TopLink源代码分析 JPA中的如何实现多数据库平台的访问

[问题的由来]

一般来说，通用的商业产品都可以访问多数据库。TopLink也不例外，其中的JPA号称支持上十种数据库平台，而且还可以动态增加。那么具体的技术细节是怎么样呢？怎么实现的呢？

我们来看一看TopLink的源代码是如何实现的。

 

[源代码分析]

第一步：在登陆访问数据库时，首先执行下面的函数

public void loginAndDetectDatasource() throws DatabaseException {     preConnectDatasource();     Connection conn = null;     try{       conn = (Connection)getReadLogin().connectToDatasource(null);         getLogin().setPlatformClassName(DBPlatformHelper.getDBPlatform(conn.getMetaData().getDatabaseProductName(), getSessionLog()));         }catch (SQLException ex){ DatabaseException dbEx =  DatabaseException.errorRetrieveDbMetadataThroughJDBCConnection();  // Typically exception would occur if user did not provide correct connection // parameters. The root cause of exception should be propogated up             dbEx.initCause(ex);             throw dbEx;         }finally{             if (conn != null){                 try{                     conn.close();                 }catch (SQLException ex){                     DatabaseException dbEx =  DatabaseException.errorRetrieveDbMetadataThroughJDBCConnection();    // Typically exception would occur if user did not provide correct connection   // parameters. The root cause of exception should be propogated up                     dbEx.initCause(ex);                     throw dbEx;                 }             }         }         connect();         postConnectDatasource();     }

 

第二步：在上面的函数中，有一个特别重要的函数，大家注意看，叫做DBPlatformHelper.getDBPlatform（）

下面我们来看看这个函数到底做了什么？

public static String getDBPlatform(String vendorName, SessionLog logger) {         initializeNameToVendorPlatform(logger);         String detectedDbPlatform = null;         if(vendorName != null) {     detectedDbPlatform = matchVendorNameInProperties(vendorName, _nameToVendorPlatform, logger);         }         if (logger.shouldLog(SessionLog.FINE) ) {      logger.log(SessionLog.FINE, "dbPlaformHelper_detectedVendorPlatform", detectedDbPlatform );  // NOI18N         }         if (detectedDbPlatform == null) {             if(logger.shouldLog(SessionLog.INFO)) {                 logger.log(SessionLog.INFO, "dbPlaformHelper_defaultingPlatform",  vendorName, DEFAULTPLATFORM); // NOI18N             }             detectedDbPlatform = DEFAULTPLATFORM;         }         return detectedDbPlatform;     }

大家注意看，上面这个函数的第一行是：

initializeNameToVendorPlatform(logger);

这个函数是干什么的？是不是就是访问多数据库强大功能的秘密呢？我们再往下面看。

 

第三步：秘密原来在这里

private static Properties initializeNameToVendorPlatform(SessionLog logger) {     synchronized(DBPlatformHelper.class) {     if(_nameToVendorPlatform == null) {                 _nameToVendorPlatform = new Properties();                 try {                loadFromResource(_nameToVendorPlatform, VENDOR_NAME_TO_PLATFORM_RESOURCE_NAME,                                           DBPlatformHelper.class.getClassLoader() );                 } catch (IOException e) {                     logger.log(SessionLog.WARNING, "dbPlaformHelper_noMappingFound", VENDOR_NAME_TO_PLATFORM_RESOURCE_NAME);                 }             }         }         return _nameToVendorPlatform;     }

大家注意看，上面这个函数中，有一个很关键的一行代码

loadFromResource(_nameToVendorPlatform,VENDOR_NAME_TO_PLATFORM_RESOURCE_NAME, DBPlatformHelper.class.getClassLoader() );

这行代码干的事情就是动态的对应到不同的数据库平台。那么到底是怎么做到的呢？

这行代码中有一个变量，名字叫做[VENDOR_NAME_TO_PLATFORM_RESOURCE_NAME]

它的真实面貌如下：

private final static String VENDOR_NAME_TO_PLATFORM_RESOURCE_NAME =PROPERTY_PATH + "VendorNameToPlatformMapping.properties";  //NOI18N

关键问题出现了，原来它是最后找到了一个配置文件，通过它解决不同数据库的访问难题，那么最后我们来看看这个配置文件里面写着什么。

 

第四步：配置文件

#Property file containing mapping between vendor name and vendor platform #The key of the property is in form of java regular expression. #At runtime, vendor name obtained from DatabaseMetaData#getDatabaseProductName() #is matched against the regular expresion to derive the DatabasePlaform we are #running against. SQL/ Anywhere=oracle.toplink.essentials.platform.database.SQLAnyWherePlatform (?i)oracle.*=oracle.toplink.essentials.platform.database.oracle.OraclePlatform (?i)(sybase.*)|(adaptive/ server/ enterprise.*)|(SQL/ Server)=oracle.toplink.essentials.platform.database.SybasePlatform (?i)microsoft.*=oracle.toplink.essentials.platform.database.SQLServerPlatform #Use JavaDBPlatform as the platform for Derby (?i).*derby=oracle.toplink.essentials.platform.database.JavaDBPlatform (?i).*db2.*=oracle.toplink.essentials.platform.database.DB2Platform (?i)pointbase.*=oracle.toplink.essentials.platform.database.PointBasePlatform (?i)mysql.*=oracle.toplink.essentials.platform.database.MySQL4Platform (?i)informix.*=oracle.toplink.essentials.platform.database.InformixPlatform (?i)postgresql.*=oracle.toplink.essentials.platform.database.PostgreSQLPlatform (?i).*symfoware.*=oracle.toplink.essentials.platform.database.SymfowarePlatform