ACE proactor example

網上看的兩個好的例子－學習ACE時經常看的。

這個文章應該是介紹ACE編程的一個很好的原創文章，個人非常推薦了！

1、WIN32下面用proactor可以達到幾乎RAW　IOCP的效率，由於封裝關係，應該是差那麼一點。

 

用戶端處理類的常規寫法：

//處理用戶端串連訊息class ClientHandler : public ACE_Service_Handler{public:/**建構函式 * * */ClientHandler(unsigned int client_recv_buf_size=SERVER_CLIENT_RECEIVE_BUF_SIZE) :_read_msg_block(client_recv_buf_size),_io_count(0){}~ClientHandler(){}/** *初始化，因為可能要用到ClientHandler記憶體池，而這個池又不一定要用NEW */void init();/**清理函數，因為可能要用到記憶體池 * */void fini();//檢查是否逾時的函數void check_time_out(time_t cur_time);public:/**用戶端串連伺服器成功後調用 * * /param handle 通訊端控制代碼 * /param &message_block 第一次讀到的資料（未用） *///由Acceptor來調用！！！virtual void open (ACE_HANDLE handle,ACE_Message_Block &message_block);/**處理網路讀操作結束訊息 * * /param &result 讀操作結果 */virtual void handle_read_stream (const ACE_Asynch_Read_Stream::Result &result);/**處理網路寫操作結束訊息 * * /param &result 寫操作結果 */virtual void handle_write_stream (const ACE_Asynch_Write_Stream::Result &result);private://**產生一個網路讀請求 * * /param void  * /return 0-成功，-1失敗 */int  initiate_read_stream  (void);/**產生一個寫請求 * * /param mb 待發送的資料 * /param nBytes 待發送資料大小 * /return 0－成功，－1失敗 */int  initiate_write_stream (ACE_Message_Block & mb, size_t nBytes );/** * * /return 檢查是否可以刪除，用的是一個引用計數。每一個外出IO的時候＋1，每一個IO成功後－1 */int check_destroy();//非同步讀ACE_Asynch_Read_Stream _rs;//非同步寫ACE_Asynch_Write_Stream _ws;//接收緩衝區只要一個就夠了，因為壓根就沒想過要多讀，我直到現在也不是很清楚為什麼要多讀，多讀的話要考慮很多問題ACE_Message_Block _read_msg_block;//通訊端控制代碼,這個可以不要了，因為基類就有個HANDLER在裡面的。//ACE_HANDLE _handle;//一個鎖，用戶端反正有東東要鎖的，注意，要用ACE_Recursive_Thread_Mutex而不是用ACE_Thread_Mutex，這裡面是可以重入的，而且在WIN32下是直接的EnterCriticalSection，可以達到很高的效率ACE_Recursive_Thread_Mutex _lock;//在外IO數量,其實就是引用計數啦，沒啥的。為0的時候就把這個東東關掉啦。long _io_count;//檢查逾時用的，過段時間沒東東就CLOSE他了。time_t _last_net_io;private://本來想用另外一種模型的，只用1個或者2個外出讀，後來想想，反正一般記憶體都是足夠的，就不管了。//ACE_Message_Block _send_msg_blocks[2];//ACE_Message_Block &_sending_msg_block;//ACE_Message_Block &_idle_msg_block;private:public://TODO:move to prriva and use friend class!!!//只是為了效率更高，不用STL的LIST是因為到現在我沒有可用的Node_Allocator，所以效率上會有問題。ClientHandler *_next;ClientHandler *next(){return _next;}void next(ClientHandler *obj){_next=obj;}}; //這是具體實現，有些地方比較亂，懶得管了，鎖的有些地方不對。懶得改了，反正在出錯或者有瓶頸的時候再做也不遲。void ClientHandler::handle_read_stream (const ACE_Asynch_Read_Stream::Result &result){_last_net_io=ACE_OS::time(NULL);int byterecved=result.bytes_transferred ();if ( (result.success ()) && (byterecved != 0)){ //ACE_DEBUG ((LM_DEBUG,  "Receiver completed:%d/n",byterecved));//處理完資料 if(handle_received_data()==true) {  //ACE_DEBUG ((LM_DEBUG,  "go on reading.../n"));//把東東推到頭部，處理粘包  _read_msg_block.crunch();  initiate_read_stream(); }}//這個地方不想用ACE_Atom_op，因為反正要有一個鎖，而且一般都會用鎖，不管了。假如不在意的話，應該直接用ACE_Atom_Op以達到最好的效率{ ACE_Guard<ACE_Recursive_Thread_Mutex> locker (_lock); _io_count--;}check_destroy ();}void ClientHandler::init(){//初始化資料，並不在建構函式裡做。_last_net_io=ACE_OS::time(NULL);_read_msg_block.rd_ptr(_read_msg_block.base());_read_msg_block.wr_ptr(_read_msg_block.base());this->handle(ACE_INVALID_HANDLE);}bool ClientHandler::handle_received_data(){...........自己處理return true;}//==================================================================void ClientHandler::handle_write_stream (const ACE_Asynch_Write_Stream::Result &result){//發送成功，RELEASE掉//這個不可能有多個RELEASE，直接XX掉//result.message_block ().release ();MsgBlockManager::get_instance().release_msg_block(&result.message_block());{ ACE_Guard<ACE_Recursive_Thread_Mutex> locker (_lock); _io_count--;}check_destroy ();}//bool ClientHandler::destroy () //{// FUNC_ENTER;// ClientManager::get_instance().release_client_handle(this);// FUNC_LEAVE;// return false ;//}int  ClientHandler::initiate_read_stream  (void){ACE_Guard<ACE_Recursive_Thread_Mutex> locker (_lock);//考慮到粘包的呀if (_rs.read (_read_msg_block, _read_msg_block.space()) == -1){ ACE_ERROR_RETURN ((LM_ERROR,"%p/n","ACE_Asynch_Read_Stream::read"),-1);}_io_count++;return 0;}/**產生一個寫請求** /param mb 待發送的資料* /param nBytes 待發送資料大小* /return 0－成功，－1失敗*/int  ClientHandler::initiate_write_stream (ACE_Message_Block & mb, size_t nBytes ){ACE_Guard<ACE_Recursive_Thread_Mutex> locker (_lock);if (_ws.write (mb , nBytes ) == -1){ mb.release (); ACE_ERROR_RETURN((LM_ERROR,"%p/n","ACE_Asynch_Write_File::write"),-1);}_io_count++;return 0;}void ClientHandler::open (ACE_HANDLE handle,ACE_Message_Block &message_block){//FUNC_ENTER;_last_net_io=ACE_OS::time(NULL);_io_count=0;if(_ws.open(*this,this->handle())==-1){ ACE_ERROR ((LM_ERROR,"%p/n","ACE_Asynch_Write_Stream::open"));}else if (_rs.open (*this, this->handle()) == -1){ ACE_ERROR ((LM_ERROR,"%p/n","ACE_Asynch_Read_Stream::open"));}else{ initiate_read_stream ();}check_destroy();//FUNC_LEAVE;}void ClientHandler::fini(){}void ClientHandler::check_time_out(time_t cur_time){//ACE_Guard<ACE_Recursive_Thread_Mutex> locker (_lock);//ACE_DEBUG((LM_DEBUG,"cur_time is %u,last io is %u/n",cur_time,_last_net_io));//檢測是否已經為0了if(this->handle()==ACE_INVALID_HANDLE) return;if(cur_time-_last_net_io>CLIENT_TIME_OUT_SECONDS){ ACE_OS::shutdown(this->handle(),SD_BOTH); ACE_OS::closesocket(this->handle()); this->handle(ACE_INVALID_HANDLE);}}int ClientHandler::check_destroy(){{ ACE_Guard<ACE_Recursive_Thread_Mutex> locker (_lock); if (_io_count> 0)  return 1;}ACE_OS::shutdown(this->handle(),SD_BOTH);ACE_OS::closesocket(this->handle());this->handle(ACE_INVALID_HANDLE);//這個地方給記憶體池吧。ClientManager::get_instance().release_client_handle(this);//delete this;return 0;} 這個也很好！ACE的好文！真是不轉我覺得後悔啊！沒啥好說的，管理所有的用戶端和記憶體池的功能。class ClientManager : public SingleTon<ClientManager>{public:ClientManager():_header(NULL){}~ClientManager(){}public:void init(unsigned int default_pool_size,unsigned int default_read_buf_size);void fini();public:ClientHandler *get_clienthandler();void release_client_handle(ClientHandler *client);void check_time_out();size_t get_client_count();private:ClientHandler *_header;std::set<ClientHandler *> _active_clients;ACE_Recursive_Thread_Mutex _lock;};  #include "clientmanager.h"#include <ace/Guard_T.h>ClientHandler *ClientManager::get_clienthandler(){FUNC_ENTER;ACE_Guard<ACE_Recursive_Thread_Mutex> locker(_lock);ClientHandler *ret=NULL;if(_header==NULL){ ACE_DEBUG((LM_DEBUG,"client > max clients!!!/n"));}else{ ret=_header; _header=_header->next(); ret->init(); _active_clients.insert(ret);}FUNC_LEAVE;return ret;}void ClientManager::release_client_handle(ClientHandler *client){//FUNC_ENTER;ACE_Guard<ACE_Recursive_Thread_Mutex> locker(_lock);client->fini();client->next(_header);_header=client;_active_clients.erase(client);//FUNC_LEAVE;}void ClientManager::init(unsigned int default_pool_size,unsigned int default_read_buf_size){//FUNC_ENTER;for(unsigned int i=0;i<default_pool_size;i++){ ClientHandler *client=new ClientHandler(default_read_buf_size); client->next(_header); _header=client;}//FUNC_LEAVE;}void ClientManager::fini(){//FUNC_ENTER;while(_header){ ClientHandler *temp=_header->next(); delete _header; _header=temp;}//FUNC_LEAVE;}void ClientManager::check_time_out(){time_t cur_time=ACE_OS::time(NULL);ACE_Guard<ACE_Recursive_Thread_Mutex> locker(_lock);for(std::set<ClientHandler *>::iterator it=_active_clients.begin();it!=_active_clients.end();it++){ (*it)->check_time_out(cur_time);}}size_t ClientManager::get_client_count(){ACE_Guard<ACE_Recursive_Thread_Mutex> locker(_lock);return _active_clients.size();}//伺服器的設計按照七貓的說話，這個架構可以達到IOCP的效率，真的利害！但是我也不知道真偽！所以大家不要認為是我說的阿！我沒有測試過，所以也不太清楚是否真的有那麼高的效率！沒有什麼可說的！ 好文章！class MyServer : public SingleTon<MyServer >{public:/**主伺服器初始化工作** /param *listenaddr 監聽地址："192.168.0.188:80"－在192.168.0.188的80連接埠進行監聽"80"－在所有IP地址的80連接埠進行監聽* /param numOfThreads 伺服器網路訊息處理的線程個數* /return 1:成功，0或者其他值：失敗*/int init(const char *listenaddr,unsigned int numOfThreads);/**最後清理工作，資源釋放工作 * */void fini();/**主伺服器開始運行** /return 1－成功，－1失敗*/int start();/**主伺服器停止運行**/void stop();private://工作管理員（線程池）ServerTask _task;//監聽地址ACE_INET_Addr _listen_addr;//網路接收器ClientAcceptor _acceptor;//網路訊息處理線程數量unsigned int _num_of_threads;private: Observer _observer;//檢查是否有用戶端逾時CheckClientTimeoutHandler _check_time_out_handler;};int MyServer::init(const char *listenaddr,unsigned int numOfThreads){//ACE_WIN32_Proactor *pImpl = new ACE_WIN32_Proactor;//static ACE_Proactor proactor(pImpl,1);//ACE_Proactor::instance( & proactor);_listen_addr=ACE_INET_Addr(listenaddr);//_num_of_threads=numOfThreads;_num_of_threads=1;_observer.init();_syn_cms_handler.init();_check_time_out_handler.init();return 1;}void MyServer::fini(){ItemManager::get_instance().purge_all_items();_observer.fini();_syn_cms_handler.fini();_check_time_out_handler.fini();}/**主伺服器開始運行** /return 1－成功，－1失敗*/int MyServer::start(){int Rc = _acceptor.open (_listen_addr,0,1);if(Rc==-1){ ACE_ERROR_RETURN ((LM_ERROR, "acceptor error./n"), -1);} //每20秒檢查一次，檢查是否有用戶端逾時ACE_Time_Value check_client_timeout_interval(120);Rc=ACE_Proactor::instance ()->schedule_timer ( _check_time_out_handler, (void *) "timeout", ACE_Time_Value::zero, check_client_timeout_interval);if(Rc==-1){ ACE_ERROR_RETURN ((LM_ERROR, "%p/n", "check_client_timeout schedule_timer"), -1);}ACE_Time_Value observertime(20);Rc=ACE_Proactor::instance ()->schedule_timer ( _observer, (void *) "observer", ACE_Time_Value::zero, observertime);if(Rc==-1){ ACE_ERROR_RETURN ((LM_ERROR, "%p/n", "observer schedule_timer"), -1);}if (_task.activate (THR_NEW_LWP, _num_of_threads ) == -1){ ACE_ERROR_RETURN ((LM_ERROR, "task start error./n", "main"), -1);}return 1;}/**主伺服器停止運行**/void MyServer::stop(){ACE_Proactor::end_event_loop () ;ACE_Thread_Manager * pTM = ACE_Thread_Manager::instance();pTM->wait_task ( & _task) ;//ACE_Proactor::instance( ( ACE_Proactor* )NULL );}int ACE_TMAIN(int argc,char *argv[]){FUNC_ENTER; std::cout<<"size of item is "<<sizeof(Item)<<std::endl;//設定日誌ACE_LOG_MSG->open (argv[0],ACE_Log_Msg::OSTREAM);//std::ofstream myostream (SERVER_LOG_FILE_NAME, ios::out | ios::trunc);//ACE_LOG_MSG->msg_ostream (&myostream);//讀入設定檔ACE_DEBUG((LM_DEBUG,"read config file from %s/n",SERVER_CONFIG_FILE));XMLConfig::get_instance().read_config_from_xml_file(SERVER_CONFIG_FILE);//初始化MSGBLOCK池ACE_DEBUG((LM_DEBUG,"starting init MsgBlockManager.../n"));MsgBlockManager::get_instance().init(XMLConfig::get_instance().get_num_of_msg_blocks(),XMLConfig::get_instance().get_size_of_write_msg_blocks());//初始化串連池ACE_DEBUG((LM_DEBUG,"starting init ClientManager.../n"));ClientManager::get_instance().init(XMLConfig::get_instance().get_num_of_clients(),XMLConfig::get_instance().get_size_of_client_read_buf()); //開始伺服器ACE_DEBUG((LM_DEBUG,"starting init MyServer.../n"));MyServer::get_instance().init(XMLConfig::get_instance().get_listen_addr(),XMLConfig::get_instance().get_num_of_proactor_threads());ACE_DEBUG((LM_DEBUG,"starting MyServer.../n"));MyServer::get_instance().start();ACE_DEBUG((LM_DEBUG,"Enter looping.../n"));while(true){ std::cout << "Input Command=>/n" << std::flush ; std::string inputcmd; std::getline(std::cin,inputcmd); if(!handle_console_cmd(inputcmd))  break;}//char c ;//std::cout << "Press any key to stop and exit=>/n" << std::flush ;//std::cin.clear ();//std::cin >> c ;MyServer::get_instance().stop();MsgBlockManager::get_instance().fini();ClientManager::get_instance().fini();return 1;}

轉帖：http://blog.csdn.net/hejianhua/article/details/7045332