Java程序员的Golang入门指南(下)
Java程序员的Golang入门指南(下)
4.高级特性
上面介绍的只是Golang的基本语法和特性,尽管像控制语句的条件不用圆括号、函数多返回值、switch-case默认break、函数闭包、集合切片等特性相比Java的确提高了开发效率,但这些在其他语言中也都有,并不是Golang能真正吸引人的地方。不仅是Golang,我们学习任何语言当然都是从基本语法特性着手,但学习时要不断地问自己:使这门语言区别于其他语言的”独到之处“在哪?这种独到之处往往反映了语言的设计思想、出发点、要解决的”痛点“,这才是一门语言或任何技术的立足之本。
4.1 goroutine
goroutine使用go关键字来调用函数,也可以使用匿名函数。可以简单的把go关键字调用的函数想像成pthread_create。如果一个goroutine没有被阻塞,那么别的goroutine就不会得到执行。也就是说goroutine阻塞时,Golang会切换到其他goroutine执行,这是非常好的特性!Java对类似goroutine这种的协程没有原生支持,像Akka最害怕的就是阻塞。因为协程不等同于线程,操作系统不会帮我们完成“现场”保存和恢复,所以要实现goroutine这种特性,就要模拟操作系统的行为,保存方法或函数在协程“上下文切换”时的Context,当阻塞结束时才能正确地切换回来。像Kilim等协程库利用字节码生成,能够胜任,而Akka完全是运行时的。
注意:如果你要真正的并发,需要调用runtime.GOMAXPROCS(CPU_NUM)设置。
package main
import "fmt"
func main() {
go f("goroutine")
go func(msg string) {
fmt.Println(msg)
}("going")
// Block main thread
var input string
fmt.Scanln(&input)
fmt.Println("done")
}
func f(msg string) {
fmt.Println(msg)
}
4.2 原子操作
像Java一样,Golang支持很多CAS操作。运行结果是unsaftCnt可能小于200,因为unsafeCnt++在机器指令层面上不是一条指令,而可能是从内存加载数据到寄存器、执行自增运算、保存寄存器中计算结果到内存这三部分,所以不进行保护的话有些更新是会丢失的。
package main
import (
"fmt"
"time"
"sync/atomic"
"runtime"
)
func main() {
// IMPORTANT!!!
runtime.GOMAXPROCS(4)
// thread-unsafe
var unsafeCnt int32 = 0
for i := 0; i < 10; i++ {
go func() {
for i := 0; i < 20; i++ {
time.Sleep(time.Millisecond)
unsafeCnt++
}
}()
}
time.Sleep(time.Second)
fmt.Println("cnt: ", unsafeCnt)
// CAS toolkit
var cnt int32 = 0
for i := 0; i < 10; i++ {
go func() {
for i := 0; i < 20; i++ {
time.Sleep(time.Millisecond)
atomic.AddInt32(&cnt, 1)
}
}()
}
time.Sleep(time.Second)
cntFinal := atomic.LoadInt32(&cnt)
fmt.Println("cnt: ", cntFinal)
}
神奇CAS的原理
Golang的AddInt32()类似于Java中AtomicInteger.incrementAndGet(),其伪代码可以表示如下。二者的基本思想是一致的,本质上是 乐观锁:首先,从内存位置M加载要修改的数据到寄存器A中;然后,修改数据并保存到另一寄存器B;最终,利用CPU提供的CAS指令(Java通过JNI调用到)用一条指令完成:1)A值与M处的原值比较;2)若相同则将B值覆盖到M处。
若不相同,则CAS指令会失败,说明从内存加载到执行CAS指令这一小段时间内,发生了上下文切换,执行了其他线程的代码修改了M处的变量值。那么重新执行前面几个步骤再次尝试。
ABA问题:即另一线程修改了M位置的数据,但是从原值改为C,又从C改回原值。这样上下文切换回来,CAS指令发现M处的值“未改变”(实际是改了两次,最后改回来了),所以CAS指令正常执行,不会失败。这种问题在Java中可以用AtomicStampedReference/AtomicMarkableReference解决。
public final int incrementAndGet() {
for (;;) {
int current = get();
int next = current + 1;
if (compareAndSet(current, next))
return next;
}
}
4.3 Channel管道
通过前面可以看到,尽管goroutine很方便很高效,但如果滥用的话很可能会导致并发安全问题。而Channel就是用来解决这个问题的,它是goroutine之间通信的桥梁,类似Actor模型中每个Actor的mailbox。多个goroutine要修改一个状态时,可以将请求都发送到一个Channel里,然后由一个goroutine负责顺序地修改状态。
Channel默认是阻塞的,也就是说select时如果没有事件,那么当前goroutine会发生读阻塞。同理,Channel是有大小的,当Channel满了时,发送方会发生写阻塞。Channel这种阻塞的特性加上goroutine可以很容易就能实现生产者-消费者模式。
用case可以给Channel设置阻塞的超时时间,避免一直阻塞。而default则使select进入无阻塞模式。
package main
import (
"fmt"
"time"
)
/**
* Output:
* received message: hello
* received message: world
*
* received from channel-1: Hello
* received from channel-2: World
*
* received message: hello
* Time out!
*
* Nothing received!
* received message: hello
* Nothing received!
* Nothing received!
* Nothing received!
* Nothing received!
* Nothing received!
* Nothing received!
* Nothing received!
* Nothing received!
* Nothing received!
* received message: world
* Nothing received!
* Nothing received!
* Nothing received!
*/
func main() {
listenOnChannel()
selectTwoChannels()
blockChannelWithTimeout()
unblockChannel()
}
func listenOnChannel() {
// Specify channel type and buffer size
channel := make(chan string, 5)
go func() {
channel <- "hello"
channel <- "world"
}()
for i := 0; i < 2; i++ {
msg := <- channel
fmt.Println("received message: " + msg)
}
}
func selectTwoChannels() {
c1 := make(chan string)
c2 := make(chan string)
go func() {
time.Sleep(time.Second)
c1 <- "Hello"
}()
go func() {
time.Sleep(time.Second)
c2 <- "World"
}()
for i := 0; i < 2; i++ {
select {
case msg1 := <- c1:
fmt.Println("received from channel-1: " + msg1)
case msg2 := <- c2:
fmt.Println("received from channel-2: " + msg2)
}
}
}
func blockChannelWithTimeout() {
channel := make(chan string, 5)
go func() {
channel <- "hello"
// Sleep 10 sec
time.Sleep(time.Second * 10)
channel <- "world"
}()
for i := 0; i < 2; i++ {
select {
case msg := <- channel:
fmt.Println("received message: " + msg)
// Set timeout 5 sec
case <- time.After(time.Second * 5):
fmt.Println("Time out!")
}
}
}
func unblockChannel() {
channel := make(chan string, 5)
go func() {
channel <- "hello"
time.Sleep(time.Second * 10)
channel <- "world"
}()
for i := 0; i < 15; i++ {
select {
case msg := <- channel:
fmt.Println("received message: " + msg)
default:
fmt.Println("Nothing received!")
time.Sleep(time.Second)
}
}
}
4.4 缓冲流
Golang的bufio包提供了方便的缓冲流操作,通过strings或网络IO得到流后,用bufio.NewReader/Writer()包装:
- 缓冲区:Peek()或Read时,数据会从底层进入到缓冲区。缓冲区默认大小为4096字节。
- 切片和拷贝:Peek()和ReadSlice()得到的都是切片(缓冲区数据的引用)而不是拷贝,所以更加节约空间。但是当缓冲区数据变化时,切片也会随之变化。而ReadBytes/String()得到的都是数据的拷贝,可以放心使用。
- Unicode支持:ReadRune()可以直接读取Unicode字符。有意思的是Golang中Unicode字符也要用单引号,这点与Java不同。
- 分隔符:ReadSlice/Bytes/String()得到的包含分隔符,bufio不会自动去掉。
- Writer:对应地,Writer提供了WriteBytes/String/Rune。
- undo方法:可以将读出的字节再放回到缓冲区,就像什么都没发生一样。
package main
import (
"fmt"
"strings"
"bytes"
"bufio"
)
/**
* Buffered: 0
* Buffered after peek: 7
* ABCDE
* AxCDE
*
* abcdefghijklmnopqrst 20 <nil>
* uvwxyz1234567890 16 <nil>
* 0 EOF
*
* "ABC "
* "DEF "
* "GHI"
*
* "ABC "
* "DEF "
* "GHI"
*
* read unicode=[你], size=[3]
* read unicode=[好], size=[3]
* read(after undo) unicode=[好], size=[3]
*
* Available: 4096
* Buffered: 0
* Available after write: 4088
* Buffered after write: 8
* Buffer after write: ""
* Available after flush: 4096
* Buffered after flush: 0
* Buffer after flush: "ABCDEFGH"
*
* Hello,世界!
*/
func main() {
testPeek()
testRead()
testReadSlice()
testReadBytes()
testReadUnicode()
testWrite()
testWriteByte()
}
func testPeek() {
r := strings.NewReader("ABCDEFG")
br := bufio.NewReader(r)
fmt.Printf("Buffered: %d\n", br.Buffered())
p, _ := br.Peek(5)
fmt.Printf("Buffered after peek: %d\n", br.Buffered())
fmt.Printf("%s\n", p)
p[1] = 'x'
p, _ = br.Peek(5)
fmt.Printf("%s\n", p)
}
func testRead() {
r := strings.NewReader("abcdefghijklmnopqrstuvwxyz1234567890")
br := bufio.NewReader(r)
b := make([]byte, 20)
n, err := br.Read(b)
fmt.Printf("%-20s %-2v %v\n", b[:n], n, err)
n, err = br.Read(b)
fmt.Printf("%-20s %-2v %v\n", b[:n], n, err)
n, err = br.Read(b)
fmt.Printf("%-20s %-2v %v\n", b[:n], n, err)
}
func testReadSlice() {
r := strings.NewReader("ABC DEF GHI")
br := bufio.NewReader(r)
w, _ := br.ReadSlice(' ')
fmt.Printf("%q\n", w)
w, _ = br.ReadSlice(' ')
fmt.Printf("%q\n", w)
w, _ = br.ReadSlice(' ')
fmt.Printf("%q\n", w)
}
func testReadBytes() {
r := strings.NewReader("ABC DEF GHI")
br := bufio.NewReader(r)
w, _ := br.ReadBytes(' ')
fmt.Printf("%q\n", w)
w, _ = br.ReadSlice(' ')
fmt.Printf("%q\n", w)
s, _ := br.ReadString(' ')
fmt.Printf("%q\n", s)
}
func testReadUnicode() {
r := strings.NewReader("你好,世界!")
br := bufio.NewReader(r)
c, size, _ := br.ReadRune()
fmt.Printf("read unicode=[%c], size=[%v]\n", c, size)
c, size, _ = br.ReadRune()
fmt.Printf("read unicode=[%c], size=[%v]\n", c, size)
br.UnreadRune()
c, size, _ = br.ReadRune()
fmt.Printf("read(after undo) unicode=[%c], size=[%v]\n", c, size)
}
func testWrite() {
b := bytes.NewBuffer(make([]byte, 0))
bw := bufio.NewWriter(b)
fmt.Printf("Available: %d\n", bw.Available())
fmt.Printf("Buffered: %d\n", bw.Buffered())
bw.WriteString("ABCDEFGH")
fmt.Printf("Available after write: %d\n", bw.Available())
fmt.Printf("Buffered after write: %d\n", bw.Buffered())
fmt.Printf("Buffer after write: %q\n", b)
bw.Flush()
fmt.Printf("Available after flush: %d\n", bw.Available())
fmt.Printf("Buffered after flush: %d\n", bw.Buffered())
fmt.Printf("Buffer after flush: %q\n", b)
}
func testWriteByte() {
b := bytes.NewBuffer(make([]byte, 0))
bw := bufio.NewWriter(b)
bw.WriteByte('H')
bw.WriteByte('e')
bw.WriteByte('l')
bw.WriteByte('l')
bw.WriteByte('o')
bw.WriteString(",")
bw.WriteRune('世')
bw.WriteRune('界')
bw.WriteRune('!')
bw.Flush()
fmt.Println(b)
}
4.5 并发控制
sync包中的WaitGroup是个很有用的类,类似信号量。wg.Add()和Done()能够加减WaitGroup(信号量)的值,而Wait()会挂起当前线程直到信号量变为0。下面的例子用WaitGroup的值表示正在运行的goroutine数量。在goroutine中,用defer Done()确保goroutine正常或异常退出时,WaitGroup都能减一。
package main
import (
"fmt"
"sync"
)
/**
* I'm waiting all goroutines on wg done
* I'm done=[0]
* I'm done=[1]
* I'm done=[2]
* I'm done=[3]
* I'm done=[4]
* I'm done=[5]
* I'm done=[6]
* I'm done=[7]
* I'm done=[8]
* I'm done=[9]
*/
func main() {
var wg sync.WaitGroup
for i := 0; i < 10; i++ {
wg.Add(1)
go func(id int) {
defer wg.Done()
fmt.Printf("I'm done=[%d]\n", id)
}(i)
}
fmt.Println("I'm waiting all goroutines on wg done")
wg.Wait()
}
4.6 网络编程
Golang的net包的抽象层次还是挺高的,用不了几行代码就能实现一个简单的TCP或HTTP服务端了。
4.6.1 Socket编程
package main
import (
"net"
"fmt"
"io"
)
/**
* Starting the server
* Accept the connection: 127.0.0.1:14071
* Warning: End of data EOF
*/
func main() {
listener, err := net.Listen("tcp", "127.0.0.1:12345")
if err != nil {
panic("error listen: " + err.Error())
}
fmt.Println("Starting the server")
for {
conn, err := listener.Accept()
if err != nil {
panic("error accept: " + err.Error())
}
fmt.Println("Accept the connection: ", conn.RemoteAddr())
go echoServer(conn)
}
}
func echoServer(conn net.Conn) {
buf := make([]byte, 1024)
defer conn.Close()
for {
n, err := conn.Read(buf)
switch err {
case nil:
conn.Write(buf[0:n])
case io.EOF:
fmt.Printf("Warning: End of data %s\n", err)
return
default:
fmt.Printf("Error: read data %s\n", err)
return
}
}
}
4.6.2 Http服务器
package main
import (
"fmt"
"log"
"net/http"
)
func main() {
http.HandleFunc("/hello", handleHello)
fmt.Println("serving on http://localhost:7777/hello")
log.Fatal(http.ListenAndServe("localhost:7777", nil))
}
func handleHello(w http.ResponseWriter, req *http.Request) {
log.Println("serving", req.URL)
fmt.Fprintln(w, "Hello, world!")
}
5.结束语
5.1 Golang初体验
Golang的某些语法的确很简洁,像行尾无分号、条件语句无括号、类型推断、函数多返回值、异常处理、原生协程支持、DuckType继承等,尽管很多并不是Golang首创,但结合到一起写起来还是很舒服的。
当然Golang也有让人“不爽”的地方。像变量和函数中的类型声明写在后面简直是“反人类”!同样是颠覆,switch的case默认会break就很实用。另外,因为Golang主要还是想替代C做系统开发,所以像类啊、包啊还是能看到C的影子,例如类声明只有成员变量而不会包含方法实现等,支持全局函数等,所以有时看到aaa.bbb()还是有点迷糊,不知道aaa是包名还是实例名。
5.2 如何学习一门语言
当我们谈到学习英语时,想到的可能是背单词、学语法、练习听说读写。对于编程语言来说,背单词(关键字)、学语法(语法规则)少不了,可听说读写只剩下了“写”,因为我们说话的对象是“冷冰冰”的计算机。所以唯一的捷径就是“写”,不断地练习!
此外,学的语言多了也能总结出一些规律。首先是基础语法,包括了变量和常量、控制语句、函数、集合、OOP、异常处理、控制台输入输出、包管理等。然后是高级特性就差别比较大了。专注高并发的语言就要看并发方面的特性,专注OOP的语言就要看有哪些抽象层次更高的特性等等。还是那句话,基础语言只能说我们会用,而能够区别一门语言的高级特性才是它的根本和灵魂,也是我们要着重学习和领悟的地方。