轉載請註明出處
作者:曉渡
文章地址:https://greatestrabit.github.io/2016/03/29/callback/
1.字面意義上的回調
字面意思上理解回調,就是A調用B,B回過頭來再調用A,即是回調.既然是這樣,當然就要求A中有B,B中有A.如下:
class A { /** * 提出問題 * @author [email protected] * @param b * @param question */ public void ask(final B b, final String question) { b.answer(this, question); } /** * 處理結果 * @author [email protected] * @param answer */ public void proce***esult(final String answer) { System.out.println(answer); } } class B { /** * 計算結果 * @author [email protected] * @param a * @param question */ public void answer(final A a, final String question) { if (question.equals("What is the answer to life, the universe and everything?")) { a.proce***esult("42"); } } } /** * 相互調用 * @author [email protected] * */ public class SyncObjectCallback { public static void main(final String[] args) { B b = new B(); A a = new A(); a.ask(b, "What is the answer to life, the universe and everything?"); } }
2.面向對象的回調
上面的寫法中,B的對象只在方法中被傳遞了.實際上,這個B對象後來又調用了A中的方法,它的作用應該不止侷限在一個方法中,而應該是A的一個部分.也就是,上面的寫法不夠”面向對象”,讓我們來改造一下:
class A { private final B b; public A(final B b) { this.b = b; } public void ask(final String question) { this.b.answer(this, question); } public void proce***esult(final String answer) { System.out.println(answer); } } class B { public void answer(final A a, final String question) { if (question.equals("What is the answer to life, the universe and everything?")) { a.proce***esult("42"); } } } /** * 面向對象的相互調用 * @author [email protected] * */ public class SyncOOCallback { public static void main(final String[] args) { B b = new B(); A a = new A(b); a.ask("What is the answer to life, the universe and everything?"); } }
3.面向接口的回調
上面的兩個例子,估計沒人會承認也是回調吧.因爲並沒什麼卵用.不過這個流程對於理解回調是很重要的.其實回調真正有用的地方,在於它的”預測”能力.
我們擴展想象一下.假設上面例子中的B,爲A提供了很多服務之後突然覺醒,想爲更多的對象提供服務,這樣一來,B就變成了Server.而且還要制定規則.規則是什麼呢,就是要Server提供服務可以,對方一定要有一個recvAnswer接口供Server調用才行,這樣Server才能把結果傳回給Client.具體如何制定規則呢?通過Interface.如下:
/** * 發出請求着需要實現的接口,要實現處理結果的方法 * @author [email protected] * */ public interface IClient { void recvAnswer(String answer); } /** * 響應請求者,即提供服務者 * @author [email protected] * */ public class Server { public void answer(final IClient client, final String question) { if (question.equals("What is the answer to life, the universe and everything?")) { calclating(); client.recvAnswer("42"); } } private void calclating() { try { Thread.sleep(new Random().nextInt(5000)); } catch (InterruptedException e) { e.printStackTrace(); } } } /** * 發出請求者,同時要處理請求結果 * @author [email protected] * */ public class ClientSync implements IClient { private final Server server; public ClientSync(final Server server) { this.server = server; } public void ask(final String question) { this.server.answer(this, question); } @Override public void recvAnswer(final String answer) { System.out.println(answer); } } /** * 面向接口的同步回調 * @author [email protected] * */ public class SyncInterfaceCallback { /** * 使用內部類來實現的方式 * @author [email protected] */ private static void innerMain() { Server server = new Server(); server.answer(new IClient() { @Override public void recvAnswer(final String answer) { System.out.println(answer); } }, "What is the answer to life, the universe and everything?"); } public static void main(final String[] args) { Server server = new Server(); ClientSync client = new ClientSync(server); client.ask("What is the answer to life, the universe and everything?"); innerMain(); } }
注意,接口IClient實際上應該是屬於Server端的,它是由Server制定的,需要Client來實現的接口,雖然看上去它跟Client很近.
爲什麼說有”預測”能力呢?想象另一個場景.Server現在是一個底層服務,這個底層服務知道遲早有一天會有高層服務來討要數據,但是數據如何向上傳遞呢?底層可以承諾,只有你實現IClient接口,我就會調用其中的recvAnswer方法,把數據傳上來.現在底層也可以調用高層的方法,算是有”預測”能力吧?
4.異步回調
上面的調用都是同步的.假設Server計算結果需要較長的時間,你一定希望它能在一個單獨的線程中被執行,這是就可以把ask方法的調用用線程包裝一下:
public class ClientAsync implements IClient { private final Server server; public ClientAsync(final Server server) { this.server = server; } /** * 在線程中發出請求 * @author [email protected] * @param question */ public void ask(final String question) { new Thread(new Runnable() { @Override public void run() { server.answer(ClientAsync.this, question); } }).start(); } @Override public void recvAnswer(final String answer) { System.out.println(answer); } } /** * 基於接口的異步回調,每次建立新的線程 * @author [email protected] * */ public class AsyncInterfaceCallback { /** * 使用內部類的實現方式,此處可見回調地獄 * @author [email protected] */ private static void innerMain() { Server server = new Server(); new Thread(new Runnable() { @Override public void run() { server.answer(new IClient() { @Override public void recvAnswer(final String answer) { System.out.println(answer); } }, "What is the answer to life, the universe and everything?"); } }).start(); System.out.println("asked ! waiting for the answer..."); } public static void main(final String[] args) { Server server = new Server(); ClientAsync client = new ClientAsync(server); client.ask("What is the answer to life, the universe and everything?"); System.out.println("asked ! waiting for the answer..."); innerMain(); } }
5.線程池異步回調
每次建立新的線程耗費資源巨大,爲了重用線程,使用線程池管理異步調用,這時候就要求Client不僅要實現IClient接口,還要同時是一個任務,才能被線程池執行,如下:
/** * 專門用來執行請求的任務,供線程池調用 * @author [email protected] * */ public class ClientRunnable implements IClient, Runnable { private final Server server; private final String question; private final int id; public ClientRunnable(final Server server, final String question, final int id) { this.server = server; this.question = question; this.id = id; } @Override public void recvAnswer(final String answer) { System.out.println("clinet " + this.id + " got answer: " + answer); } @Override public void run() { server.answer(ClientRunnable.this, this.question); } } /** * 基於線程池的異步回調 * @author [email protected] * */ public class ThreadpoolCallback { public static void main(final String[] args) { ExecutorService es = Executors.newCachedThreadPool(); Server server = new Server(); for (int i = 0; i < 100; i++) { ClientRunnable cr = new ClientRunnable(server, "What is the answer to life, the universe and everything?", i); es.execute(cr); System.out.println("client " + i + " asked !"); } es.shutdown(); } }
至此,我們就實現了線程池異步回調.
完整源碼請移步:github