Libcurl为一个免费开源的,客户端url传输库,支持FTP,FTPS,TFTP,HTTP,HTTPS,GOPHER,TELNET,DICT,FILE和LDAP,跨平台,支持Windows,Unix,Linux等,线程安全,支持Ipv6.
首先一个基本原则就是:绝对不应该在线程之间共享同一个libcurl handle(CURL *对象),不管是easy handle还是multi handle(本文只介绍easy_handle)。一个线程每次只能使用一个handle。
libcurl是线程安全的,但有两点例外:信号(signals)和SSL/TLS handler。 信号用于超时失效名字解析(timing out name resolves)。libcurl依赖其他的库来支持SSL/STL,所以用多线程的方式访问HTTPS或FTPS的URL时,应该满足这些库对多线程 操作的一些要求。
libcurl总体框架分为五部分:
1. 使用curl_global_init来初始化libcurl
2. 调用curl_easy_init()得到easy interface指针
3. 调用curl_easy_setopt 设置传输选项 (断点续传,超时设置,回调函数都在这里面设置)
4. 调用curl_easy_perform()完成传输任务
5. 调用curl_easy_cleanup()释放内存
6. 调用curl_global_cleanup();
这里打个比方:curl_easy_setopt类似于演员化妆,进行各种造型,而curl_easy_perform()类似上台表演,给演员们一个展示自己的舞台。
Downloader.h :
#ifndef __Downloader_LibCurl_H__
#define __Downloader_LibCurl_H__
#pragma once
#include "curl\curl.h"
#include <atlstr.h>
#define MAXWORK 200
typedef struct DownloadInfo
{
char url[512];
char filePath[256];
}DLIO;
typedef struct CurDownloadInfor
{
char url[512]; //url
char fileName[256]; //文件名称
long preLocalLen; //本地已下载的长度(大小)
double totalFileLen; //文件总长度(大小)
double CurDownloadLen; //每次下载的文件长度(大小)
}CURDI;
class CDownloader
{
public:
CDownloader(void);
~CDownloader(void);
int StartDownloadThread();
double GetTotalFileLenth(const char* url); //获取将要下载的文件长度
long GetLocalFileLenth(const char* fileName); //获取本地问价长度
void GetFileNameFormUrl(char* fileName, const char* url); //从URL中获取文件名
void AddDownloadWork(DLIO downloadWork);
int SetConnectTimeOut(DWORD nConnectTimeOut); //设置连接的超时时间
int GetCurrentDownloadInfo(CURDI* lpCurDownloadInfor);
BOOL CreateMultiDir(const char* pathName); //是否在本地创建目录,没有就创建
BOOL IsDownloadBegin();
BOOL IsDownloadEnd();
protected:
static DWORD WINAPI SingleDownloadProc(LPVOID lpParameter); //线程函数
static size_t WriteFunc(char *str, size_t size, size_t nmemb, void *stream); //写入数据(回调函数)
static size_t ProgressFunc(double* fileLen, double t, double d, double ultotal, double ulnow); //下载进度
private:
char m_filePath[512];
char m_downloadUrl[256];
int m_downloadCourse; //-1 还未下载 0正在下载 1下载完成
long m_curLocalFileLenth; //因为下载的时候已经计算了本地文件的大小用来设置断点,所以对于每个文件,该数字只会被设置一次;就是下载前的本地大小;
long m_nConnectTimeOut; //连接的超时时间
DLIO m_dowloadWork[MAXWORK];
CURDI m_curDownloadInfo;
int m_curIndex;
CURL* m_pCurl;
};
#endif
Downloader.cpp :
#include "StdAfx.h"
#include "Downloader.h"
#include <io.h>
CDownloader::CDownloader(void)
{
m_downloadCourse = -1;
m_nConnectTimeOut = 0;
curl_global_init (CURL_GLOBAL_ALL);
for(int i=0; i<MAXWORK; i++)
{
memset(m_dowloadWork->url, 0, 512);
memset(m_dowloadWork->filePath, 0, 256);
}
m_curIndex = 0;
}
CDownloader::~CDownloader(void)
{
curl_global_cleanup();
}
BOOL CDownloader::IsDownloadBegin()
{
if(m_downloadCourse == 0)
return TRUE;
return FALSE;
}
BOOL CDownloader::IsDownloadEnd()
{
if(m_downloadCourse == 1)
return TRUE;
return FALSE;
}
BOOL CDownloader::CreateMultiDir(const char* pathName)
{
if(pathName == NULL) return FALSE;
char filePath[256] = {0};
strcpy(filePath, pathName);
int i = 0, pathLen = strlen(pathName);
CString curPath;
char curFilePath[256] = {0};
WIN32_FIND_DATA swf;
if(filePath[pathLen - 1] != '\\') //最后一个非0字符不是‘\\’则加上
{
filePath[pathLen] = '\\';
}
while(filePath[i] != '\0')
{
if(filePath[i] == ':')
{
i+=2;
continue;
}
if(filePath[i] == '\\')
{
memcpy(curFilePath, filePath, i);
curFilePath[i] = '\0';
curPath = curFilePath;
if(FindFirstFile(curPath, &swf) == INVALID_HANDLE_VALUE) //目录不存在就创建
{
if(!CreateDirectory(curPath, NULL))
{
return FALSE;
}
}
}
i++;
}
return TRUE;
}
void CDownloader::AddDownloadWork(DLIO downloadWork)
{
char filePath[256] = {0};
char mUrl[512] = {0};
strcpy(mUrl, downloadWork.url);
strcpy(filePath, downloadWork.filePath);
int i = strlen(filePath) -1;
BOOL isPath = TRUE;
while(filePath[i] != '\\')
{
if(filePath[i] == '.' && filePath[i+1] != '\0')
{
isPath = FALSE;
}
i--;
}
if(isPath)
{
if(!CreateMultiDir(filePath))
return;
char fileName[256] = {0};
GetFileNameFormUrl(fileName,mUrl);
if(filePath[strlen(filePath)-1] != '\\')
{
strcat(filePath, "\\");
}
strcat(filePath, fileName);
}
else
{
char realPath[256] = {0};
for(int k=0; k<i; k++)
{
realPath[k] = filePath[k];
}
realPath[i] = '\\';
if(!CreateMultiDir(realPath))
return;
}
strcpy(m_dowloadWork[m_curIndex].url, mUrl);
strcpy(m_dowloadWork[m_curIndex].filePath, filePath);
m_curIndex++;
}
void CDownloader::GetFileNameFormUrl(char* fileName, const char* url)
{
int urlLen = strlen(url);
char mUrl[512] = {0};
char fName[256] = {0};
strcpy(mUrl, url);
int cutIndex = 0;
int i = urlLen - 1, j = 0;
while(mUrl[--i] != '/');
i++;
while(mUrl[i] != '\0' && mUrl[i] != '?' &&mUrl[i] != '&')
{
fName[j++] = mUrl[i++];
}
fName[j] = '\0';
strcpy(fileName, fName);
return ;
}
long CDownloader::GetLocalFileLenth(const char* fileName)
{
if(m_downloadCourse == 0) //文件已经开始下载的时候,取到的是下载前本地文件的大小;
return m_curLocalFileLenth;
char strTemp[256] = {0};
strcpy(strTemp,fileName);
FILE* fp = fopen(strTemp, "rb");
if(fp != NULL)
{
m_curLocalFileLenth = filelength(fileno(fp));
fclose(fp);
return m_curLocalFileLenth;
}
return 0;
}
double CDownloader::GetTotalFileLenth(const char* url)
{
char mUrl[512] = {0};
strcpy(mUrl, url);
double downloadFileLenth = 0;
CURL* pCurl = curl_easy_init();
curl_easy_setopt(pCurl, CURLOPT_URL, mUrl);
curl_easy_setopt(pCurl, CURLOPT_HEADER, 1L);
curl_easy_setopt(pCurl, CURLOPT_NOBODY, 1L);
if(curl_easy_perform(pCurl) == CURLE_OK)
{
curl_easy_getinfo(pCurl, CURLINFO_CONTENT_LENGTH_DOWNLOAD, &downloadFileLenth);
}
else
{
downloadFileLenth = -1;
}
curl_easy_cleanup(pCurl);
return downloadFileLenth;
}
size_t CDownloader::WriteFunc(char *str, size_t size, size_t nmemb, void *stream)
{
return fwrite(str, size, nmemb, (FILE*)stream);
}
size_t CDownloader::ProgressFunc(
double* pFileLen,
double t,// 下载时总大小
double d, // 已经下载大小
double ultotal, // 上传是总大小
double ulnow) // 已经上传大小
{
if(t == 0) return 0;
*pFileLen = d;
return 0;
}
int CDownloader::StartDownloadThread()
{
if(m_downloadCourse == -1||m_downloadCourse == 1)
{
HANDLE downloadThread = CreateThread(NULL, 0, SingleDownloadProc, this, 0, NULL);
CloseHandle(downloadThread);
return 0;
}
return -1;
}
DWORD WINAPI CDownloader::SingleDownloadProc(LPVOID lpParameter)
{
CDownloader* pDownload = (CDownloader*)lpParameter;
int curDLIndex = 0;
CURL* pCurl = curl_easy_init();
while(curDLIndex <= pDownload->m_curIndex)
{
char fileName[256] = {0};
char url[512] = {0};
strcpy(fileName, pDownload->m_dowloadWork[curDLIndex].filePath);
strcpy(url, pDownload->m_dowloadWork[curDLIndex].url);
strcpy(pDownload->m_curDownloadInfo.url, url);
strcpy(pDownload->m_curDownloadInfo.fileName, fileName);
long localFileLen = pDownload->GetLocalFileLenth(fileName);
pDownload->m_curLocalFileLenth = localFileLen;
pDownload->m_curDownloadInfo.preLocalLen = pDownload->m_curLocalFileLenth;
double totalFileLen = pDownload->m_curDownloadInfo.totalFileLen = pDownload->GetTotalFileLenth(url);
if(localFileLen >= (long)totalFileLen) //如果需要下载文件的大小大于等于本地文件的大小,直接下载下一个文件
{
curDLIndex++;
pDownload->m_downloadCourse = -1;
continue;
}
FILE* fp = fopen(fileName,"ab+");
if(fp == NULL) //文件打开错误,进行下一个文件的下载
{
pDownload->m_downloadCourse = -1;
continue;
}
curl_easy_setopt(pCurl, CURLOPT_URL, url);
curl_easy_setopt(pCurl, CURLOPT_TIMEOUT, pDownload->m_nConnectTimeOut);
curl_easy_setopt(pCurl, CURLOPT_HEADER, 0L);
curl_easy_setopt(pCurl, CURLOPT_NOBODY, 0L);
curl_easy_setopt(pCurl, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(pCurl, CURLOPT_RESUME_FROM, localFileLen);
curl_easy_setopt(pCurl, CURLOPT_WRITEFUNCTION, WriteFunc);
curl_easy_setopt(pCurl, CURLOPT_WRITEDATA, fp);
curl_easy_setopt(pCurl, CURLOPT_NOPROGRESS, 0L);
curl_easy_setopt(pCurl, CURLOPT_PROGRESSFUNCTION, ProgressFunc);
curl_easy_setopt(pCurl, CURLOPT_PROGRESSDATA, &(pDownload->m_curDownloadInfo.CurDownloadLen));
pDownload->m_downloadCourse = 0;
if(!curl_easy_perform(pCurl))
{
curDLIndex++;
pDownload->m_downloadCourse = -1;
}
fclose(fp);
}
curl_easy_cleanup(pCurl);
pDownload->m_downloadCourse = 1;
return 0;
}
int CDownloader::GetCurrentDownloadInfo(CURDI* lpCurDownloadInfor)
{
*lpCurDownloadInfor = m_curDownloadInfo;
return 0;
}
int CDownloader::SetConnectTimeOut(DWORD nConnectTimeOut)
{
if(m_downloadCourse == 0) return -1;
else
m_nConnectTimeOut = nConnectTimeOut;
return 0;
}
最后CurlTest.cpp :
// test.cpp : 定义控制台应用程序的入口点。
#include "stdafx.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <curl/curl.h>
#include "Downloader.h"
#pragma comment(lib,"libcurl.lib")
int _tmain(int argc, _TCHAR* argv[])
{
DWORD tick = GetTickCount();
CDownloader murl;
DLIO mDlWork;
strcpy(mDlWork.url, "http://sw.bos.baidu.com/sw-search-sp/software/f69ab46476e8e/TGPSetup_2.3.2.4083.exe");
strcpy(mDlWork.filePath, ".\\DownloadSoft\\");
murl.AddDownloadWork(mDlWork); //添加到下载任务中
strcpy(mDlWork.url, "http://sw.bos.baidu.com/sw-search-sp/software/16f6d358815f2/iTunes_12.5.1.21.exe");
strcpy(mDlWork.filePath, ".\\DownloadSoft\\");
murl.AddDownloadWork(mDlWork); //添加到下载任务中
murl.StartDownloadThread(); //开启下载线程
CURDI curInfo;
double curDownloadLen,preLen = 0.0;
while(1)
{
if(murl.IsDownloadBegin())
{
murl.GetCurrentDownloadInfo(&curInfo); //获取每次下载的信息(一次相当于毫秒级,这里速度也用毫秒计算)
curDownloadLen = curInfo.CurDownloadLen;
printf("正在下载:%s,下载进度:%6.2lf%%,下载速度:%9.2lfKB/s\r",curInfo.fileName,
((double)curInfo.preLocalLen+curInfo.CurDownloadLen)/curInfo.totalFileLen*100,(curDownloadLen-preLen)/(double)(GetTickCount()-tick));
tick = GetTickCount();
Sleep(500);
}
if(murl.IsDownloadEnd()) break;
preLen = curDownloadLen;
}
return 0;
}
1. CURLcode curl_global_init(long flags); 在多线程应用中,需要在主线程中调用这个函数。这个函数设置libcurl所需的环境。通常情况,如果不显式的调用它,第一次调用 curl_easy_init()时,curl_easy_init 会调用 curl_global_init,在单线程环境下,这不是问题。但是多线程下就不行了,因为curl_global_init不是线程安全的。在多个线 程中调用curl_easy_int,然后如果两个线程同时发现curl_global_init还没有被调用,同时调用 curl_global_init,悲剧就发生了。这种情况发生的概率很小,但可能性是存在的。
2. libcurl 有个很好的特性,它甚至可以控制域名解析的超时。但是在默认情况下,它是使用alarm + siglongjmp 实现的。用alarm在多线程下做超时,本身就几乎不可能。如果只是使用alarm,并不会导致程序崩溃,但是,再加上siglongjmp,就要命了 (程序崩溃的很可怕,core中几乎看不出有用信息),因为其需要一个sigjmp_buf型的全局变量,多线程修改它。(通常情况下,可以每个线程一个 sigjmp_buf 型的变量,这种情况下,多线程中使用 siglongjmp 是没有问题的,但是libcurl只有一个全局变量,所有的线程都会用)。
具体是类似 curl_easy_setopt(curl, CURLOPT_TIMEOUT, 30L) 的超时设置,导致alarm的使用(估计发生在域名解析阶段),如前所述,这在多线程中是不行的。解决方式是禁用掉alarm这种超时, curl_easy_setopt(curl, CURLOPT_NOSIGNAL, 1L)。
这样,多线程中使用超时就安全了。但是域名解析就没了超时机制,碰到很慢的域名解析,也很麻烦。文档的建议是 Consider building libcurl with c-ares support to enable asynchronous DNS lookups, which enables nice timeouts for name resolves without signals. c-ares 是异步的 DNS 解决方案。