Mybatis中SqlMapper配置的扩展与应用(1)

奋斗了好几个晚上调试程序，写了好几篇博客，终于建立起了Mybatis配置的扩展机制。虽然扩展机制是重要的，然而如果没有真正实用的扩展功能，那也至少是不那么鼓舞人心的，这篇博客就来举几个扩展的例子。

这次研读源码的起因是Oracle和MySQL数据库的兼容性，比如在Oracle中使用双竖线作为连接符，而MySQL中使用CONCAT函数；比如Oracle中可以使用DECODE函数，而MySQL中只能使用标准的CASE WHEN；又比如Oracle中可以执行DELETE FORM TABLE WHERE FIELD1 IN (SELECT FIELD1 FORM TABLE WHERE FIELD2=?)，但是MySQL中会抛出异常，等等。下面就从解决这些兼容性问题开始，首先需要在配置中添加数据库标识相关的配置：

    
<bean id="mybatisConfig" class="org.dysd.dao.mybatis.schema.SchemaConfiguration"/>
<bean id="sqlSessionFactory" p:dataSource-ref="dataSource" 
    class="org.dysd.dao.mybatis.schema.SchemaSqlSessionFactoryBean">
    
    <property name="configuration" ref="mybatisConfig"/>
    
    <property name="mapperLocations">
        <array>
            <value>classpath*:**/*.sqlmapper.xmlvalue>
        array>
    property>
    
    <property name="databaseIdProvider">
        <bean class="org.apache.ibatis.mapping.VendorDatabaseIdProvider">
            <property name="properties">
                <props>
                    
                    <prop key="MySQL">mysqlprop>
                    <prop key="oracle">oracleprop>
                    <prop key="H2">h2prop>
                    <prop key="db2">db2prop>
                props>
            property>
        bean>
    property>
bean>

 

一、连接符问题

1、编写SQL配置函数实现类

public class ConcatSqlConfigFunction extends AbstractSqlConfigFunction{//抽象父类中设定了默认的order级别

    @Override
    public String getName() {
        return "concat";
    }

    @Override
    public String eval(String databaseId, String[] args) {
        if(args.length < 2){
            Throw.throwException("the concat function require at least two arguments.");
        }
        if("mysql".equalsIgnoreCase(databaseId)){
            return "CONCAT("+Tool.STRING.join(args, ",")+")";
        }else{
            return Tool.STRING.join(args, "||");
        }
    }
}

2、在SchemaHandlers类的静态代码块中注册，或者在启动初始化类中调用SchemaHandlers的方法注册

static {
    //注册默认命名空间的StatementHandler
    register("cache-ref", new CacheRefStatementHandler());
    register("cache", new CacheStatementHandler());
    register("parameterMap", new ParameterMapStatementHandler());
    register("resultMap", new ResultMapStatementHandler());
    register("sql", new SqlStatementHandler());
    register("select|insert|update|delete", new CRUDStatementHandler());
    
    //注册默认命名空间的ScriptHandler
    register("trim", new TrimScriptHandler());
    register("where", new WhereScriptHandler());
    register("set", new SetScriptHandler());
    register("foreach", new ForEachScriptHandler());
    register("if|when", new IfScriptHandler());
    register("choose", new ChooseScriptHandler());
    //register("when", new IfScriptHandler());
    register("otherwise", new OtherwiseScriptHandler());
    register("bind", new BindScriptHandler());
    
    // 注册自定义命名空间的处理器
    registerExtend("db", new DbStatementHandler(), new DbScriptHandler());
    
    // 注册SqlConfigFunction
    register(new DecodeSqlConfigFunction());
    register(new ConcatSqlConfigFunction());
    
    // 注册SqlConfigFunctionFactory
    register(new LikeSqlConfigFunctionFactory());
}

上面代码除了注册ConcatSQLConfigFunction外，还有一些其它的注册代码，这里一并给出，下文将省略。

3、修改SqlMapper配置

<select id="selectString" resultType="string">
    select PARAM_NAME, $concat{PARAM_CODE, PARAM_NAME} AS CODE_NAME 
      from BF_PARAM_ENUM_DEF
    <if test="null != paramName and '' != paramName">
     where PARAM_NAME LIKE $CONCAT{'%', #{paramName, jdbcType=VARCHAR}, '%'}
    if>
select>

4、编写dao接口类

@Repository
public interface IExampleDao {
    
    public String selectString(@Param("paramName")String paramName);
}

5、编写测试类

@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(locations={
    "classpath:spring/applicationContext.xml"    
})
@Component
public class ExampleDaoTest {

    @Resource
    private IExampleDao dao;
    
    @Test
    public void testSelectString(){
        String a = dao.selectString("显示");
        Assert.assertEquals("显示区域", a);
    }
}

6、分别在MySQL和H2中运行如下（将mybatis日志级别调整为TRACE）

(1)MySQL

20161108 00:12:55,235 [main]-[DEBUG] ==>  Preparing: select PARAM_NAME, CONCAT(PARAM_CODE,PARAM_NAME) AS CODE_NAME from BF_PARAM_ENUM_DEF where PARAM_NAME LIKE CONCAT('%',?,'%') 
20161108 00:12:55,269 [main]-[DEBUG] ==> Parameters: 显示(String)
20161108 00:12:55,287 [main]-[TRACE] <==    Columns: PARAM_NAME, CODE_NAME
20161108 00:12:55,287 [main]-[TRACE] <==        Row: 显示区域, DISPLAY_AREA显示区域
20161108 00:12:55,289 [main]-[DEBUG] <==      Total: 1

(2)H2

20161108 00:23:08,348 [main]-[DEBUG] ==>  Preparing: select PARAM_NAME, PARAM_CODE||PARAM_NAME AS CODE_NAME from BF_PARAM_ENUM_DEF where PARAM_NAME LIKE '%'||?||'%' 
20161108 00:23:08,364 [main]-[DEBUG] ==> Parameters: 显示(String)
20161108 00:23:08,411 [main]-[TRACE] <==    Columns: PARAM_NAME, CODE_NAME
20161108 00:23:08,411 [main]-[TRACE] <==        Row: 显示区域, DISPLAY_AREA显示区域
20161108 00:23:08,411 [main]-[DEBUG] <==      Total: 1

可以看到，已经解决连接符的兼容性问题了。

另外，我们也发现，使用LIKE关键字时，写起来比较麻烦，那我们就给它一组新的SQL配置函数吧：

public class LikeSqlConfigFunctionFactory implements ISqlConfigFunctionFactory{

    @Override
    public Collection getSqlConfigFunctions() {
        return Arrays.asList(getLeftLikeSqlConfigFunction(),getRightLikeSqlConfigFunction(),getLikeSqlConfigFunction());
    }
    
    private ISqlConfigFunction getLeftLikeSqlConfigFunction(){
        return new AbstractLikeSqlConfigFunction(){
            @Override
            public String getName() {
                return "llike";
            }

            @Override
            protected String eval(String arg) {
                return "LIKE $concat{'%',"+arg+"}";
            }
        };
    }
    
    private ISqlConfigFunction getRightLikeSqlConfigFunction(){
        return new AbstractLikeSqlConfigFunction(){
            @Override
            public String getName() {
                return "rlike";
            }

            @Override
            protected String eval(String arg) {
                return "LIKE $concat{"+arg+", '%'}";
            }
        };
    }
    
    private ISqlConfigFunction getLikeSqlConfigFunction(){
        return new AbstractLikeSqlConfigFunction(){
            @Override
            public String getName() {
                return "like";
            }

            @Override
            protected String eval(String arg) {
                return "LIKE $concat{'%',"+arg+", '%'}";
            }
        };
    }

    private abstract class AbstractLikeSqlConfigFunction extends AbstractSqlConfigFunction{
        @Override
        public String eval(String databaseId, String[] args) {
            if(args.length != 1){
                Throw.throwException("the like function require one and only one argument.");
            }
            return eval(args[0]);
        }
        protected abstract String eval(String arg);
    }
}

View Code

这里，定义了一组SQL配置函数，左相似，右相似以及中间相似匹配，并且SQL配置函数还可以嵌套。于是，SqlMapper的配置文件简化为：

<select id="selectString" resultType="string">
    select PARAM_NAME, $concat{PARAM_CODE, PARAM_NAME} AS CODE_NAME 
      from BF_PARAM_ENUM_DEF
    <if test="null != paramName and '' != paramName">
     where PARAM_NAME $like{#{paramName, jdbcType=VARCHAR}}
    if>
select>

运行结果完全相同。

如果还觉得麻烦，因为PARAM_NAME和paramName是驼峰式对应，甚至还可以添加一个fieldLike函数，并将配置修改为

where  $fieldLike{#{PARAM_NAME, jdbcType=VARCHAR}}

如果再结合数据字典，jdbcType的配置也可自动生成：

where  $fieldLike{#{PARAM_NAME}}

这种情形下，如果有多个参数，也不会出现歧义（或者新定义一个配置函数$likes{}消除歧义），于是可将多个条件简化成：

where  $likes{#{PARAM_NAME, PARAM_NAME2， PARAM_NAME3}}

当然，还有更多可挖掘的简化，已经不止是兼容性的范畴了，这里就不再进一步展开了。

二、DECODE函数/CASE ... WHEN

Oracle中的DECODE函数非常方便，语法如下：

DECODE(条件,值1,返回值1,值2,返回值2,...值n,返回值n[,缺省值])

等价的标准写法：

CASE 条件
  WHEN 值1 THEN 返回值1
  WHEN 值2 THEN 返回值2
  ...
  WHEN 值n THEN 返回值n
 [ELSE 缺省值]
END

现在我们来实现一个$decode配置函数：

public class DecodeSqlConfigFunction extends AbstractSqlConfigFunction{

    @Override
    public String getName() {
        return "decode";
    }

    @Override
    public String eval(String databaseId, String[] args) {
        if(args.length < 3){
            Throw.throwException("the decode function require at least three arguments.");
        }
        if("h2".equalsIgnoreCase(databaseId)){//测试时，使用h2代替oracle，正式程序中修改为oracle
            return "DECODE("+Tool.STRING.join(args, ",")+")";
        }else{
            StringBuffer sb = new StringBuffer();
            sb.append("CASE ").append(args[0]);
            int i=2, l = args.length;
            for(; i < l; i= i+2){
                sb.append(" WHEN ").append(args[i-1]).append(" THEN ").append(args[i]);
            }
            if(i == l){//结束循环时，两者相等说明最后一个参数未使用
                sb.append(" ELSE ").append(args[l-1]);
            }
            sb.append(" END");
            return sb.toString();
        }
    }
}

然后使用SchemaHandlers注册，修改SqlMapper中配置：

<select id="selectString" resultType="string">
    select PARAM_NAME, $decode{#{paramName}, '1', 'A', '2', 'B','C'} AS DECODE_TEST 
      from BF_PARAM_ENUM_DEF
    <if test="null != paramName and '' != paramName">
     where PARAM_NAME $like{#{paramName, jdbcType=VARCHAR}}
    if>
select>

测试如下：

（1）H2中(以H2代替Oracle)

20161108 06:53:29,747 [main]-[DEBUG] ==>  Preparing: select PARAM_NAME, DECODE(?,'1','A','2','B','C') AS DECODE_TEST from BF_PARAM_ENUM_DEF where PARAM_NAME LIKE '%'||?||'%' 

（2）MySQL中

20161108 06:50:55,998 [main]-[DEBUG] ==>  Preparing: select PARAM_NAME, CASE ? WHEN '1' THEN 'A' WHEN '2' THEN 'B' ELSE 'C' END AS DECODE_TEST from BF_PARAM_ENUM_DEF where PARAM_NAME LIKE '%'||?||'%'