mybatis 源码分析(八)ResultSetHandler 详解

本篇博客就是 myabtis 系列的最后一篇了,还剩 ResultSetHandler 没有分析;作为整个 mybatis 最复杂最繁琐的部分,我不打算按步骤一次详解,因为里面的主要内容就是围绕 resultMap 按层次结构依次解析的,其中运用最多的就是反射,所以我这里将围绕延迟加载重点分析,另外本文使用的测试代码都是源码的测试案例;

一、ResultSetHandler 主体结构

public interface ResultSetHandler {
  // 负责结果集处理,完成映射返回结果对象
   List handleResultSets(Statement stmt) throws SQLException;
  // 负责游标对象处理
   Cursor handleCursorResultSets(Statement stmt) throws SQLException;
  // 负责存储过程的输出参数
  void handleOutputParameters(CallableStatement cs) throws SQLException;
}

以上就是 ResultSetHandler 的接口方法(mybatis 中只提供了唯一的实现类 DefaultResultSetHandler),在本篇博客中将主要以 handleResultSets 结果集处理作为主线分析;

在分析之前首先要清楚 handleResultSets 方法的处理流程就是围绕 resultMap 依次解析的,这里先看一个比较复杂的 resultMap 映射:


  
    
  
  
  
    
    
    
  
  
    
    
    
    
      
    
    
      
    
  
  
  

当 mybatis 初始化完成后上面的配置都放到 MappedStatement.resultMaps 里面,在解析的时候就是通过 resultMap.id 取到对应的 resultMap 然后逐次解析;

1. 嵌套查询

这里之所以说 ResultSetHandler 是整个 mybatis 里面最复杂的,主要是嵌套查询的解析(association 一对一,collection 一对多),值得注意的是这里的嵌套查询是有两种方式的:

内部嵌套


  
  
  



  
  
  
  
    
  
  
    
  

像这种嵌套查询是直接在同一个 resultMap 依次映射对应结果的,使用的是 SQL 多表连接,例如:

这里还有一种分离的内部嵌套:


  
  
  



  
  
  
  
  

这中写法只是将 association、collection 部分分离出来,其实质都是一样的,都是一条多表连接的 SQL;

外部嵌套

还有另外一种是将多表连接的 SQL 拆分,每个属性单独发一条 SQL:


  




像这样在 association、collection 中使用 select 属性指定外部 SQL,其查询结果也是发两条 SQL,这里之所以没有详细写出每个属性的映射,是因为指定了 type 和 ofType,并开启的自动映射,mybatis 在执行的时候使用反射推断出来的;

这里的两种嵌套查询在初始化的时候就进行了单独的区分:

// org.apache.ibatis.builder.xml.XMLMapperBuilder
String nestedResultMap = context.getStringAttribute("resultMap", processNestedResultMappings(context, Collections.emptyList(), resultType));

// org.apache.ibatis.submitted.nestedresulthandler.Mapper.mapper_resultMap[personResult]_collection[items]
private String processNestedResultMappings(XNode context, List resultMappings, Class enclosingType) throws Exception {
  if ("association".equals(context.getName())
      || "collection".equals(context.getName())
      || "case".equals(context.getName())) {
    if (context.getStringAttribute("select") == null) {
      validateCollection(context, enclosingType);
      ResultMap resultMap = resultMapElement(context, resultMappings, enclosingType);
      return resultMap.getId();
    }
  }
  return null;
}

// org.apache.ibatis.mapping.ResultMap
// resultMap.hasNestedResultMaps = resultMap.hasNestedResultMaps || (resultMapping.getNestedResultMapId() != null && resultMapping.getResultSet() == null);

这里代码代码比较多所以只放了关键代码,其最终结果是内部嵌套由 ResultMap.hasNestedResultMaps 标识;

// org.apache.ibatis.builder.xml.XMLMapperBuilder
private ResultMapping buildResultMappingFromContext(XNode context, Class resultType, List flags) throws Exception {
  ...
  String nestedSelect = context.getStringAttribute("select");
  String nestedResultMap = context.getStringAttribute("resultMap",
      processNestedResultMappings(context, Collections.emptyList(), resultType));
  ...
}

外部查询的最终结果是由 ResultMapping.nestedQueryId 保存的,之所以这里强调这些是因为在 ResultSetHandler 解析的时候是分了内外部嵌套两种大的情况的;

2. 多结果集

此外分析之前首先还要知道 CallableStatement 调用存储过程的时候,会有多结果集的情况,例如:

create procedure sptest.getnamesanditemsbyid(in nameId integer)
modifies sql data
dynamic result sets 2
BEGIN ATOMIC
  declare cur1 cursor for select * from sptest.names where id = nameId;
  declare cur2 cursor for select * from sptest.items where name_id in (select id from sptest.names where id = nameId);
  open cur1;
  open cur2;
END

  
  
  
  


2. 整体流程

mybatis 源码分析(八)ResultSetHandler 详解_第1张图片

上图就是 ResultSetHandler.handleResultSet 的主要流程,这里只保留了重要的部分:

  • 内外部嵌套查询的分支;
  • 外部嵌套查询与一级缓存;
  • 外部嵌套查询的延迟加载,主要是代理对象、ResultLoader、ResultLoaderMap三个对象;

其余的部分这里就不再详细分析了,一下将主要讲解外部嵌套查询的延迟加载;

二、cglib 和 javassisit 动态代理

在讲解延迟加载之前,需要首先简单了解一下动态代理,因为普通的 JavaBean 对象一般都没有实现接口,所以不能使用 java.lang.reflect.Proxy,在 mybatis 中提供了另外两种动态代理 cglib 和 javassisit;

1. cglib

public class Car {
  String name;
  
  public String getName() { return name; }
  public void setName(String name) { this.name = name; }
}

@Test
public void test() {
  Enhancer enhancer = new Enhancer();
  enhancer.setSuperclass(Car.class);
  enhancer.setCallback((MethodInterceptor) (obj, method, args, proxy) -> {
    System.out.println("代理开始");
    Object object = proxy.invokeSuper(obj, args);
    System.out.println("result: " + object);
    System.out.println("代理结束");
    return object;
  });

  Car car = (Car) enhancer.create();
  car.setName("Test");
  car.getName();
}

打印:

代理开始
result: null
代理结束
代理开始
result: Test
代理结束

2. javassist

public class Car {
  String name;
  
  public String getName() { return name; }
  public void setName(String name) { this.name = name; }
}

@Test
public void test() throws IllegalAccessException, InstantiationException {
  ProxyFactory proxyFactory = new ProxyFactory();
  proxyFactory.setSuperclass(Car.class);
  // 设置拦截目标
  proxyFactory.setFilter(m -> m.getName().startsWith("get") || m.getName().startsWith("set"));
  proxyFactory.setHandler((self, thisMethod, proceed, arg) -> {
    System.out.println("代理开始");
    Object object = proceed.invoke(self, arg);
    System.out.println("result: " + object);
    System.out.println("代理结束");
    return object;
  });

  Class clazz = proxyFactory.createClass();
  Car car = (Car) clazz.newInstance();
  car.setName("Test");
  car.getName();
}

打印:

代理开始
result: null
代理结束
代理开始
result: Test
代理结束

三、延迟加载

通过上面的讲解大家应该清楚只有外部嵌套查询才有延迟加载功能;此外和延迟加载相关的配置:

  • proxyFactory:(CGLIB | JAVASSIST-默认)指定 mybatis 延迟加载的代理工具;
  • lazyLoadingEnabled:(true | false-默认)延迟加载的全局开关。可使用 association、collection 的 fetchType (lazy|eager)属性覆盖;
  • aggressiveLazyLoading:(false| true-默认)当开启时,任何方法的调用都会加载该对象的所有属性。 否则每个属性会按需加载;

1. demo




  
  



  
  
  


public class User {
  private Integer id;
  private String name;
  private User lazy1;
  private User lazy2;
  private List lazy3;
  ...
 }

@Test
void test() {
  try (SqlSession sqlSession = sqlSessionFactory.openSession()) {
    Mapper mapper = sqlSession.getMapper(Mapper.class);
    User user = mapper.getUser(1);
    System.out.println("----getLazy1: " + user.getLazy1());
    System.out.println("----getLazy2: " + user.getLazy2());
    System.out.println("----getLazy3: " + user.getLazy3());
  }
}

打印:

DEBUG [main] - ==> Preparing: select * from users where id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - ====> Preparing: select 12 id, 'lazy2' name from (values(0))
DEBUG [main] - ====> Parameters:
DEBUG [main] - <==== Total: 1
DEBUG [main] - <== Total: 1
DEBUG [main] - ==> Preparing: select 11 id, 'lazy1' name from (values(0))
DEBUG [main] - ==> Parameters:
DEBUG [main] - <== Total: 1
----getLazy1: User{id=11, name='lazy1'}
----getLazy2: User{id=12, name='lazy2'}
DEBUG [main] - ==> Preparing: select 13 id, 'lazy3' name from (values(0))
DEBUG [main] - ==> Parameters:
DEBUG [main] - <== Total: 1
----getLazy3: [User{id=13, name='lazy3'}]

从打印的顺序可以看出当 mapper.getUser(1) 的时候,就已经获取了 user 和 lazy2,而 lazy1 和 lazy3 则是在 get 的时候才加载;这里在看一下 aggressiveLazyLoading = true 的效果:

打印:

DEBUG [main] - ==> Preparing: select * from users where id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - ====> Preparing: select 12 id, 'lazy2' name from (values(0))
DEBUG [main] - ====> Parameters:
DEBUG [main] - <==== Total: 1
DEBUG [main] - ====> Preparing: select 11 id, 'lazy1' name from (values(0))
DEBUG [main] - ====> Parameters:
DEBUG [main] - <==== Total: 1
DEBUG [main] - ====> Preparing: select 13 id, 'lazy3' name from (values(0))
DEBUG [main] - ====> Parameters:
DEBUG [main] - <==== Total: 1
DEBUG [main] - <== Total: 1
----getLazy1: User{id=11, name='lazy1'}
----getLazy2: User{id=12, name='lazy2'}
----getLazy3: [User{id=13, name='lazy3'}]

这里也能看到首先是获取 user 和 lazy2,然后在 user.getLazy1() 的时候同时加载了 lazy1 和 lazy3;

2. 创建代理

在上面已经讲过了,在使用延迟加载的时候:

  • 首先判断是否有延迟加载属性,有就使用代理包装结果集对象;
  • 然后判断一级缓存中时候有对应的外部嵌套,有就取缓存;如果没有就将外部嵌套包装为 ResultLoader 对象;
  • 然后判断外部嵌套是否需要延迟加载,如果是就将 ResultLoader 加入到 ResultLoaderMap 中,如果不需要就直接加载 resultLoader.loadResult();

创建代理:首先获取代理工厂,然后创建代理类;

private Object createResultObject(ResultSetWrapper rsw, ResultMap resultMap, ResultLoaderMap lazyLoader, String columnPrefix) throws SQLException {
  this.useConstructorMappings = false; // reset previous mapping result
  final List> constructorArgTypes = new ArrayList<>();
  final List constructorArgs = new ArrayList<>();
  Object resultObject = createResultObject(rsw, resultMap, constructorArgTypes, constructorArgs, columnPrefix);
  if (resultObject != null && !hasTypeHandlerForResultObject(rsw, resultMap.getType())) {
    final List propertyMappings = resultMap.getPropertyResultMappings();
    for (ResultMapping propertyMapping : propertyMappings) {
      if (propertyMapping.getNestedQueryId() != null && propertyMapping.isLazy()) {
        resultObject = configuration.getProxyFactory().createProxy(resultObject, lazyLoader, configuration, objectFactory, constructorArgTypes, constructorArgs);
        break;
      }
    }
  }
  this.useConstructorMappings = resultObject != null && !constructorArgTypes.isEmpty();
  return resultObject;
} 
 

3. 代理工厂

mybatis 源码分析(八)ResultSetHandler 详解_第2张图片

这里 CglibProxyFactory 和 JavassistProxyFactory 的流程都是一样的,所以我们就以 CglibProxyFactory 为例进行简单分析:

crateProxy:

static Object crateProxy(Class type, Callback callback, List> constructorArgTypes, List constructorArgs) {
  Enhancer enhancer = new Enhancer();
  enhancer.setCallback(callback);
  enhancer.setSuperclass(type);
  try {
    type.getDeclaredMethod(WRITE_REPLACE_METHOD);
    // ObjectOutputStream will call writeReplace of objects returned by writeReplace
    if (LogHolder.log.isDebugEnabled()) {
      LogHolder.log.debug(WRITE_REPLACE_METHOD + " method was found on bean " + type + ", make sure it returns this");
    }
  } catch (NoSuchMethodException e) {
    enhancer.setInterfaces(new Class[]{WriteReplaceInterface.class});
  } catch (SecurityException e) {
    // nothing to do here
  }
  Object enhanced;
  if (constructorArgTypes.isEmpty()) {
    enhanced = enhancer.create();
  } else {
    Class[] typesArray = constructorArgTypes.toArray(new Class[constructorArgTypes.size()]);
    Object[] valuesArray = constructorArgs.toArray(new Object[constructorArgs.size()]);
    enhanced = enhancer.create(typesArray, valuesArray);
  }
  return enhanced;
} 
 

这里创建大致和上面给出的 demo 差不多,都是指定父类,设置回调;接下来我们继续看拦截的具体内容:

private final ResultLoaderMap lazyLoader;
public Object intercept(Object enhanced, Method method, Object[] args, MethodProxy methodProxy) throws Throwable {
  final String methodName = method.getName();
  try {
    // 锁定 ResultLoaderMap 对象
    synchronized (lazyLoader) {
      // 创建代理的时候设置的 WriteReplaceInterface 接口
      if (WRITE_REPLACE_METHOD.equals(methodName)) {
        Object original;
        if (constructorArgTypes.isEmpty()) {
          original = objectFactory.create(type);
        } else {
          original = objectFactory.create(type, constructorArgTypes, constructorArgs);
        }
        PropertyCopier.copyBeanProperties(type, enhanced, original);
        if (lazyLoader.size() > 0) {
          return new CglibSerialStateHolder(original, lazyLoader.getProperties(), objectFactory, constructorArgTypes, constructorArgs);
        } else {
          return original;
        }
      // 真正延迟加载的逻辑处理
      } else {
        // ResultLoaderMap 数量大于 0,就表示还有待加载的属性
        if (lazyLoader.size() > 0 && !FINALIZE_METHOD.equals(methodName)) {
          // aggressive = true,或者 equals、clone、hashCode、toString 之一,就加载全部方法
          if (aggressive || lazyLoadTriggerMethods.contains(methodName)) {
            lazyLoader.loadAll();
          // 调用某属性的 set 方法时,表示不需要从数据库再加载了,所以将其移除
          } else if (PropertyNamer.isSetter(methodName)) {
            final String property = PropertyNamer.methodToProperty(methodName);
            lazyLoader.remove(property);
          // 加载单个属性
          } else if (PropertyNamer.isGetter(methodName)) {
            final String property = PropertyNamer.methodToProperty(methodName);
            if (lazyLoader.hasLoader(property)) {
              lazyLoader.load(property);
            }
          }
        }
      }
    }
    return methodProxy.invokeSuper(enhanced, args);
  } catch (Throwable t) {
    throw ExceptionUtil.unwrapThrowable(t);
  }
}

4. 延迟加载属性加载

// org.apache.ibatis.executor.loader.ResultLoaderMap
public boolean load(String property) throws SQLException {
  // 先删除 key,防止第二次查询数据库
  LoadPair pair = loaderMap.remove(property.toUpperCase(Locale.ENGLISH));
  if (pair != null) {
    // 查数据库
    pair.load();
    return true;
  }
  return false;
}

public void load(final Object userObject) throws SQLException {
  if (this.metaResultObject == null || this.resultLoader == null) {
    ...
    this.metaResultObject = config.newMetaObject(userObject);
    this.resultLoader = new ResultLoader(config, new ClosedExecutor(), ms, this.mappedParameter,
            metaResultObject.getSetterType(this.property), null, null);
  }

  /* We are using a new executor because we may be (and likely are) on a new thread
   * and executors aren't thread safe. (Is this sufficient?)
   *
   * A better approach would be making executors thread safe. */
  if (this.serializationCheck == null) {
    final ResultLoader old = this.resultLoader;
    this.resultLoader = new ResultLoader(old.configuration, new ClosedExecutor(), old.mappedStatement,
                                         old.parameterObject, old.targetType, old.cacheKey, old.boundSql);
  }
  // 查询数据库,并反射设置属性
  this.metaResultObject.setValue(property, this.resultLoader.loadResult());
}
// org.apache.ibatis.executor.loader.ResultLoader
public Object loadResult() throws SQLException {
  // 查询结果
  List list = selectList();
  // 转换结果类型
  resultObject = resultExtractor.extractObjectFromList(list, targetType);
  return resultObject;
}

// 这里又是从 Executor 出发,再查数据库了
private  List selectList() throws SQLException {
  Executor localExecutor = executor;
  if (Thread.currentThread().getId() != this.creatorThreadId || localExecutor.isClosed()) {
    localExecutor = newExecutor();
  }
  try {
    return localExecutor. query(mappedStatement, parameterObject, RowBounds.DEFAULT, Executor.NO_RESULT_HANDLER, cacheKey, boundSql);
  } finally {
    if (localExecutor != executor) {
      localExecutor.close(false);
    }
  }
} 
 

以上就是延迟加载的全部流程了,

5. 延迟加载与一级缓存

上面我们将了当一级缓存中有外部嵌套查询缓存的时候,会直接取缓存,而不是延迟加载:

private Object getNestedQueryMappingValue(ResultSet rs, MetaObject metaResultObject, ResultMapping propertyMapping, ResultLoaderMap lazyLoader, String columnPrefix)
    throws SQLException {
  final String nestedQueryId = propertyMapping.getNestedQueryId();
  final String property = propertyMapping.getProperty();
  final MappedStatement nestedQuery = configuration.getMappedStatement(nestedQueryId);
  final Class nestedQueryParameterType = nestedQuery.getParameterMap().getType();
  final Object nestedQueryParameterObject = prepareParameterForNestedQuery(rs, propertyMapping, nestedQueryParameterType, columnPrefix);
  Object value = null;
  if (nestedQueryParameterObject != null) {
    final BoundSql nestedBoundSql = nestedQuery.getBoundSql(nestedQueryParameterObject);
    final CacheKey key = executor.createCacheKey(nestedQuery, nestedQueryParameterObject, RowBounds.DEFAULT, nestedBoundSql);
    final Class targetType = propertyMapping.getJavaType();
    // 判断一级缓存
    if (executor.isCached(nestedQuery, key)) {
      executor.deferLoad(nestedQuery, metaResultObject, property, key, targetType);
      value = DEFERRED;
    } else {
      final ResultLoader resultLoader = new ResultLoader(configuration, executor, nestedQuery, nestedQueryParameterObject, targetType, key, nestedBoundSql);
      if (propertyMapping.isLazy()) {
        lazyLoader.addLoader(property, metaResultObject, resultLoader);
        value = DEFERRED;
      } else {
        value = resultLoader.loadResult();
      }
    }
  }
  return value;
}

下面我们就实验一下:





  
  
  



  
  


@Test
void test1() {
  try (SqlSession sqlSession = sqlSessionFactory.openSession()) {
    GrandFatherMapper grandFatherMapper = sqlSession.getMapper(GrandFatherMapper.class);
    FatherMapper fatherMapper = sqlSession.getMapper(FatherMapper.class);

    GrandFather grandFather = grandFatherMapper.selectById(1);
    System.out.println("----- get grandFather: " + grandFather);

    Father father = fatherMapper.selectById(1);
    System.out.println("----- get father: " + father.getName());
    System.out.println("----- get father.grandFather: " + father.getGrandFather());
  }
}

打印:

DEBUG [main] - ==> Preparing: SELECT id, name FROM GrandFather WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
----- get grandFather: GrandFather{id=1, name='John Smith sen'}
DEBUG [main] - ==> Preparing: SELECT id, name, grand_father_id FROM Father WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
----- get father: John Smith
----- get father.grandFather: GrandFather{id=1, name='John Smith sen'}

这里我们首先获取了一次 GrandFather,保证一级缓存中有,然后获取 Father,延迟加载 GrandFather;从上面的结果可以看到,确实延迟加载是从一级缓存中取的;

6. 延迟加载与二级缓存

上面我们讲过了外部嵌套查询的时候是从 Executor 开始的,那么必然有一级缓存和二级缓存;这里先说结论嵌套查询使用二级缓存一定要在同一个 namespace 里面,否则会出现脏读现象;下面举例说明:





// org/apache/ibatis/submitted/lazyload_common_property/FatherMapper.xml

  

  
    
    
    
  

  

  
    update Father set name = #{name} where id = #{id}
  
// org/apache/ibatis/submitted/lazyload_common_property/GrandFatherMapper.xml

  

  
    
    
  

  

  
    update GrandFather set name = #{name} where id = #{id}
  
@Test
void test2() {
  try (SqlSession sqlSession1 = sqlSessionFactory.openSession();
       SqlSession sqlSession2 = sqlSessionFactory.openSession();
  ) {
    GrandFatherMapper grandFatherMapper1 = sqlSession1.getMapper(GrandFatherMapper.class);
    GrandFatherMapper grandFatherMapper2 = sqlSession2.getMapper(GrandFatherMapper.class);
    FatherMapper fatherMapper1 = sqlSession1.getMapper(FatherMapper.class);
    FatherMapper fatherMapper2 = sqlSession2.getMapper(FatherMapper.class);

    Father father1 = fatherMapper1.selectById(1);
    System.out.println("----- session1 get father(put cache): " + father1);
    sqlSession1.commit();

    Father father2 = fatherMapper2.selectById(1);
    System.out.println("----- session2 get father(get cache): " + father2);

    // 测试重点
    // fatherMapper1.updateById(1, "TestName");
    grandFatherMapper1.updateById(1, "TestName");
    sqlSession1.commit();
    System.out.println("----- session1 update(put cache)");

    Father father3 = fatherMapper2.selectById(1);
    System.out.println("----- session2 get father(get cache): " + father3);
  }
}

测试流程:

  • 首先 session1 查询并提交二级缓存
  • 然后 session2 查询检查二级缓存是否生效
  • 然后 session1 修改缓存,并提交
  • 最后 session2 再查查询,看是否检查到缓存的修改

打印:

DEBUG [main] - ==> Preparing: SELECT id, name, grand_father_id FROM Father WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.GrandFatherMapper]: 0.0
DEBUG [main] - ==> Preparing: SELECT id, name FROM GrandFather WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
----- session1 get father(put cache): Father{id=1, name='John Smith', grandFather=GrandFather{id=1, name='John Smith sen'}}
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.FatherMapper]: 0.5
----- session2 get father(get cache): Father{id=1, name='John Smith', grandFather=GrandFather{id=1, name='John Smith sen'}}
DEBUG [main] - ==> Preparing: update GrandFather set name = ? where id = ?
DEBUG [main] - ==> Parameters: TestName(String), 1(Integer)
DEBUG [main] - <== Updates: 1
DEBUG [main] - Committing JDBC Connection [org.hsqldb.jdbc.JDBCConnection@2f01783a]
----- session1 update(put cache)
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.FatherMapper]: 0.6666666666666666
----- session2 get father(get cache): Father{id=1, name='John Smith', grandFather=GrandFather{id=1, name='John Smith sen'}}

注意看这里二级缓存生效了,但是出现了脏读:

然后我们将上面的注释打开:

DEBUG [main] - ==> Preparing: SELECT id, name, grand_father_id FROM Father WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.GrandFatherMapper]: 0.0
DEBUG [main] - ==> Preparing: SELECT id, name FROM GrandFather WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
----- session1 get father(put cache): Father{id=1, name='John Smith', grandFather=GrandFather{id=1, name='John Smith sen'}}
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.FatherMapper]: 0.5
----- session2 get father(get cache): Father{id=1, name='John Smith', grandFather=GrandFather{id=1, name='John Smith sen'}}
DEBUG [main] - ==> Preparing: update Father set name = ? where id = ?
DEBUG [main] - ==> Parameters: TestName(String), 1(Integer)
DEBUG [main] - <== Updates: 1
DEBUG [main] - ==> Preparing: update GrandFather set name = ? where id = ?
DEBUG [main] - ==> Parameters: TestName(String), 1(Integer)
DEBUG [main] - <== Updates: 1
DEBUG [main] - Committing JDBC Connection [org.hsqldb.jdbc.JDBCConnection@2f01783a]
----- session1 update(put cache)
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.FatherMapper]: 0.3333333333333333
DEBUG [main] - ==> Preparing: SELECT id, name, grand_father_id FROM Father WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
DEBUG [main] - Cache Hit Ratio [org.apache.ibatis.submitted.lazyload_common_property.GrandFatherMapper]: 0.0
DEBUG [main] - ==> Preparing: SELECT id, name FROM GrandFather WHERE id = ?
DEBUG [main] - ==> Parameters: 1(Integer)
DEBUG [main] - <== Total: 1
----- session2 get father(get cache): Father{id=1, name='TestName', grandFather=GrandFather{id=1, name='TestName'}}

这次发现脏读消失了??其原因就是第一次之修改了 GrandFather,虽然 Father 中有 GrandFather 属性,但是刷新缓存的时候并不会刷新 Father,所以出现的脏读;其解决办法就是使用 将缓存放在同一个命名空间内;

这里再提醒一下本文中使用的测试案例都能在 mybatis 源码的单元测试用找到;

你可能感兴趣的:(mybatis 源码分析(八)ResultSetHandler 详解)