Go语言sync package提供了条件变量(condition variable)类型:

type Cond struct {
        // L is held while observing or changing the condition
        L Locker
        // contains filtered or unexported fields
}
type Cond
    func NewCond(l Locker) *Cond
    func (c *Cond) Broadcast()
    func (c *Cond) Signal()
    func (c *Cond) Wait()
type Locker

type Locker interface {
        Lock()
        Unlock()
}
A Locker represents an object that can be locked and unlocked.

NewCond()函数输入参数是一个Locker接口类型,即实现了锁功能的变量。Broadcast()函数通知所有等待在condition variablegoroutine,而Signal()函数只会通知其中的一个goroutineWait()会让goroutine阻塞在condition variable,等待条件成立。通常的做法是:

c.L.Lock()
for !condition() {
    c.Wait()
}
... make use of condition ...
c.L.Unlock()

进入Wait()函数会解锁,离开Wait()函数会重新加锁。由于在“解锁->收到通知->重新加锁”这段逻辑中间有可能另一个同样wait()goroutine抢先一步改变了条件,导致当前goroutine的条件不再成立,所以这块要使用循环检测。参考下例:

package main

import (
    "sync"
    "time"
    "fmt"
)

func main() {
    c := sync.NewCond(&sync.Mutex{})
    var num int

    for i := 1; i <= 2; i++ {
        go func(id int) {
            fmt.Println("Enter Thread ID:", id)
            c.L.Lock()
            for num != 1 {
                fmt.Println("Enter loop: Thread ID:", id)
                c.Wait()
                fmt.Println("Exit loop: Thread ID:", id)
            }
            num++
            c.L.Unlock()
            fmt.Println("Exit Thread ID:", id)
        }(i)
    }

    time.Sleep(time.Second)
    fmt.Println("Sleep 1 second")

    num++
    c.Broadcast()
    time.Sleep(time.Second)
    fmt.Println("Program exit")
}

一次执行结果如下:

Enter Thread ID: 2
Enter loop: Thread ID: 2
Enter Thread ID: 1
Enter loop: Thread ID: 1
Sleep 1 second
Exit loop: Thread ID: 2
Exit Thread ID: 2
Exit loop: Thread ID: 1
Enter loop: Thread ID: 1
Program exit

从上面例子可以看出由于goroutine 2改变了条件,导致goroutine 1重新进入循环,即多个goroutine阻塞在一个condition variable上存在着竞争的关系。

参考资料:
Package sync
Condition Variables