View Code
packagecom.sreach.spider;
importjava.io.BufferedWriter;
importjava.io.File;
importjava.io.FileOutputStream;
importjava.io.IOException;
importjava.io.OutputStreamWriter;
importjava.util.regex.Matcher;
importjava.util.regex.Pattern;
publicclassFunctionUtils
{
/**
* 匹配超链接的正则表达式
*/
privatestaticString pat ="http://www\\.oschina\\.net/code/explore/.*/\\w+\\.[a-zA-Z]+";
privatestaticPattern pattern = Pattern.compile(pat);
privatestaticBufferedWriter writer =null;
/**
* 爬虫搜索深度
*/
publicstaticintdepth =0;
/**
* 以"/"来分割URL,获得超链接的元素
*
* @param url
* @return
*/
publicstaticString[] divUrl(String url)
{
returnurl.split("/");
}
/**
* 判断是否创建文件
*
* @param url
* @return
*/
publicstaticbooleanisCreateFile(String url)
{
Matcher matcher = pattern.matcher(url);
returnmatcher.matches();
}
/**
* 创建对应文件
*
* @param content
* @param urlPath
*/
publicstaticvoidcreateFile(String content, String urlPath)
{
/* 分割url */
String[] elems = divUrl(urlPath);
StringBuffer path =newStringBuffer();
File file =null;
for(inti =1; i
{
if(i != elems.length -1)
{
path.append(elems[i]);
path.append(File.separator);
file =newFile("D:"+ File.separator + path.toString());
}
if(i == elems.length -1)
{
Pattern pattern = Pattern.compile("\\w+\\.[a-zA-Z]+");
Matcher matcher = pattern.matcher(elems[i]);
if((matcher.matches()))
{
if(!file.exists())
{
file.mkdirs();
}
String[] fileName = elems[i].split("\\.");
file =newFile("D:"+ File.separator + path.toString()
+ File.separator + fileName[0] +".txt");
try
{
file.createNewFile();
writer =newBufferedWriter(newOutputStreamWriter(
newFileOutputStream(file)));
writer.write(content);
writer.flush();
writer.close();
System.out.println("创建文件成功");
}catch(IOException e)
{
e.printStackTrace();
}
}
}
}
}
/**
* 获取页面的超链接并将其转换为正式的A标签
*
* @param href
* @return
*/
publicstaticString getHrefOfInOut(String href)
{
/* 内外部链接最终转化为完整的链接格式 */
String resultHref =null;
/* 判断是否为外部链接 */
if(href.startsWith("http://"))
{
resultHref = href;
}else
{
/* 如果是内部链接,则补充完整的链接地址,其他的格式忽略不处理,如:a href="#" */
if(href.startsWith("/"))
{
resultHref ="http://www.oschina.net"+ href;
}
}
returnresultHref;
}
/**
* 截取网页网页源文件的目标内容
*
* @param content
* @return
*/
publicstaticString getGoalContent(String content)
{
intsign = content.indexOf("
String signContent = content.substring(sign);
intstart = signContent.indexOf(">");
intend = signContent.indexOf("
");returnsignContent.substring(start +1, end);
}
/**
* 检查网页源文件中是否有目标文件
*
* @param content
* @return
*/
publicstaticintisHasGoalContent(String content)
{
returncontent.indexOf("
}
}
HrefOfPage.java 此类为获取页面的超链接View Code
packagecom.sreach.spider;
publicclassHrefOfPage
{
/**
* 获得页面源代码中超链接
*/
publicstaticvoidgetHrefOfContent(String content)
{
System.out.println("开始");
String[] contents = content.split("
for(inti =1; i
{
intendHref = contents[i].indexOf("\"");
String aHref = FunctionUtils.getHrefOfInOut(contents[i].substring(
, endHref));
if(aHref !=null)
{
String href = FunctionUtils.getHrefOfInOut(aHref);
if(!UrlQueue.isContains(href)
&& href.indexOf("/code/explore") != -1
&& !VisitedUrlQueue.isContains(href))
{
UrlQueue.addElem(href);
}
}
}
System.out.println(UrlQueue.size() +"--抓取到的连接数");
System.out.println(VisitedUrlQueue.size() +"--已处理的页面数");
}
}
UrlDataHanding.java 此类主要是从未访问队列中获取url,下载页面,分析url,保存已访问url等操作,实现Runnable接口View Code
packagecom.sreach.spider;
publicclassUrlDataHandingimplementsRunnable
{
/**
* 下载对应页面并分析出页面对应的URL放在未访问队列中。
* @param url
*/
publicvoiddataHanding(String url)
{
HrefOfPage.getHrefOfContent(DownloadPage.getContentFormUrl(url));
}
publicvoidrun()
{
while(!UrlQueue.isEmpty())
{
dataHanding(UrlQueue.outElem());
}
}
}
UrlQueue.java 此类主要是用来存放未访问的URL队列View Code
packagecom.sreach.spider;
importjava.util.LinkedList;
publicclassUrlQueue
{
/**超链接队列*/
publicstaticLinkedList urlQueue =newLinkedList();
/**队列中对应最多的超链接数量*/
publicstaticfinalintMAX_SIZE =10000;
publicsynchronizedstaticvoidaddElem(String url)
{
urlQueue.add(url);
}
publicsynchronizedstaticString outElem()
{
returnurlQueue.removeFirst();
}
publicsynchronizedstaticbooleanisEmpty()
{
returnurlQueue.isEmpty();
}
publicstaticintsize()
{
returnurlQueue.size();
}
publicstaticbooleanisContains(String url)
{
returnurlQueue.contains(url);
}
}
VisitedUrlQueue.java 主要是保存已访问过的URL,使用HashSet来保存,主要是考虑到每个访问过的URL是不同。HashSet刚好符合这个要求View Code
packagecom.sreach.spider;
importjava.util.HashSet;
/**
* 已访问url队列
* @author HHZ
*
*/
publicclassVisitedUrlQueue
{
publicstaticHashSet visitedUrlQueue =newHashSet();
publicsynchronizedstaticvoidaddElem(String url)
{
visitedUrlQueue.add(url);
}
publicsynchronizedstaticbooleanisContains(String url)
{
returnvisitedUrlQueue.contains(url);
}
publicsynchronizedstaticintsize()
{
returnvisitedUrlQueue.size();
}
}
Test.java 此类为测试类View Code
importjava.sql.SQLException;
importcom.sreach.spider.UrlDataHanding;
importcom.sreach.spider.UrlQueue;
publicclassTest
{
publicstaticvoidmain(String[] args)throwsSQLException
{
String url ="http://www.oschina.net/code/explore/achartengine/client/AndroidManifest.xml";
String url1 ="http://www.oschina.net/code/explore";
String url2 ="http://www.oschina.net/code/explore/achartengine";
String url3 ="http://www.oschina.net/code/explore/achartengine/client";
UrlQueue.addElem(url);
UrlQueue.addElem(url1);
UrlQueue.addElem(url2);
UrlQueue.addElem(url3);
UrlDataHanding[] url_Handings =newUrlDataHanding[10];
for(inti =0; i <10; i++)
{
url_Handings[i] =newUrlDataHanding();
newThread(url_Handings[i]).start();
}
}
}
说明一下:由于我抓取的是针对oschina的,所以里面的url正则表达式不适合其他网站,需要自己修改一下。你也可以写成xml来配置。