MySQL由于它本身的小巧和操作的高效, 在数据库应用中越来越多的被采用.我在开发一个P2P应用的时候曾经使用MySQL来保存P2P节点,由于P2P的应用中,结点数动辄上万个,而且节点变化频繁,因此一定要保持查询和插入的高效.以下是我在使用过程中做的提高效率的三个有效的尝试.
l
使用statement进行绑定查询
使用statement可以提前构建查询语法树,在查询时不再需要构建语法树就直接查询.因此可以很好的提高查询的效率. 这个方法适合于查询条件固定但查询非常频繁的场合.
使用方法是:
绑定, 创建一个MYSQL_STMT变量,与对应的查询字符串绑定,字符串中的问号代表要传入的变量,每个问号都必须指定一个变量.
查询, 输入每个指定的变量, 传入MYSQL_STMT变量用可用的连接句柄执行.
代码如下:
//1.绑定
bool CDBManager::BindInsertStmt(MYSQL * connecthandle)
{
//作插入操作的绑定
MYSQL_BIND insertbind[FEILD_NUM];
if(m_stInsertParam == NULL)
m_stInsertParam = new CHostCacheTable;
m_stInsertStmt = mysql_stmt_init(connecthandle);
//构建绑定字符串
char insertSQL[SQL_LENGTH];
strcpy(insertSQL, "insert into HostCache(SessionID, ChannelID, ISPType, "
"ExternalIP, ExternalPort, InternalIP, InternalPort) "
"values(?, ?, ?, ?, ?, ?, ?)");
mysql_stmt_prepare(m_stInsertStmt, insertSQL, strlen(insertSQL));
int param_count= mysql_stmt_param_count(m_stInsertStmt);
if(param_count != FEILD_NUM)
return false;
//填充bind结构数组, m_sInsertParam是这个statement关联的结构变量
memset(insertbind, 0, sizeof(insertbind));
insertbind[0].buffer_type = MYSQL_TYPE_STRING;
insertbind[0].buffer_length = ID_LENGTH /* -1 */;
insertbind[0].buffer = (char *)m_stInsertParam->sessionid;
insertbind[0].is_null = 0;
insertbind[0].length = 0;
insertbind[1].buffer_type = MYSQL_TYPE_STRING;
insertbind[1].buffer_length = ID_LENGTH /* -1 */;
insertbind[1].buffer = (char *)m_stInsertParam->channelid;
insertbind[1].is_null = 0;
insertbind[1].length = 0;
insertbind[2].buffer_type = MYSQL_TYPE_TINY;
insertbind[2].buffer = (char *)&m_stInsertParam->ISPtype;
insertbind[2].is_null = 0;
insertbind[2].length = 0;
insertbind[3].buffer_type = MYSQL_TYPE_LONG;
insertbind[3].buffer = (char *)&m_stInsertParam->externalIP;
insertbind[3].is_null = 0;
insertbind[3].length = 0;
insertbind[4].buffer_type = MYSQL_TYPE_SHORT;
insertbind[4].buffer = (char *)&m_stInsertParam->externalPort;
insertbind[4].is_null = 0;
insertbind[4].length = 0;
insertbind[5].buffer_type = MYSQL_TYPE_LONG;
insertbind[5].buffer = (char *)&m_stInsertParam->internalIP;
insertbind[5].is_null = 0;
insertbind[5].length = 0;
insertbind[6].buffer_type = MYSQL_TYPE_SHORT;
insertbind[6].buffer = (char *)&m_stInsertParam->internalPort;
insertbind[6].is_null = 0;
insertbind[6].is_null = 0;
//绑定
if (mysql_stmt_bind_param(m_stInsertStmt, insertbind))
return false;
return true;
}
//2.查询
bool CDBManager::InsertHostCache2(MYSQL * connecthandle, char * sessionid, char * channelid, int ISPtype, \
unsigned int eIP, unsigned short eport, unsigned int iIP, unsigned short iport)
{
//填充结构变量m_sInsertParam
strcpy(m_stInsertParam->sessionid, sessionid);
strcpy(m_stInsertParam->channelid, channelid);
m_stInsertParam->ISPtype = ISPtype;
m_stInsertParam->externalIP = eIP;
m_stInsertParam->externalPort = eport;
m_stInsertParam->internalIP = iIP;
m_stInsertParam->internalPort = iport;
//执行statement,性能瓶颈处
if(mysql_stmt_execute(m_stInsertStmt))
return false;
return true;
}
l
随机的获取记录
在某些数据库的应用中, 我们并不是要获取所有的满足条件的记录,而只是要随机挑选出满足条件的记录. 这种情况常见于数据业务的统计分析,从大容量数据库中获取小量的数据的场合.
有两种方法可以做到
1. 常规方法,首先查询出所有满足条件的记录,然后随机的挑选出部分记录.这种方法在满足条件的记录数很多时效果不理想.
2. 使用limit语法,先获取满足条件的记录条数, 然后在sql查询语句中加入limit来限制只查询满足要求的一段记录. 这种方法虽然要查询两次,但是在数据量大时反而比较高效.
示例代码如下:
//1.常规的方法
//性能瓶颈,10万条记录时,执行查询140ms, 获取结果集500ms,其余可忽略
int CDBManager::QueryHostCache(MYSQL* connecthandle, char * channelid, int ISPtype, CDBManager::CHostCacheTable * &hostcache)
{
char selectSQL[SQL_LENGTH];
memset(selectSQL, 0, sizeof(selectSQL));
sprintf(selectSQL,"select * from HostCache where ChannelID = '%s' and ISPtype = %d", channelid, ISPtype);
if(mysql_real_query(connecthandle, selectSQL, strlen(selectSQL)) != 0) //检索
return 0;
//获取结果集
m_pResultSet = mysql_store_result(connecthandle);
if(!m_pResultSet) //获取结果集出错
return 0;
int iAllNumRows = (int)(mysql_num_rows(m_pResultSet)); ///<所有的搜索结果数
//计算待返回的结果数
int iReturnNumRows = (iAllNumRows <= RETURN_QUERY_HOST_NUM)? iAllNumRows:RETURN_QUERY_HOST_NUM;
if(iReturnNumRows <= RETURN_QUERY_HOST_NUM)
{
//获取逐条记录
for(int i = 0; i<iReturnNumRows; i++)
{
//获取逐个字段
m_Row = mysql_fetch_row(m_pResultSet);
if(m_Row[0] != NULL)
strcpy(hostcache.sessionid, m_Row[0]);
if(m_Row[1] != NULL)
strcpy(hostcache.channelid, m_Row[1]);
if(m_Row[2] != NULL)
hostcache.ISPtype = atoi(m_Row[2]);
if(m_Row[3] != NULL)
hostcache.externalIP = atoi(m_Row[3]);
if(m_Row[4] != NULL)
hostcache.externalPort = atoi(m_Row[4]);
if(m_Row[5] != NULL)
hostcache.internalIP = atoi(m_Row[5]);
if(m_Row[6] != NULL)
hostcache.internalPort = atoi(m_Row[6]);
}
}
else
{
//随机的挑选指定条记录返回
int iRemainder = iAllNumRows%iReturnNumRows; ///<余数
int iQuotient = iAllNumRows/iReturnNumRows; ///<商
int iStartIndex = rand()%(iRemainder + 1); ///<开始下标
//获取逐条记录
for(int iSelectedIndex = 0; iSelectedIndex < iReturnNumRows; iSelectedIndex++)
{
mysql_data_seek(m_pResultSet, iStartIndex + iQuotient * iSelectedIndex);
m_Row = mysql_fetch_row(m_pResultSet);
if(m_Row[0] != NULL)
strcpy(hostcache[iSelectedIndex].sessionid, m_Row[0]);
if(m_Row[1] != NULL)
strcpy(hostcache[iSelectedIndex].channelid, m_Row[1]);
if(m_Row[2] != NULL)
hostcache[iSelectedIndex].ISPtype = atoi(m_Row[2]);
if(m_Row[3] != NULL)
hostcache[iSelectedIndex].externalIP = atoi(m_Row[3]);
if(m_Row[4] != NULL)
hostcache[iSelectedIndex].externalPort = atoi(m_Row[4]);
if(m_Row[5] != NULL)
hostcache[iSelectedIndex].internalIP = atoi(m_Row[5]);
if(m_Row[6] != NULL)
hostcache[iSelectedIndex].internalPort = atoi(m_Row[6]);
}
}
//释放结果集内容
mysql_free_result(m_pResultSet);
return iReturnNumRows;
}
//2.使用limit版
int CDBManager::QueryHostCache(MYSQL * connecthandle, char * channelid, unsigned int myexternalip, int ISPtype, CHostCacheTable * hostcache)
{
//首先获取满足结果的记录条数,再使用limit随机选择指定条记录返回
MYSQL_ROW row;
MYSQL_RES * pResultSet;
char selectSQL[SQL_LENGTH];
memset(selectSQL, 0, sizeof(selectSQL));
sprintf(selectSQL,"select count(*) from HostCache where ChannelID = '%s' and ISPtype = %d", channelid, ISPtype);
if(mysql_real_query(connecthandle, selectSQL, strlen(selectSQL)) != 0) //检索
return 0;
pResultSet = mysql_store_result(connecthandle);
if(!pResultSet)
return 0;
row = mysql_fetch_row(pResultSet);
int iAllNumRows = atoi(row[0]);
mysql_free_result(pResultSet);
//计算待取记录的上下范围
int iLimitLower = (iAllNumRows <= RETURN_QUERY_HOST_NUM)?
0:(rand()%(iAllNumRows - RETURN_QUERY_HOST_NUM));
int iLimitUpper = (iAllNumRows <= RETURN_QUERY_HOST_NUM)?
iAllNumRows:(iLimitLower + RETURN_QUERY_HOST_NUM);
//计算待返回的结果数
int iReturnNumRows = (iAllNumRows <= RETURN_QUERY_HOST_NUM)?
iAllNumRows:RETURN_QUERY_HOST_NUM;
//使用limit作查询
sprintf(selectSQL,"select SessionID, ExternalIP, ExternalPort, InternalIP, InternalPort "
"from HostCache where ChannelID = '%s' and ISPtype = %d limit %d, %d"
, channelid, ISPtype, iLimitLower, iLimitUpper);
if(mysql_real_query(connecthandle, selectSQL, strlen(selectSQL)) != 0) //检索
return 0;
pResultSet = mysql_store_result(connecthandle);
if(!pResultSet)
return 0;
//获取逐条记录
for(int i = 0; i<iReturnNumRows; i++)
{
//获取逐个字段
row = mysql_fetch_row(pResultSet);
if(row[0] != NULL)
strcpy(hostcache.sessionid, row[0]);
if(row[1] != NULL)
hostcache.externalIP = atoi(row[1]);
if(row[2] != NULL)
hostcache.externalPort = atoi(row[2]);
if(row[3] != NULL)
hostcache.internalIP = atoi(row[3]);
if(row[4] != NULL)
hostcache.internalPort = atoi(row[4]);
}
//释放结果集内容
mysql_free_result(pResultSet);
return iReturnNumRows;
}
l
使用连接池管理连接.
在有大量节点访问的数据库设计中,经常要使用到连接池来管理所有的连接.
一般方法是:建立两个连接句柄队列,空闲的等待使用的队列和正在使用的队列.
当要查询时先从空闲队列中获取一个句柄,插入到正在使用的队列,再用这个句柄做数据库操作,完毕后一定要从使用队列中删除,再插入到空闲队列.
设计代码如下:
//定义句柄队列
typedef std::list<MYSQL *> CONNECTION_HANDLE_LIST;
typedef std::list<MYSQL *>::iterator CONNECTION_HANDLE_LIST_IT;
//连接数据库的参数结构
class CDBParameter
{
public:
char *host; ///<主机名
char *user; ///<用户名
char *password; ///<密码
char *database; ///<数据库名
unsigned int port; ///<端口，一般为0
const char *unix_socket; ///<套接字，一般为NULL
unsigned int client_flag; ///<一般为0
};
//创建两个队列
CONNECTION_HANDLE_LIST m_lsBusyList; ///<正在使用的连接句柄
CONNECTION_HANDLE_LIST m_lsIdleList; ///<未使用的连接句柄
//所有的连接句柄先连上数据库,加入到空闲队列中,等待使用.
bool CDBManager::Connect(char * host /* = "localhost" */, char * user /* = "chenmin" */, \
char * password /* = "chenmin" */, char * database /* = "HostCache" */)
{
CDBParameter * lpDBParam = new CDBParameter();
lpDBParam->host = host;
lpDBParam->user = user;
lpDBParam->password = password;
lpDBParam->database = database;
lpDBParam->port = 0;
lpDBParam->unix_socket = NULL;
lpDBParam->client_flag = 0;
try
{
//连接
for(int index = 0; index < CONNECTION_NUM; index++)
{
MYSQL * pConnectHandle = mysql_init((MYSQL*) 0); //初始化连接句柄
if(!mysql_real_connect(pConnectHandle, lpDBParam->host, lpDBParam->user, lpDBParam->password,\
lpDBParam->database,lpDBParam->port,lpDBParam->unix_socket,lpDBParam->client_fla))
return false;
//加入到空闲队列中
m_lsIdleList.push_back(pConnectHandle);
}
}
catch(...)
{
return false;
}
return true;
}
//提取一个空闲句柄供使用
MYSQL * CDBManager::GetIdleConnectHandle()
{
MYSQL * pConnectHandle = NULL;
m_ListMutex.acquire();
if(m_lsIdleList.size())
{
pConnectHandle = m_lsIdleList.front();
m_lsIdleList.pop_front();
m_lsBusyList.push_back(pConnectHandle);
}
else //特殊情况,闲队列中为空,返回为空
{
pConnectHandle = 0;
}
m_ListMutex.release();
return pConnectHandle;
}
//从使用队列中释放一个使用完毕的句柄,插入到空闲队列
void CDBManager::SetIdleConnectHandle(MYSQL * connecthandle)
{
m_ListMutex.acquire();
m_lsBusyList.remove(connecthandle);
m_lsIdleList.push_back(connecthandle);
m_ListMutex.release();
}
//使用示例,首先获取空闲句柄,利用这个句柄做真正的操作,然后再插回到空闲队列
bool CDBManager::DeleteHostCacheBySessionID(char * sessionid)
{
MYSQL * pConnectHandle = GetIdleConnectHandle();
if(!pConnectHandle)
return 0;
bool bRet = DeleteHostCacheBySessionID(pConnectHandle, sessionid);
SetIdleConnectHandle(pConnectHandle);
return bRet;
}
//传入空闲的句柄,做真正的删除操作
bool CDBManager::DeleteHostCacheBySessionID(MYSQL * connecthandle, char * sessionid)
{
char deleteSQL[SQL_LENGTH];
memset(deleteSQL, 0, sizeof(deleteSQL));
sprintf(deleteSQL,"delete from HostCache where SessionID = '%s'", sessionid);
if(mysql_query(connecthandle,deleteSQL) != 0) //删除
return false;
return true;
}