HttpClient实现HTTP文件通用下载类

时间:2023-03-09 05:51:50
HttpClient实现HTTP文件通用下载类
 import java.io.File;
import java.io.FileOutputStream;
import java.io.InputStream; import org.apache.http.Header;
import org.apache.http.HeaderElement;
import org.apache.http.HttpEntity;
import org.apache.http.HttpResponse;
import org.apache.http.NameValuePair;
import org.apache.http.client.HttpClient;
import org.apache.http.client.methods.HttpGet;
import org.apache.http.impl.client.DefaultHttpClient; /**
* 说明
* 利用httpclient下载文件
* maven依赖
* <dependency>
* <groupId>org.apache.httpcomponents</groupId>
* <artifactId>httpclient</artifactId>
* <version>4.0.1</version>
* </dependency>
* 可下载http文件、图片、压缩文件
* bug:获取response header中Content-Disposition中filename中文乱码问题
* @author tanjundong
*
*/
public class HttpDownload { public static final int cache = 10 * 1024;
public static final boolean isWindows;
public static final String splash;
public static final String root;
static {
if (System.getProperty("os.name") != null && System.getProperty("os.name").toLowerCase().contains("windows")) {
isWindows = true;
splash = "\\";
root="D:";
} else {
isWindows = false;
splash = "/";
root="/search";
}
} /**
* 根据url下载文件,文件名从response header头中获取
* @param url
* @return
*/
public static String download(String url) {
return download(url, null);
} /**
* 根据url下载文件,保存到filepath中
* @param url
* @param filepath
* @return
*/
public static String download(String url, String filepath) {
try {
HttpClient client = new DefaultHttpClient();
HttpGet httpget = new HttpGet(url);
HttpResponse response = client.execute(httpget); HttpEntity entity = response.getEntity();
InputStream is = entity.getContent();
if (filepath == null)
filepath = getFilePath(response);
File file = new File(filepath);
file.getParentFile().mkdirs();
FileOutputStream fileout = new FileOutputStream(file);
/**
* 根据实际运行效果 设置缓冲区大小
*/
byte[] buffer=new byte[cache];
int ch = 0;
while ((ch = is.read(buffer)) != -1) {
fileout.write(buffer,0,ch);
}
is.close();
fileout.flush();
fileout.close(); } catch (Exception e) {
e.printStackTrace();
}
return null;
}
/**
* 获取response要下载的文件的默认路径
* @param response
* @return
*/
public static String getFilePath(HttpResponse response) {
String filepath = root + splash;
String filename = getFileName(response); if (filename != null) {
filepath += filename;
} else {
filepath += getRandomFileName();
}
return filepath;
}
/**
* 获取response header中Content-Disposition中的filename值
* @param response
* @return
*/
public static String getFileName(HttpResponse response) {
Header contentHeader = response.getFirstHeader("Content-Disposition");
String filename = null;
if (contentHeader != null) {
HeaderElement[] values = contentHeader.getElements();
if (values.length == 1) {
NameValuePair param = values[0].getParameterByName("filename");
if (param != null) {
try {
//filename = new String(param.getValue().toString().getBytes(), "utf-8");
//filename=URLDecoder.decode(param.getValue(),"utf-8");
filename = param.getValue();
} catch (Exception e) {
e.printStackTrace();
}
}
}
}
return filename;
}
/**
* 获取随机文件名
* @return
*/
public static String getRandomFileName() {
return String.valueOf(System.currentTimeMillis());
}
public static void outHeaders(HttpResponse response) {
Header[] headers = response.getAllHeaders();
for (int i = 0; i < headers.length; i++) {
System.out.println(headers[i]);
}
}
public static void main(String[] args) {
// String url = "http://bbs.btwuji.com/job.php?action=download&pid=tpc&tid=320678&aid=216617";
String url="http://www.dy1000.com/img/20120701/1999311085_150_200.JPG";
// String filepath = "D:\\test\\a.torrent";
String filepath = "D:\\test\\a.jpg";
HttpDownload.download(url, filepath);
}
}

参考资料: 
http://eyecm.com/httpclient4-1-tutorial-official-examples-to-explain-httpclient4-1-usage/ 
http://eyecm.com/httpclient-download/