Hibernate实现Clob和Blob对象的存取

作为测试，我们新建一个TUser对象，其image属性中，保存了一个图片文件的二进制内容。而其resume属性，我们以一个简单的字符串作为填充。

 

 1 TUser user  = new  TUser(); 
 2
 3 user.setAge( new  Integer( 20 )); 
 4
 5 user.setName( " Shark " ); 
 6
 7 FilelnputStream imgis  = new  FileinputStream( " C:\\inimage.jpg "  
 8
 9 Blob img  =  Hibernate.createBlob(imgis); 
10
11 user.setlmage(img); 
12
13 Clob resume  =  Hibernate.createClob( " This is Clob " ); 
14
15 user. setResume(resume); 
16
17 Transaction tx  =  session.beginTransaction(); 
18
19 session.save(user); 
20
21 tx.commit(); 
22
23

上面的代码中，我们通过Hibemate.createBlob和Hibemate.createClob创建了对应的Blob和Clob对象。其中Blob对象基于一个FileInputStream构建，而Clob对象基于一个字符串构建。

 

完成了写入操作，对应的读取操作代码如下:

 

 1 //  假设库表记录的id字段等于3 
 2 TUser user = (TUser)  session.load(TUger.elaa., load(TUser. class ,   new  Integer( 3 )); 
 3 Clob resume = user.getResume(); 
 4 //  通过Clob.getSubString()方法获取Clob字段内容 
 5 System.out.println( " User resume=> " + resume.getSubString( 1 ,( int )resume.length())); 
 6 Blob img  =  user.getImage(); 
 7 //  通过Blob.getBinaryS=ream()方法获取二进制流 
 8 InputStream is  =  img.getBinaryStream(); 
 9 FileOutputStream fos = new  FileOutputStream( " C:\\outimage.jpg " ); 
10 byte [] buf = new   byte ( 102400 ); 
11 int  len; 
12 while ((len  =  is.read(buf)) !=- 1 ) { 
13     fos.write(buf, 0 ,len); 
14 }  
15 fos.close(); 
16 is.close(); 
17
18

 

通过上面的代码，我们完成了针对SQLServer的Blob/Clob型字段操作.看起来非常简单，不过，并非侮种数据库都如此友善。让我们接着来看看Oracle数据库下的Blob/Clob字段读写，

通过修改hibernate.cfg.xml中的Dialect配置完成数据库切换后，我们再次运行上面的TUser对象保存例程。

程序运行期间抛出异常:

 

Hibernate:select hibernate_sequence.nextval from dual

Hibernate:insert into T_USER (name, age,  image,resume. id) values(?, ?, ?, ?, ?)

17:27:24,161 ERROR JDBCExceptionReporter:58 - - 不允许的操作： Streams type cannot be used in batching

17:27:24,171 ERROR Sessionlmpl:2399 - Could not synchronize database state with session

net.sf.hibernate.exception.GenericJDBCException:could not insert:[com.redsaga.hibernate.db.entity.TUser#6]

...

 

观察异常信息:streams type cannot be used in batching.这意味着Oracle JDBC不允许流操作以批量方式执行(Oracle CLOB采用流机制作为数据读写方式)。

这种错误一般发生在hibernate.cfg.xml中的hibernate jdbc.batch_size设定大于0的情况，将hibernate.jdbc.batch_size修改为0即可消除。

 

 

 

             ...

            

preStmt.setint(l,5);

ResultSet rset=preStmt.executeQuery();

// 注意我们这里需要引用Oracle原生BLOB定义，如果使用了Weblogic JDBC Vendor

// 则应使用weblogic.jdbc.vendor.oracle. OracleThinBLob/OracleThinCLOb

rset.next();

oracle.sql.BLOB imqBlob = (oracle.sql.BLOB) rset.getBlob(1);

oracle.sql.CLOB resClob = (oracle.sql.CLOB) rset.getClob(2);

//======= 将二进创数据写入Blob

FileInputStream inStream = new FileinputStream("c\\inimage.jpg");

OutputStream outStream = imgBlob.getBinaryOutputStream();

byte[] buf=new byte[10240];//10K 读取缓存

int len;

while((len=inStream.read(buf))>0){

   outStream.write(buf,0,len);

}

inStream.close();

outStream.close():

//======= 将字符串写入Clob

resClob.putString(1 ，"This is my Glob"

//======= 将Blob/Clob字段更新到数据序

preStmt= dbconn.prepareStatement("update T_USER set  image=?,  resume=? where id=?");

preStmt.setBlob(1,imgBlob);

preStmt.setClob(2,resClob):

preStmt.setlnt(3 ，5);

preStmt.executeUpdate();

preStmt.close():

dbconn.commit();

dbconn.close():

 

上面的代码说明了Oracle中Blob/Clob字段操作的一般机制，那么，基于Hibernate的持久层实现中，应该如何对Blob/Clob字段进行处理?

我们知道，Hibernate底层数据访问机制仍然是基于JDBC实现，那么也就意味着我们必须在Hibernate中模拟JDBC的访问流程：

TUser user=new TUser();

user.setAge(new Integer(20));

user.setName("Shark');

user.setImage(Hibernate.createSlob(new byte [1])):

user.setResume(Hibernate.createClob(" "));// 注意这里的参教是一个空格

Transaction tx=session.beginTransaction();

session.save(user):

// 调用flush方法，强制Hibernate立即执行insert sql

session.flush();

// 通过refresh方法，强制Hibernate执行select for update

session.refresh(user, LockMode.UPGRADE);

// 向Blob写入实际内容

oracle.sql.BLOB blob=(oracle.sql.BLOB)user.getImage();

OutputStream out=blob. getBinaryOutputStream();

FileInputStream imgis=new FileInputStream("C:\\inimage.jpg");

byte[] buf=new byte[10240];//10K 缓存

int len;

while((len=imgis.read(buf))>0){

  out.write(buf,0,len);

}

imgis.close();

out.close();

// 向Clob写入实际内容

oracle.sql.CLOB clob=(oracle.sgl.CLOB)  user.getResume();

java.io.Writer writer = clob.getCharacterOutputStream();

writer.write("this is my  resume');

writer.close();

session.save(user);

tx.commit();

实际应用中,对于Clob字段，我们可以简单地将其映射为String类型，不过在这种情况下需要注意，Oracle Thin Driver对Clob字段支持尚有欠缺，当Clob内容超出4000字节时将无法读取，而Oracle OCI Driver(需要在本地安装Oracle客户端组件)则可以成功地完成大容量Clob字段的操作。

上面的代码中，我们通过Session.save/flush/refresh方法的组合使用，实现了上面JDBC代码中的Blob/Clob访问逻辑。

Blob/Clob 字段的Hibernate保存实现如上所述，相对来讲，读取则没有太多的障碍，之前的读取代码依然可以正常运行。

对于上面的实现，相信大家都感觉到了一些Bad Smell，如果Blob/Clob字段普遍存在，那么我们的持久层逻辑中可能遍布如此复杂的数据存储逻辑、并与数据库原生类紧密祸

如何解决这些问题?

回忆之前关于自定义数据类型的讨论。通过自定义数据类型我们可以对数据的通用特征进行抽象，那么，对于Oracle的Blob/Clob字段，我们是否可以也对其进行抽象，并以其作为所有Oracle Blob/Clob字段的映射类型?

下面的StringClobType实现了这一目标:

public class StringClobType implements UserType{

    private static final String ORACLE_DRIVER_NAME="Oracle JDBC driver";

       private static final int ORACLE_DRIVER_MAJOR_VERSION=9;

    private static final int ORACLE_DRIVER_MINOR_VERSION=0;

    public int[] sqlTypes(){

       return new int[] {Types.CLOB};

    }

    public Class returnedClass{

       return String.class;

    }

    public boolean equals(Object x, object y){

       return org.apache.commons.lang.ObjectUtils.equals(x, y);

    }

    public Object nullSafeGet(ResultSet rs, String[] names, Object owner)

    throws HibernateException,SQLException{

       Clob clob=rs.getClob(names(O]);

       return(clob==null ? null:clob.getSubString(l,  (int) clob.length())):

    }

    public void nullSafeSet(PreparedStatement st ，Object value, int index) 

    throws HibernateException, SQLException{

       DatabaseMetaData dbMetaData=st.getConnection().getMetaData();

       if (value==null)

           st.setNull(index,  sqiTypes()(0));

       else

//         本实现仅仅适用于Oracle数据序9.0以上版本

           if

           (ORACLE_DRIVER_NAME.equals(dbMetaData.getDriverName( ，))(

              if((dbMetaData.getDriverMajorVersion() 

                     >=ORACLE-DRIVER-MAJOR-VERSION)

                  &&(dbMetaData.getDriverMinorVersion()  

                     >=ORACLE-DRIVER-MINOR-VERSION)) ｛

                         try ｛

//                通过动态加载方式进免编译期对Oracle JDBC的依赖

                         Class oracleClobClass=Class.forName('oracle.sgl.CLOB");

//                              动态调用createTemporary方法

                                Class partypes[]=new Class[3];

                                partypes[0]=Connection.class;

                                partypes[1]=Boolean.TYPE;

                                partypes(2]=Integer.TYPE;

                                Method createTemporaryMethod=

                                   oracleClobClass.getDeclaredMethod(

                                          "createTemporaxy “，

                                          partypes);

                                Field durationSessionField=

                                   oracleClobClass.getField("DURATION-SESSION");

                                Object arglist[]=new 0bject[3]:

                                   Connection conn=

                                       st.getConnection().getMetaData().getConnection();

//                              数据库连接类型必须为OracleConnection

//                              莱些应用服务器会使用自带Oracle JDBC Wrapper，如Weblogic

//                              这里需要特别注意

                                Class oracleConnectionClass=

                                   Class.forName("oracle.jdbc.OracleConnection");

                                if(!oracleConnectionClass

                                       .isAssignableFrom(conn.getClass())){

                                   throw new HibernateException(

                                          "Must be a oracle.jdbc.OracleConnection:.

                                          +conn.getClass().getName());

                                }

                                arglist[0] = conn;

                                arglist(1] = Boolean.TRUE;

                                arolist[2] = durationSessionField.get(null);

                                Object tempClob =createTemporaryMethod.invoke(null,arglist);

                                partypes=new Class[l];

                                partypes[0]=Integer.TYPE;

                                Method openMethod =oracleClobClass.getDeclaredMethod("open",partypes);

                                Field modeReadWriteField =oracleClobClass.getField("MODE_READWRITE");

                                arglist = new Object(l];

                                arglis[0] = modeReadWriteField.get(null);

                                openMethod.invoke(tempClob, arglist);

                                Method getCharacterOutputStreamMethod=oracleClobClass.getDeclaredMethod("getCharacterOutputStream',null) ;

//                                     call the getCharacterOutpitStream method

                                       Writer tempClobWriter =(Writer)getCharacterOutputStreamMethod.invoke(tempClob,null);

//                                     将参数写入Clob

                                       tempClobwriter.write((String) value);

                                       tempClobWriter.flush();

                                       tempClobWriter.Close();

                                      

//                                     close  clob

                                       Method closeMethod=oracleClobClass.getDeclaredMethod("close", null);

                                       closeMethod.invoke(tempClob, null);

                                       st.setClob(index,  (Clob) tempClob);

                                )catch  (ClassNotFoundException e){

                                   throw new HibernateException("Unable to find a required class.\n"+e.getMessage()):

                                }catch (NOSuchMethodException e){

                                   throw new HibernateException("Unable to find a required method.\n"+e.getMessage()):

                                }catch (NoSuchFieldException e){

                                   throw new HibernateException("Unable to find a required field.\n"+e.getMessage());

                                }catch (IllegalAccessException e){

                                   throw new HibernateException("Unable to access a required method or field.\n"+e.getMessage());

                                   catch (InvocationTargetException e){

                                       throw new HibernateException(e.getMessage());

                                       {  catch (IOException e){

                                          throw new HibernateException(e.getMessage());

                                       }

                                       else ｛

                                       throw new HibernateException(

                                              "No CLOBS support.Use driver version"

                                              +ORACLE_DRIVER_MAJOR_VERSION

                                              +" ，minor"

                                              +ORACLE_DRIVER_MINOR_VERSION);

                                       }

                                   }else ｛

                                   String str = (String)  value;

                                    StrinaReader r = new StringReader(str);

                                   St.setCharacterStream(index, r, str.length());

                                }

    }

    public Object deepCopy(Object value){

       if(value==null)

           return null;

       return new String((String)  value);

    }

    public boolean isMutable(){

       return false

    }

}

上面这段代码，重点在于nullSafeSet方法的实现，nullSafeSet中通过Java Reflection机制，解除了编译期的Oralce JDBC原生类依赖。同时，借助Oracle JDBC提供的原生功能完成了Clob字段的写入，Clob字段的写入操作由于涉及特定数据库内部实现细节，这里就不多费唇舌，大家可参见Oracle JDBC Java Doc.

这段代码是由笔者根据Ali Ibrahim, Scott Miller的代码修改而来的(原版请参见httpJ/www.hibemate， org /56.html )，支持Oracle 9以上版本，Oracle 8对应的实现请参见上述网址。

同样的道理，读者可以根据以上例程，编写自己的ByteBlobType以实现byte[]到Blob的映射。

另外，此代码必须运行在最新版的Oracle JDBC Driver上(笔者所用版本为Oracle9i9.2.0.5 for JDK1.4,如果使用9.2.0.3或之前版本则在新建l更却删除数据时可能会遇到“nomore data read from socket”错误)。