Calling stored procedures in Java (details)

This article explains how to use DBMS stored procedures. I covered basic and advanced features of using stored procedures, such as returning a ResultSet. This article assumes that you are already very familiar with DBMS and JDBC, and also assumes that you can read codes written in other languages ​​​​(that is, languages ​​other than Java) without any obstacles. However, it does not require you to have any experience in stored procedure programming.
Stored procedures refer to programs that are saved in the database and executed on the database side. You can call stored procedures from Java classes using special syntax. When calling, the name of the stored procedure and the specified parameters are sent to the DBMS through the JDBC connection, the stored procedure is executed and the results are returned through the connection (if any).
Using stored procedures has the same benefits as using an application server based on EJB or CORBA. The difference is that stored procedures are available for free from many popular DBMSs, while application servers are mostly very expensive. It's not just about license fees. The management and coding costs of using an application server, as well as the added complexity of client programs, can all be replaced by stored procedures in a DBMS.
You can write stored procedures in Java, Python, Perl, or C, but generally use the specific language specified by your DBMS. Oracle uses PL/SQL, PostgreSQL uses pl/pgsql, and DB2 uses Procedural SQL. These languages ​​are all very similar. Porting stored procedures between them is no more difficult than porting Session Beans between different implementations of Sun's EJB specification. Moreover, stored procedures are designed for embedding SQL, which makes them a more friendly way to express the database mechanism than languages ​​such as Java or C.
Because stored procedures run in the DBMS itself, this can help reduce waiting time in the application. Instead of executing 4 or 5 SQL statements in Java code, only 1 stored procedure needs to be executed on the server side. Reducing the number of data round-trips on the network can dramatically improve performance.

Using stored procedures

Simple old JDBC supports the calling of stored procedures through the CallableStatement class. This class is actually a subclass of PreparedStatement. Suppose we have a poets database. There is a stored procedure in the database that sets the poet's death age. The following is the detailed code for calling old soak Dylan Thomas (old soak Dylan Thomas, do not specify whether it is related to allusions or culture, please criticize and correct. Translation):

try{
       int age = 39;
      String poetName = "dylan thomas";
      CallableStatement proc = connection.prepareCall("{ call set_death_age(?, ?) }");
      proc.setString(1, poetName);
      proc.setInt(2, age);
      cs.execute();
}catch (SQLException e){ // ....}

The string passed to the prepareCall method is the writing specification for stored procedure calls. It specifies the name of the stored procedure,? Represents the parameters you need to specify.
Integration with JDBC is a great convenience for stored procedures: in order to call stored procedures from an application, no stub classes or configuration files are needed, nothing is required except the JDBC driver for your DBMS.
When this code is executed, the database stored procedure is called. We did not get the result because the stored procedure does not return a result. Success or failure of execution will be known through exceptions. Failure may mean a failure in calling the stored procedure (such as supplying a parameter of an incorrect type), or an application failure (such as throwing an exception indicating that "Dylan Thomas" does not exist in the poets database)

Combining SQL operations with stored procedures

Mapping Java objects to rows in a SQL table is quite simple, but it usually requires executing several SQL statements; perhaps a SELECT to find the ID, and then an INSERT to insert the specified ID. data. In a highly normalized database schema, updates to multiple tables may be required, thus requiring more statements. Java code can grow quickly, and the network overhead of each statement can quickly increase.
Moving these SQL statements into a stored procedure will greatly simplify the code, involving only one network call. All associated SQL operations can occur inside the database. Also, stored procedure languages, such as PL/SQL, allow the use of SQL syntax, which is more natural than Java code. The following is our early stored procedure, written in Oracle's PL/SQL language:

create procedure set_death_age(poet VARCHAR2, poet_age NUMBER)
      poet_id NUMBER;
      begin SELECT id INTO poet_id FROM poets WHERE name = poet;
      INSERT INTO deaths (mort_id, age) VALUES (poet_id, poet_age);
end set_death_age;

Is it unique? No. I bet you were expecting to see an UPDATE on the poets table. This also hints at how easy it is to implement using stored procedures. set_death_age is almost certainly a bad implementation. We should add a column to the poets table to store the death age. The Java code doesn't care how the database schema is implemented, because it only calls stored procedures. We can change the database schema later to improve performance, but we don't have to modify our code.
The following is the Java code that calls the above stored procedure:

public static void setDeathAge(Poet dyingBard, int age) throws SQLException{
      Connection con = null;
      CallableStatement proc = null;
      try {
           con = connectionPool.getConnection();
           proc = con.prepareCall("{ call set_death_age(?, ?) }");
           proc.setString(1, dyingBard.getName());
            proc.setInt(2, age);
            proc.execute();
}
     finally {  
           try { proc.close(); }
           catch (SQLException e) {}  
            con.close();
     }
}

为了确保可维护性，建议使用像这儿这样的static方法。这也使得调用存储过程的代码集中在一个简单的模版代码中。如果你用到许多存储过程，就会发现仅需要拷贝、粘贴就可以创建新的方法。因为代码的模版化，甚至也可以通过脚本自动生产调用存储过程的代码。 

Functions 

存储过程可以有返回值，所以CallableStatement类有类似getResultSet这样的方法来获取返回值。当存储过程返回一个值时，你必须使用registerOutParameter方法告诉JDBC驱动器该值的SQL类型是什么。你也必须调整存储过程调用来指示该过程返回一个值。 
下面接着上面的例子。这次我们查询Dylan Thomas逝世时的年龄。这次的存储过程使用PostgreSQL的pl/pgsql： 

create function snuffed_it_when (VARCHAR) returns integer 'declare
                 poet_id NUMBER;
                 poet_age NUMBER;
begin
--first get the id associated with the poet.
               SELECT id INTO poet_id FROM poets WHERE name = $1;
--get and return the age.
                 SELECT age INTO poet_age FROM deaths WHERE mort_id = poet_id;
return age;
end;' language 'pl/pgsql';

另外，注意pl/pgsql参数名通过Unix和DOS脚本的$n语法引用。同时，也注意嵌入的注释，这是和Java代码相比的另一个优越性。在Java中写这样的注释当然是可以的，但是看起来很凌乱，并且和SQL语句脱节，必须嵌入到Java String中。 
下面是调用这个存储过程的Java代码： 

connection.setAutoCommit(false);
CallableStatement proc = connection.prepareCall("{ ? = call snuffed_it_when(?) }");
proc.registerOutParameter(1, Types.INTEGER);
proc.setString(2, poetName);
cs.execute();
int age = proc.getInt(2);

如果指定了错误的返回值类型会怎样？那么，当调用存储过程时将抛出一个RuntimeException，正如你在ResultSet操作中使用了一个错误的类型所碰到的一样。 

复杂的返回值 

关于存储过程的知识，很多人好像就熟悉我们所讨论的这些。如果这是存储过程的全部功能，那么存储过程就不是其它远程执行机制的替换方案了。存储过程的功能比这强大得多。 
当你执行一个SQL查询时，DBMS创建一个叫做cursor（游标）的数据库对象，用于在返回结果中迭代每一行。ResultSet是当前时间点的游标的一个表示。这就是为什么没有缓存或者特定数据库的支持，你只能在ResultSet中向前移动。 
某些DBMS允许从存储过程中返回游标的一个引用。JDBC并不支持这个功能，但是Oracle、PostgreSQL和DB2的JDBC驱动器都支持在ResultSet上打开到游标的指针（pointer）。 
设想列出所有没有活到退休年龄的诗人，下面是完成这个功能的存储过程，返回一个打开的游标，同样也使用PostgreSQL的pl/pgsql语言： 

create procedure list_early_deaths () return refcursor as 'declare
      toesup refcursor;
begin
      open toesup for SELECT poets.name, deaths.age FROM poets, deaths -- all entries in deaths are for poets. -- but the table might become generic.
      WHERE poets.id = deaths.mort_id AND deaths.age < 60;
      return toesup;
end;&#39; language &#39;plpgsql&#39;;

下面是调用该存储过程的Java方法，将结果输出到PrintWriter： 
PrintWriter: 

static void sendEarlyDeaths(PrintWriter out){
      Connection con = null;
      CallableStatement toesUp = null;
      try {
          con = ConnectionPool.getConnection();
           // PostgreSQL needs a transaction to do this... con.
          setAutoCommit(false); // Setup the call.
          CallableStatement toesUp = connection.prepareCall("{ ? = call list_early_deaths () }");
           toesUp.registerOutParameter(1, Types.OTHER);
           toesUp.execute();
           ResultSet rs = (ResultSet) toesUp.getObject(1);
           while (rs.next()) {
                  String name = rs.getString(1);
                  int age = rs.getInt(2);
                  out.println(name + " was " + age + " years old.");
            }
            rs.close();
         }  
       catch (SQLException e) { // We should protect these calls. toesUp.close(); con.close();
      }
}

因为JDBC并不直接支持从存储过程中返回游标，我们使用Types.OTHER来指示存储过程的返回类型，然后调用getObject()方法并对返回值进行强制类型转换。 
这个调用存储过程的Java方法是mapping的一个好例子。Mapping是对一个集上的操作进行抽象的方法。不是在这个过程上返回一个集，我们可以把操作传送进去执行。本例中，操作就是把ResultSet打印到一个输出流。这是一个值得举例的很常用的例子，下面是调用同一个存储过程的另外一个方法实现： 

public class ProcessPoetDeaths{
      public abstract void sendDeath(String name, int age);
}
      static void mapEarlyDeaths(ProcessPoetDeaths mapper){
      Connection con = null;
      CallableStatement toesUp = null;
      try {
           con = ConnectionPool.getConnection();
           con.setAutoCommit(false);
           CallableStatement toesUp = connection.prepareCall("{ ? = call list_early_deaths () }");
            toesUp.registerOutParameter(1, Types.OTHER);
            toesUp.execute();
            ResultSet rs = (ResultSet) toesUp.getObject(1);
           while (rs.next()) {
           String name = rs.getString(1);
           int age = rs.getInt(2);
          mapper.sendDeath(name, age);
}
rs.close();
} catch (SQLException e) { // We should protect these calls. toesUp.close();
con.close();
}
}

这允许在ResultSet数据上执行任意的处理，而不需要改变或者复制获取ResultSet的方法： 

static void sendEarlyDeaths(final PrintWriter out){
            ProcessPoetDeaths myMapper = new ProcessPoetDeaths() {
                                public void sendDeath(String name, int age) {
                                                    out.println(name + " was " + age + " years old.");
                               }
           };
mapEarlyDeaths(myMapper);
}

这个方法使用ProcessPoetDeaths的一个匿名实例调用mapEarlyDeaths。该实例拥有sendDeath方法的一个实现，和我们上面的例子一样的方式把结果写入到输出流。当然，这个技巧并不是存储过程特有的，但是和存储过程中返回的ResultSet结合使用，是一个非常强大的工具。 

结论 

存储过程可以帮助你在代码中分离逻辑，这基本上总是有益的。这个分离的好处有： 
• 快速创建应用，使用和应用一起改变和改善的数据库模式。  
• 数据库模式可以在以后改变而不影响Java对象，当我们完成应用后，可以重新设计更好的模式。 
• 存储过程通过更好的SQL嵌入使得复杂的SQL更容易理解。 
• 编写存储过程比在Java中编写嵌入的SQL拥有更好的工具－－大部分编辑器都提供语法高亮！ 
• 存储过程可以在任何SQL命令行中测试，这使得调试更加容易。 

并不是所有的数据库都支持存储过程，但是存在许多很棒的实现，包括免费/开源的和非免费的，所以移植并不是一个问题。Oracle、PostgreSQL和DB2都有类似的存储过程语言，并且有在线的社区很好地支持。 
存储过程工具很多，有像TOAD或TORA这样的编辑器、调试器和IDE，提供了编写、维护PL/SQL或pl/pgsql的强大的环境。 
存储过程确实增加了你的代码的开销，但是它们和大多数的应用服务器相比，开销小得多。如果你的代码复杂到需要使用DBMS，我建议整个采用存储过程的方式。

以上就是在Java中调用存储过程(详细)的内容，更多相关内容请关注PHP中文网（www.php.cn）！