time.Millisecond)
    fmt.Printf("worker %2d process job %2d, time %dms\n", n, job, size)
}

// 随机时间100~500
func randTime() time.Duration {
    rand.Seed(time.Now().UnixNano())
    return time.Duration(rand.Intn(400) + 100)
}

func main() {
    jobChan := make(chan Job, 10)

    // 使用context控制worker是否停止，适合多级函数传递和控制，并且有超时取消
    ctx, cancel := context.WithCancel(context.Background())
    //ctx, cancel := context.WithTimeout(context.Background(), time.Second) // 可以设置延时的context
    workPool(5, jobChan, ctx)
    for i := 0; i < 20; i++ {
        jobChan <- Job(i)
    }

    cancel()
    time.Sleep(5 * time.Second)
}

/* 当次执行结果如下：
worker  2 process job  3, time 471ms
worker  3 process job  4, time 471ms
worker  1 process job  1, time 471ms
worker  4 process job  0, time 471ms
worker  0 process job  2, time 471ms
worker  3 process job  6, time 200ms
cancel worker 3
worker  4 process job  8, time 200ms
worker  2 process job  5, time 200ms
cancel worker 2
worker  0 process job  9, time 200ms
cancel worker 0 by context
worker  1 process job  7, time 200ms
cancel worker 1
worker  4 process job 10, time 310ms
cancel worker 4
*/

4.2 使用channel停止worker

package main

import (
    "fmt"
    "math/rand"
    "time"
)

type Job int

func workPool(workerNum int, jobChan chan Job, ch chan struct{}) {
    for i := 0; i < workerNum; i++ {
        go func(i int) {
            worker(i, jobChan, ch)
        }(i)
    }
}

// 通过channel取消未完成的job
func worker(n int, jobChan <-chan Job, ch chan struct{}) {
    for {
        select {
        case job := <-jobChan:
            Process(n, job)

        case <-ch:
            fmt.Printf("cancel worker %d\n", n)
            return
        }
    }
}

func Process(n int, job Job) {
    size := randTime()
    time.Sleep(size * time.Millisecond)
    fmt.Printf("worker %2d process job %2d, time %dms\n", n, job, size)
}

// 随机时间100~500
func randTime() time.Duration {
    rand.Seed(time.Now().UnixNano())
    return time.Duration(rand.Intn(400) + 100)
}

func main() {
    jobChan := make(chan Job, 10)

    // 使用channel控制worker是否停止
    ch := make(chan struct{})
    workPool(5, jobChan, ch)
    for i := 0; i < 20; i++ {
        jobChan <- Job(i)
    }

    close(ch)
    time.Sleep(5 * time.Second)

}

/* 当次执行结果如下：
worker  1 process job  1, time 313ms
worker  3 process job  4, time 313ms
worker  2 process job  3, time 313ms
worker  0 process job  2, time 313ms
worker  4 process job  0, time 313ms
worker  3 process job  6, time 258ms
cancel worker 3
worker  2 process job  7, time 258ms
worker  4 process job  9, time 258ms
cancel worker 4
worker  0 process job  8, time 258ms
worker  1 process job  5, time 258ms
cancel worker 1
worker  0 process job 11, time 190ms
cancel worker 0
worker  2 process job 10, time 226ms
worker  2 process job 12, time 485ms
cancel worker 2
*/

专题「golang相关」的其它文章 »

使用开发框架sponge快速把单体web服务拆分为微服务 (Sep 18, 2023)
使用开发框架sponge一天多开发完成一个简单版社区后端服务 (Jul 30, 2023)
sponge —— 一个强大的go开发框架，以 (Jan 06, 2023)
go test命令 (Apr 15, 2022)
go应用程序性能分析 (Mar 29, 2022)
channel原理和应用 (Mar 22, 2022)
go runtime (Mar 14, 2022)
go调试工具 (Mar 13, 2022)
cobra (Mar 10, 2022)
grpc使用实践 (Nov 27, 2020)
配置文件viper库 (Nov 22, 2020)
根据服务名称查看golang程序的profile信息 (Sep 03, 2019)
go语言开发规范 (Aug 28, 2019)
golang中的context包 (Aug 28, 2018)