java 9 多线程 HTTP 客户端无法正常工作

Question

我正在尝试将以下顺序代码转换为多线程代码，但结果对我来说听起来不合理。

package com.net;

import jdk.incubator.http.HttpClient;
import jdk.incubator.http.HttpRequest;
import jdk.incubator.http.HttpResponse;
import org.jsoup.Jsoup;
import org.jsoup.nodes.Document;

import java.io.IOException;
import java.net.URI;
import java.net.URISyntaxException;

public class Req {

    HttpClient client = HttpClient.newHttpClient();

    private String getResource(String someUrl) {
        String body = "";
        try {
            URI url = new URI(someUrl);
            HttpRequest request = HttpRequest.newBuilder()
                    .uri(url).GET().build();
            HttpResponse<String> response = client.send(request, HttpResponse.BodyHandler.asString());
            body = response.body();
        } catch (URISyntaxException e) {
            System.out.println("URL " + someUrl + "is not valid");
        } catch (IOException | InterruptedException e) {
            System.out.println(e.getMessage());
        }
        return body;
    }


    public static void main(String[] args){

        String[] topIranianSites = {
                "https://www.aparat.com/",
                "http://www.varzesh3.com/",
                "http://namnak.com/",
                "http://www.telewebion.com/",
                "https://divar.ir/",
                "https://www.ninisite.com/",
                "https://www.blogfa.com/",
                "http://www.namasha.com/",
                "http://www.yjc.ir/"
        };

        Req singleThreadReq = new Req();
        float totalElapsedTime = 0F;

        for (String site : topIranianSites){
            long fetchStartTime = System.currentTimeMillis();
            String html = singleThreadReq.getResource(site);
            float elapsed = (float) (System.currentTimeMillis() - fetchStartTime) / 1000;

            Document doc = Jsoup.parse(html);
            System.out.println("It took " + elapsed + " seconds to fetch " + site + " with title " + doc.title());
            totalElapsedTime += elapsed;
        }

        System.out.println("Total Elapsed Time: " + totalElapsedTime + "\nTotal Number of sites: " + topIranianSites.length);

    }
}

这是输出

WARNING: Using incubator modules: jdk.incubator.httpclient
It took 2.622 seconds to fetch https://www.aparat.com/ with title آپارات - سرویس اشتراک ویدیو
It took 0.455 seconds to fetch http://www.varzesh3.com/ with title 
It took 0.521 seconds to fetch http://namnak.com/ with title نمناک
It took 2.172 seconds to fetch http://www.telewebion.com/ with title تلوبیون | مرجع پخش زنده و دانلود فیلم ، سریال و سایر برنامه های تلویزیون
General SSLEngine problem
It took 0.229 seconds to fetch https://divar.ir/ with title 
It took 1.769 seconds to fetch https://www.ninisite.com/ with title نی نی سایت | راهنمای بارداری و بچه داری
Received fatal alert: handshake_failure
It took 0.382 seconds to fetch https://www.blogfa.com/ with title 
It took 2.641 seconds to fetch http://www.namasha.com/ with title نماشا - سرویس رایگان اشتراک ویدیو
It took 0.503 seconds to fetch http://www.yjc.ir/ with title 
Total Elapsed Time: 11.294001
Total Number of sites: 9

从顺序输出中，我猜想正确的多线程代码应该需要大约 2.8 秒才能获取所有 9 个站点。但是我的多线程代码实现需要更多

package com.net;

import jdk.incubator.http.HttpClient;
import jdk.incubator.http.HttpRequest;
import jdk.incubator.http.HttpResponse;
import org.jsoup.Jsoup;
import org.jsoup.nodes.Document;

import java.io.IOException;
import java.net.URI;
import java.net.URISyntaxException;

public class ReqThreaded implements Runnable {

    class Site {
        String url;
        String title;
        float fetchTime;
    }

    private HttpClient client = HttpClient.newHttpClient();

    private Thread[] threadPool;
    private String[] rawSites;
    private Site[] sitesArr;
    private int sitesDone = 0;
    long startTime = System.currentTimeMillis();
    float totalElapsed = 0F;

    public ReqThreaded(String[] sites) {
        threadPool = new Thread[sites.length];
        sitesArr = new Site[sites.length];
        rawSites = sites;

        for (int i = 0; i < sites.length; i++) {
            startThread(i);
        }

        while (sitesDone < sites.length) {
            try {
                Thread.sleep(1000);
                totalElapsed = (float) (System.currentTimeMillis() - startTime) / 1000;
                System.out.print("\rElapsed time: " + totalElapsed + "Sites Done: " + sitesDone);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }

        System.out.println("\n\nStatistics:\n\n");

        for (Site someSite : sitesArr) {
            System.out.println("URL " + someSite.url + "\nTitle: " + someSite.title + "\nFetch Time: " + someSite.fetchTime + "\n\n");
        }

    }

    private void startThread(int i) {
        if (threadPool[i] == null) {
            threadPool[i] = new Thread(this);
            threadPool[i].start();
        }
    }


    private String getResource(String someUrl) {
        String body = "";
        try {
            URI url = new URI(someUrl);
            HttpRequest request = HttpRequest.newBuilder()
                    .uri(url).GET().build();
            HttpResponse<String> response = client.send(request, HttpResponse.BodyHandler.asString());
            body = response.body();
        } catch (URISyntaxException e) {
            System.out.println("URL " + someUrl + "is not valid");
        } catch (IOException | InterruptedException e) {
            System.out.println(e.getMessage());
        }
        return body;
    }


    @Override
    public void run() {
        Thread thisThread = Thread.currentThread();
        int sitesIndex = 0;

        for (int j = 0; j < threadPool.length; j++) {
            if (thisThread == threadPool[j]) {
                sitesIndex = j;
            }
        }
        long fetchStartTime = System.currentTimeMillis();
        String html = getResource(rawSites[sitesIndex]);
        float elapsed = (float) (System.currentTimeMillis() - fetchStartTime) / 1000;
        sitesDone++;
        Document doc = Jsoup.parse(html);
        sitesArr[sitesIndex] = new Site();
        sitesArr[sitesIndex].url = rawSites[sitesIndex];
        sitesArr[sitesIndex].title = doc.title();
        sitesArr[sitesIndex].fetchTime = elapsed;
    }

    public static void main(String[] args) {

        String[] topIranianSites = {
                "https://www.aparat.com/",
                "http://www.varzesh3.com/",
                "http://namnak.com/",
                "http://www.telewebion.com/",
                "https://divar.ir/",
                "https://www.ninisite.com/",
                "https://www.blogfa.com/",
                "http://www.namasha.com/",
                "http://www.yjc.ir/"
        };

        new ReqThreaded(topIranianSites);

    }
}

这是多线程代码的输出。每个 url 的总时间和获取时间似乎都不正确。我认为有些东西在这里阻塞或某种竞争条件。这里有什么问题？

WARNING: Using incubator modules: jdk.incubator.httpclient
General SSLEngine problem
Received fatal alert: handshake_failure
Elapsed time: 7.068Sites Done: 9

Statistics:


URL https://www.aparat.com/
Title: آپارات - سرویس اشتراک ویدیو
Fetch Time: 4.808


URL http://www.varzesh3.com/
Title: 
Fetch Time: 5.904


URL http://namnak.com/
Title: نمناک
Fetch Time: 1.056


URL http://www.telewebion.com/
Title: تلوبیون | مرجع پخش زنده و دانلود فیلم ، سریال و سایر برنامه های تلویزیون
Fetch Time: 6.569


URL https://divar.ir/
Title: 
Fetch Time: 0.53


URL https://www.ninisite.com/
Title: نی نی سایت | راهنمای بارداری و بچه داری
Fetch Time: 4.287


URL https://www.blogfa.com/
Title: 
Fetch Time: 0.767


URL http://www.namasha.com/
Title: نماشا - سرویس رایگان اشتراک ویدیو
Fetch Time: 4.539


URL http://www.yjc.ir/
Title: 
Fetch Time: 0.836

Answer 1

我修复了类 ReqThreaded 中的一些明显错误，例如不正确的同步，将方法 run() 移至类 Site，并使结果打印与单线程变体中的结果相同。未使用线程池，为每个请求创建了一个单独的线程。

结果如下：

Received fatal alert: handshake_failure
It took 0.263 seconds to fetch https://www.blogfa.com/ with title 
General SSLEngine problem
It took 0.491 seconds to fetch https://divar.ir/ with title 
It took 1.02 seconds to fetch http://www.yjc.ir/ with title 
It took 1.056 seconds to fetch http://www.telewebion.com/ with title تلوبیون | مرجع پخش زنده و دانلود فیلم ، سریال و سایر برنامه های تلویزیون
It took 1.262 seconds to fetch https://www.ninisite.com/ with title نی نی سایت | راهنمای بارداری و بچه داری
It took 1.411 seconds to fetch http://namnak.com/ with title نمناک
It took 1.608 seconds to fetch http://www.varzesh3.com/ with title ورزش سه :: صفحه اصلی
It took 2.221 seconds to fetch http://www.namasha.com/ with title نماشا - سرویس رایگان اشتراک ویدیو
It took 2.247 seconds to fetch https://www.aparat.com/ with title آپارات - سرویس اشتراک ویدیو
Elapsed time: 2.253
Sites Done: 9
Process finished with exit code 0

也就是说，总时间仅比从网站获得结果的最长时间稍长。

多线程工作！

修改后的代码如下：

import jdk.incubator.http.HttpClient;
import jdk.incubator.http.HttpRequest;
import jdk.incubator.http.HttpResponse;
import org.jsoup.Jsoup;
import org.jsoup.nodes.Document;

import java.io.IOException;
import java.net.URI;
import java.net.URISyntaxException;
import java.util.concurrent.CountDownLatch;

public class ReqThreaded {

class Site implements Runnable {
    final String url;
    String title;
    float fetchTime;

    Site(String url) {
        this.url = url;
    }

    @Override
    public void run() {
        long fetchStartTime = System.currentTimeMillis();
        String html = getResource(url);
        float elapsed = (float) (System.currentTimeMillis() - fetchStartTime) / 1000;
        Document doc = Jsoup.parse(html);
        title = doc.title();
        fetchTime = elapsed;
        System.out.println("It took " + fetchTime + " seconds to fetch " + url + " with title " + title);
        sitesDone.countDown();
    }
}

private HttpClient client = HttpClient.newHttpClient();

private CountDownLatch sitesDone;

public ReqThreaded(String[] sites) throws InterruptedException {
    int siteNumber = sites.length;
    sitesDone = new CountDownLatch(siteNumber);
    long startTime = System.currentTimeMillis();

    for (int i = 0; i < siteNumber; i++) {
        Runnable site = new Site(sites[i]);
        Thread thread = new Thread(site);
        thread.start();
    }

    sitesDone.await();
    float totalElapsed = (float) (System.currentTimeMillis() - startTime) / 1000;
    System.out.print("\rElapsed time: " + totalElapsed + "\nSites Done: " + siteNumber);
}

private String getResource(String someUrl) {
    String body = "";
    try {
        URI url = new URI(someUrl);
        HttpRequest request = HttpRequest.newBuilder()
                .uri(url).GET().build();
        HttpResponse<String> response = client.send(request, HttpResponse.BodyHandler.asString());
        body = response.body();
    } catch (URISyntaxException e) {
        System.out.println("URL " + someUrl + "is not valid");
    } catch (IOException | InterruptedException e) {
        System.out.println(e.getMessage());
    }
    return body;
}

public static void main(String[] args) throws InterruptedException {

    String[] topIranianSites = {
            "https://www.aparat.com/",
            "http://www.varzesh3.com/",
            "http://namnak.com/",
            "http://www.telewebion.com/",
            "https://divar.ir/",
            "https://www.ninisite.com/",
            "https://www.blogfa.com/",
            "http://www.namasha.com/",
            "http://www.yjc.ir/"
    };

    new ReqThreaded(topIranianSites);

}
}

即：

• 切勿在不同线程上同时对同一对象运行相同的方法

• 永远不要通过普通变量在线程之间交换信息。仅使用专门的设施。这里我使用CountDownLatch，表示每个线程的结束。如果我想向主线程返回一些信息，我会改用BlockingQueue。