网络爬虫爬取全国省市区（动态ip代理的获取，实现对ip限制的突破）

项目中用到的包结构

项目使用Jsoup进行网络的链接与网页的解析，使用dbutils进行dao操作，使用c3p0进行链接的管理

源代码下载地址：http://download.csdn.net/detail/chen1chen2chen3/9598202点击打开链接

爬虫程序的入口：

[java]  view plain  copy

1. package com.crawlercity.main;  
2.   
3.   
4. import org.jsoup.nodes.Document;  
5.   
6. import com.crawlercity.util.HttpUtils;  
7. import com.crawlercity.util.JsoupUtils;  
8.   
9. public class Main {  
10.     public static void main(String[] args) {  
11.         String url = "http://www.stats.gov.cn/tjsj/tjbz/tjyqhdmhcxhfdm/2015/index.html";// 初始解析网页地址  
12.         // 设置代理ip  
13.         HttpUtils.setProxyIp();  
14.         Document document = JsoupUtils.getDocument(url);// 得到的document一定是正常 的document  
15.         JsoupUtils.analysisDocument(document);  
16.           
17.     }  
18. }  

用于动态ip代理的工具类HttpUtils

[java]  view plain  copy

1. package com.crawlercity.util;  
2.   
3.   
4. import java.io.BufferedReader;  
5. import java.io.IOException;  
6. import java.io.InputStreamReader;  
7. import java.util.ArrayList;  
8. import java.util.List;  
9. import java.util.Random;  
10.   
11. public class HttpUtils {  
12.     /** 
13.      * 设置代理ip 
14.      * @throws IOException 
15.      */  
16.     public static void setProxyIp() {  
17.         try {  
18.             List<String> ipList = new ArrayList<>();  
19.             BufferedReader proxyIpReader = new BufferedReader(new InputStreamReader(HttpUtils.class.getResourceAsStream("/proxyip.txt")));  
20.               
21.             String ip = "";  
22.             while((ip = proxyIpReader.readLine()) != null) {  
23.                 ipList.add(ip);  
24.             }  
25.               
26.             Random random = new Random();  
27.             int randomInt = random.nextInt(ipList.size());  
28.             String ipport = ipList.get(randomInt);  
29.             String proxyIp = ipport.substring(0, ipport.lastIndexOf(":"));  
30.             String proxyPort = ipport.substring(ipport.lastIndexOf(":") + 1, ipport.length());  
31.               
32.             System.setProperty("http.maxRedirects", "50");    
33.             System.getProperties().setProperty("proxySet", "true");     
34.             System.getProperties().setProperty("http.proxyHost", proxyIp);    
35.             System.getProperties().setProperty("http.proxyPort", proxyPort);  
36.               
37.             System.out.println("设置代理ip为：" + proxyIp + "端口号为：" + proxyPort);  
38.         } catch (Exception e) {  
39.             System.out.println("重新设置代理ip");  
40.             setProxyIp();  
41.         }  
42.             
43.           
44.     }  
45. }  

用于获取document对象的工具类JsoupUtils

[html]  view plain  copy

1. public static Document getDocument(String url) {  
2.             try {  
3.                 Document document = Jsoup.connect(url).timeout(70).get();  
4.                   
5.                 if(document == null || document.toString().trim().equals("")) {// 表示ip被拦截或者其他情况  
6.                     System.out.println("出现ip被拦截或者其他情况");  
7.                     HttpUtils.setProxyIp();  
8.                     getDocument(url);  
9.                 }  
10.                   
11.                 return document;  
12.             } catch (Exception e) { // 链接超时等其他情况  
13.                 System.out.println("出现链接超时等其他情况");  
14.                 HttpUtils.setProxyIp();// 换代理ip  
15.                 getDocument(url);// 继续爬取网页  
16.             }  
17.             return getDocument(url);  
18.         }  

用于解析html文档的工具类 JsoupUtils

[java]  view plain  copy

1. public static void analysisDocument(Document document) {  
2.         try {  
3.             String baseUrl = "http://www.stats.gov.cn/tjsj/tjbz/tjyqhdmhcxhfdm/2015/";  
4.             CityInfo cityInfo1 = new CityInfo();  
5.             CityInfo cityInfo2 = new CityInfo();  
6.             CityInfo cityInfo3 = new CityInfo();  
7.             CityInfo cityInfo4 = new CityInfo();  
8.             CityInfo cityInfo5 = new CityInfo();  
9.             // 将类型为1（省）的数据分析并且插入  
10.             Elements elements1 = document.getElementsByAttributeValue("class", "provincetr");  
11.             for(Element element1 : elements1) {  
12.                 Elements elements2 = element1.getElementsByTag("a");  
13.                 for(Element element2 : elements2) {  
14.                     cityInfo1.setName(element2.text());  
15.                     cityInfo1.setParentId(0);  
16.                     cityInfo1.setType(1);  
17.                     cityInfo1.setUrl(baseUrl + element2.attr("href"));  
18. //                  System.out.println("cityInfo1" + cityInfo1.toString());  
19.                     int key1 = DBUtils.insertCityInfo(cityInfo1);  
20.                     Document document2 = getDocument(cityInfo1.getUrl());  
21.                     Elements elements3 = document2.getElementsByAttributeValue("class", "citytr");  
22.                     for(Element element3 : elements3) {  
23.                         Elements elements4 = element3.getElementsByTag("a");  
24.                         if(elements4.toString().trim().equals("")) {  
25.                             Elements diffElements = element3.getElementsByTag("td");  
26.                             cityInfo2.setCode(diffElements.get(0).text());  
27.                             cityInfo2.setName(diffElements.get(1).text());  
28.                             cityInfo2.setParentId(key1);  
29.                             cityInfo2.setType(2);  
30.                             continue;  
31.                         }  
32.                         cityInfo2.setCode(elements4.get(0).text());  
33.                         cityInfo2.setName(elements4.get(1).text());  
34.                         cityInfo2.setUrl(baseUrl + elements4.get(1).attr("href"));  
35.                         cityInfo2.setParentId(key1);  
36.                         cityInfo2.setType(2);  
37.                         /*System.out.println("cityInfo2" + cityInfo2.toString());*/  
38.                         int key2 = DBUtils.insertCityInfo(cityInfo2);  
39.                         Document document3 = getDocument(cityInfo2.getUrl());  
40.                         Elements elements5 = document3.getElementsByAttributeValue("class", "countytr");  
41.                         for(Element element5 : elements5) {  
42.                             Elements elements6 = element5.getElementsByTag("a");  
43.                             if(elements6.toString().trim().equals("")) {  
44.                                 Elements diffElements = element5.getElementsByTag("td");  
45.                                 cityInfo3.setCode(diffElements.get(0).text());  
46.                                 cityInfo3.setName(diffElements.get(1).text());  
47.                                 cityInfo3.setParentId(key2);  
48.                                 cityInfo3.setType(3);  
49.                                 continue;  
50.                             }  
51.                             cityInfo3.setCode(elements6.get(0).text());  
52.                             cityInfo3.setName(elements6.get(1).text());  
53.                             String cityInfo2Url = cityInfo2.getUrl();  
54.                             cityInfo3.setUrl(cityInfo2Url.substring(0, cityInfo2Url.lastIndexOf("/") + 1) + elements6.get(1).attr("href"));  
55.                             cityInfo3.setParentId(key2);  
56.                             cityInfo3.setType(3);  
57.                     /*      System.out.println("cityInfo3" + cityInfo3.toString());*/  
58.                             int key3 = DBUtils.insertCityInfo(cityInfo3);  
59.                             Document document4 = getDocument(cityInfo3.getUrl());  
60.                             Elements elements7 = document4.getElementsByAttributeValue("class", "towntr");  
61.                             for(Element element7 : elements7) {  
62.                                 Elements elements8 = element7.getElementsByTag("a");  
63.                                 System.out.println(elements8.toString());  
64.                                 if(elements8.toString().trim().equals("")) {// 表示没有a标签  
65.                                     Elements diffElements = element7.getElementsByTag("td");  
66.                                     cityInfo4.setCode(diffElements.get(0).text());  
67.                                     cityInfo4.setName(diffElements.get(1).text());  
68.                                     cityInfo4.setParentId(key3);  
69.                                     cityInfo4.setType(4);  
70.                                     continue;  
71.                                 }  
72.                                 cityInfo4.setCode(elements8.get(0).text());  
73.                                 cityInfo4.setName(elements8.get(1).text());  
74.                                 String cityInfo3Url = cityInfo3.getUrl();  
75.                                 cityInfo4.setUrl(cityInfo3Url.substring(0, cityInfo3Url.lastIndexOf("/") + 1) + elements8.get(1).attr("href"));  
76.                                 cityInfo4.setParentId(key3);  
77.                                 cityInfo4.setType(4);  
78. //                              System.out.println("cityInfo4" + cityInfo4.toString());  
79.                                 int key4 = DBUtils.insertCityInfo(cityInfo4);  
80.                                 Document document5 = getDocument(cityInfo4.getUrl());  
81.                                 Elements elements9 = document5.getElementsByAttributeValue("class", "villagetr");  
82.                                 for(Element element8 : elements9) {  
83.                                     Elements elements10 = element8.getElementsByTag("td");  
84.                                     cityInfo5.setCode(elements10.get(0).text());  
85.                                     cityInfo5.setName(elements10.get(2).text());  
86.                                     cityInfo5.setParentId(key4);  
87.                                     cityInfo5.setType(5);  
88.                                     /*System.out.println("cityInfo5" + cityInfo5.toString());*/  
89.                                     DBUtils.insertCityInfo(cityInfo5);  
90.                                 }  
91.                             }  
92.                         }  
93.                     }  
94.                 }  
95.             }  
96.               
97.         } catch (Exception e) {  
98.             e.printStackTrace();  
99.         }  

用于dao操作的工具类DbUtils

[html]  view plain  copy

1. package com.crawlercity.util;  
2.   
3. import java.sql.Connection;  
4. import java.sql.ResultSet;  
5. import java.sql.SQLException;  
6. import java.sql.Statement;  
7.   
8. import javax.sql.DataSource;  
9.   
10. import org.apache.commons.dbutils.QueryRunner;  
11. import org.apache.commons.dbutils.handlers.ScalarHandler;  
12.   
13. import com.crawlercity.model.CityInfo;  
14. import com.mchange.v2.c3p0.ComboPooledDataSource;  
15.   
16. public class DBUtils {  
17.   
18.     private static DataSource ds = null;  
19.   
20.     public static Connection getConnection() {  
21.         if(ds == null) {  
22.             ds = new ComboPooledDataSource();  
23.         }  
24.         try {  
25.             return ds.getConnection();  
26.         } catch (SQLException e) {  
27.             e.printStackTrace();  
28.         }  
29.         return null;  
30.     }  
31.   
32.     public static DataSource getDataSource() {  
33.         return ds == null ? new ComboPooledDataSource() : ds;  
34.     }  
35.       
36.   
37.     public static void releaseSource(Connection conn, Statement st, ResultSet rs) {  
38.         try {  
39.             if(rs != null && !rs.isClosed()) {  
40.                 rs.close();  
41.             }  
42.             if(st != null && !st.isClosed()) {  
43.                 st.close();  
44.             }  
45.             if(conn != null && !conn.isClosed()) {  
46.                 conn.close();  
47.             }  
48.         } catch (Exception e) {  
49.             e.printStackTrace();  
50.         }  
51.     }  
52.       
53.     public static int insertCityInfo(CityInfo cityInfo) {  
54.           
55.         Connection connection = DBUtils.getConnection();  
56.         QueryRunner qr = new QueryRunner();  
57.         String sql1 = "insert into cityinfo values (?,?,?,?,?,?)";  
58.         // 返回主键  
59.         String sql2 = "SELECT LAST_INSERT_ID()";  
60.           
61.         try {  
62.             int result = qr.update(connection, sql1, null, cityInfo.getParentId(), cityInfo.getType(), cityInfo.getName(), cityInfo.getCode(), cityInfo.getUrl());  
63.             int key = Integer.parseInt(qr.query(connection, sql2, new ScalarHandler<>()).toString());  
64.             releaseSource(connection, null, null);  
65.             return key;  
66.         } catch (SQLException e) {  
67.             e.printStackTrace();  
68.         }  
69.         return 0;  
70.     }  
71. }  

写代码的过程中出现了一些问题如： Jsoup如何在设置编码的同时设置连接超时，如何在超时或者动态ip代理无效的时候重新获取动态ip代理，如何在解析html失败后继续解析等。

通过这次编程发现自己在java网络方面的只是还是有待提高，以后继续努力！