欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

WebMagic爬取网页信息

程序员文章站 2022-05-02 20:48:51
...
package cc.mrbird.febs.system.collection.cs;

import cc.mrbird.febs.common.utils.StringUtils;
import cc.mrbird.febs.system.collection.MysqlPipeline;
import cc.mrbird.febs.system.collection.rmw.RmwCollection;
import cc.mrbird.febs.system.entity.CjData;
import org.jsoup.Jsoup;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;
import us.codecraft.webmagic.Page;
import us.codecraft.webmagic.Site;
import us.codecraft.webmagic.Spider;
import us.codecraft.webmagic.processor.PageProcessor;
import us.codecraft.webmagic.scheduler.BloomFilterDuplicateRemover;
import us.codecraft.webmagic.scheduler.QueueScheduler;
import us.codecraft.webmagic.selector.Selectable;

import java.util.Date;
import java.util.List;

@Component
public class CsCollection implements PageProcessor {

    @Autowired
    private MysqlPipeline mysqlPipeline;
    // 抓取网站的相关配置,包括编码、抓取间隔、重试次数等
    private Site site = Site.me().setRetryTimes(3).setSleepTime(100);

    private String url = "https://www.casl.org.cn/policy/getByPolicy/zcfg";

    @Override
    public Site getSite() {

        return site;
    }

    @Override
    public void process(Page page) {
        //文章详情链接正则表达式/html/body/div[4]/div[2]/div[1]/div[1]/a/div[3]/h2
        String describeLink = "http://localhost:8080/policy/selectDetail/\\d{3}";
        //文章标题/html/body/div[4]/div[1]
        String titleXpath = "/html/body/div[4]/div[1]/text()";
        //文章来源/html/body/div[4]/div[2]/div[1]/span[2]
        String sourceXpath = "/html/body/div[4]/div[2]/div[1]/span[2]/text()";
        //发布时间
        String publishTime="/html/body/div[4]/div[2]/div[1]/span[1]/text()";

        String lbPublishTime ="/html/body/div[4]/div[2]/div[1]/div[1]/a/div[3]/p[2]/span";

        //判断是列表页还是详情页
        List<Selectable> listurl = page.getHtml().css("div.zcwj_con div.zcwj_list div.hui").nodes();
        if (listurl.size() == 0) {//详情页
            String sourceStr=page.getHtml().xpath(sourceXpath).regex(".*原创.*").toString();
            if(sourceStr!=null && !sourceStr.equals("")){
                CjData cjData=new CjData();
                cjData.setTitle(page.getHtml().xpath(titleXpath).toString());
                String publishStr=page.getHtml().xpath(publishTime).toString();
                cjData.setPublishTime(publishStr);
                cjData.setSource(page.getHtml().xpath(sourceXpath).toString());
                cjData.setContext(Jsoup.parse(page.getHtml().css("div#content1").toString()).text());
                cjData.setUrl(page.getUrl().toString());
                cjData.setCreateTime(new Date());
                page.putField("cjData",cjData);
            }
        } else {
            //如果是列表页
            //获取列表页所有文章的url
            for (Selectable selectable : listurl) {
                String strurl = selectable.links().toString();
                page.addTargetRequest(strurl);
            }
        }
    }

    public void process() {
        Spider.create(new CsCollection())
                .addUrl(url)
                .setScheduler(new QueueScheduler().setDuplicateRemover(new BloomFilterDuplicateRemover(1000000)))//url去重
                .addPipeline(this.mysqlPipeline)
                .thread(3)
                .run();
    }


}

教程点我

相关标签: webMagic java