經過源碼分析MyBatis的緩存

MyBatis緩存介紹

MyBatis支持聲明式數據緩存(declarative data caching)。當一條SQL語句被標記爲「可緩存」後,首次執行它時從數據庫獲取的全部數據會被存儲在一段高速緩存中,從此執行這條語句時就會從高速緩存中讀取結果,而不是再次命中數據庫。MyBatis提供了默認下基於Java HashMap的緩存實現,以及用於與OSCache、Ehcache、Hazelcast和Memcached鏈接的默認鏈接器。MyBatis還提供API供其餘緩存實現使用。html

重點的那句話就是:MyBatis執行SQL語句以後,這條語句就是被緩存,之後再執行這條語句的時候,會直接從緩存中拿結果,而不是再次執行SQLjava

這也就是你們常說的MyBatis一級緩存,一級緩存的做用域scope是SqlSession。node

MyBatis同時還提供了一種全局做用域global scope的緩存,這也叫作二級緩存,也稱做全局緩存。sql

一級緩存

測試

同個session進行兩次相同查詢:數據庫

1apache

2設計模式

3緩存

4session

5mybatis

6

7

8

9

10

11

12

@Test

public void test() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    try {

        User user = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user);

        User user2 = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user2);

    } finally {

        sqlSession.close();

    }

}

MyBatis只進行1次數據庫查詢:

1

2

3

4

5

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

同個session進行兩次不一樣的查詢:

1

2

3

4

5

6

7

8

9

10

11

12

@Test

public void test() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    try {

        User user = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user);

        User user2 = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 2);

        log.debug(user2);

    } finally {

        sqlSession.close();

    }

}

MyBatis進行兩次數據庫查詢:

1

2

3

4

5

6

7

8

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 2(Integer)

<==      Total: 1

User{id=2, name='FFF', age=50, birthday=Sat Dec 06 17:12:01 CST 2014}

不一樣session,進行相同查詢:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

@Test

public void test() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    SqlSession sqlSession2 = sqlSessionFactory.openSession();

    try {

        User user = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user);

        User user2 = (User)sqlSession2.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user2);

    } finally {

        sqlSession.close();

        sqlSession2.close();

    }

}

MyBatis進行了兩次數據庫查詢:

1

2

3

4

5

6

7

8

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

同個session,查詢以後更新數據,再次查詢相同的語句:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

@Test

public void test() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    try {

        User user = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user);

        user.setAge(100);

        sqlSession.update("org.format.mybatis.cache.UserMapper.update", user);

        User user2 = (User)sqlSession.selectOne("org.format.mybatis.cache.UserMapper.getById", 1);

        log.debug(user2);

        sqlSession.commit();

    } finally {

        sqlSession.close();

    }

}

更新操做以後緩存會被清除:

1

2

3

4

5

6

7

8

9

10

11

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

==>  Preparing: update USERS SET NAME = ? , AGE = ? , BIRTHDAY = ? where ID = ?

==> Parameters: format(String), 23(Integer), 2014-10-12 23:20:13.0(Timestamp), 1(Integer)

<==    Updates: 1

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

很明顯,結果驗證了一級緩存的概念,在同個SqlSession中,查詢語句相同的sql會被緩存,可是一旦執行新增或更新或刪除操做,緩存就會被清除

源碼分析

在分析MyBatis的一級緩存以前,咱們先簡單看下MyBatis中幾個重要的類和接口:

org.apache.ibatis.session.Configuration類:MyBatis全局配置信息類

org.apache.ibatis.session.SqlSessionFactory接口:操做SqlSession的工廠接口,具體的實現類是DefaultSqlSessionFactory

org.apache.ibatis.session.SqlSession接口:執行sql,管理事務的接口,具體的實現類是DefaultSqlSession

org.apache.ibatis.executor.Executor接口:sql執行器,SqlSession執行sql最終是經過該接口實現的,經常使用的實現類有SimpleExecutor和CachingExecutor,這些實現類都使用了裝飾者設計模式

一級緩存的做用域是SqlSession,那麼咱們就先看一下SqlSession的select過程:

這是DefaultSqlSession(SqlSession接口實現類,MyBatis默認使用這個類)的selectList源碼(咱們例子上使用的是selectOne方法,調用selectOne方法最終會執行selectList方法):

1

2

3

4

5

6

7

8

9

10

11

public <E> List<E> selectList(String statement, Object parameter, RowBounds rowBounds) {

    try {

      MappedStatement ms = configuration.getMappedStatement(statement);

      List<E> result = executor.query(ms, wrapCollection(parameter), rowBounds, Executor.NO_RESULT_HANDLER);

      return result;

    } catch (Exception e) {

      throw ExceptionFactory.wrapException("Error querying database.  Cause: " + e, e);

    } finally {

      ErrorContext.instance().reset();

    }

}

咱們看到SqlSession最終會調用Executor接口的方法。

接下來咱們看下DefaultSqlSession中的executor接口屬性具體是哪一個實現類。

DefaultSqlSession的構造過程(DefaultSqlSessionFactory內部):

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

private SqlSession openSessionFromDataSource(ExecutorType execType, TransactionIsolationLevel level, boolean autoCommit) {

    Transaction tx = null;

    try {

      final Environment environment = configuration.getEnvironment();

      final TransactionFactory transactionFactory = getTransactionFactoryFromEnvironment(environment);

      tx = transactionFactory.newTransaction(environment.getDataSource(), level, autoCommit);

      final Executor executor = configuration.newExecutor(tx, execType, autoCommit);

      return new DefaultSqlSession(configuration, executor);

    } catch (Exception e) {

      closeTransaction(tx); // may have fetched a connection so lets call close()

      throw ExceptionFactory.wrapException("Error opening session.  Cause: " + e, e);

    } finally {

      ErrorContext.instance().reset();

    }

}

咱們看到DefaultSqlSessionFactory構造DefaultSqlSession的時候,Executor接口的實現類是由Configuration構造的:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

public Executor newExecutor(Transaction transaction, ExecutorType executorType, boolean autoCommit) {

    executorType = executorType == null ? defaultExecutorType : executorType;

    executorType = executorType == null ? ExecutorType.SIMPLE : executorType;

    Executor executor;

    if (ExecutorType.BATCH == executorType) {

      executor = new BatchExecutor(this, transaction);

    } else if (ExecutorType.REUSE == executorType) {

      executor = new ReuseExecutor(this, transaction);

    } else {

      executor = new SimpleExecutor(this, transaction);

    }

    if (cacheEnabled) {

      executor = new CachingExecutor(executor, autoCommit);

    }

    executor = (Executor) interceptorChain.pluginAll(executor);

    return executor;

}

Executor根據ExecutorType的不一樣而建立,最經常使用的是SimpleExecutor,本文的例子也是建立這個實現類。 最後咱們發現若是cacheEnabled這個屬性爲true的話,那麼executor會被包一層裝飾器,這個裝飾器是CachingExecutor。其中cacheEnabled這個屬性是mybatis總配置文件中settings節點中cacheEnabled子節點的值,默認就是true,也就是說咱們在mybatis總配置文件中不配cacheEnabled的話,它也是默認爲打開的。

如今,問題就剩下一個了,CachingExecutor執行sql的時候到底作了什麼?

帶着這個問題,咱們繼續走下去(CachingExecutor的query方法):

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

public <E> List<E> query(MappedStatement ms, Object parameterObject, RowBounds rowBounds, ResultHandler resultHandler, CacheKey key, BoundSql boundSql) throws SQLException {

    Cache cache = ms.getCache();

    if (cache != null) {

      flushCacheIfRequired(ms);

      if (ms.isUseCache() && resultHandler == null) {

        ensureNoOutParams(ms, parameterObject, boundSql);

        if (!dirty) {

          cache.getReadWriteLock().readLock().lock();

          try {

            @SuppressWarnings("unchecked")

            List<E> cachedList = (List<E>) cache.getObject(key);

            if (cachedList != null) return cachedList;

          } finally {

            cache.getReadWriteLock().readLock().unlock();

          }

        }

        List<E> list = delegate.<E> query(ms, parameterObject, rowBounds, resultHandler, key, boundSql);

        tcm.putObject(cache, key, list); // issue #578. Query must be not synchronized to prevent deadlocks

        return list;

      }

    }

    return delegate.<E>query(ms, parameterObject, rowBounds, resultHandler, key, boundSql);

}

其中Cache cache = ms.getCache();這句代碼中,這個cache實際上就是個二級緩存,因爲咱們沒有開啓二級緩存(二級緩存的內容下面會分析),所以這裏執行了最後一句話。這裏的delegate也就是SimpleExecutor,SimpleExecutor沒有Override父類的query方法,所以最終執行了SimpleExecutor的父類BaseExecutor的query方法。

因此一級緩存最重要的代碼就是BaseExecutor的query方法!

BaseExecutor的屬性localCache是個PerpetualCache類型的實例,PerpetualCache類是實現了MyBatis的Cache緩存接口的實現類之一,內部有個Map類型的屬性用來存儲緩存數據。 這個localCache的類型在BaseExecutor內部是寫死的。 這個localCache就是一級緩存!

接下來咱們看下爲什麼執行新增或更新或刪除操做,一級緩存就會被清除這個問題。

首先MyBatis處理新增或刪除的時候,最終都是調用update方法,也就是說新增或者刪除操做在MyBatis眼裏都是一個更新操做。

咱們看下DefaultSqlSession的update方法:

1

2

3

4

5

6

7

8

9

10

11

public int update(String statement, Object parameter) {

    try {

      dirty = true;

      MappedStatement ms = configuration.getMappedStatement(statement);

      return executor.update(ms, wrapCollection(parameter));

    } catch (Exception e) {

      throw ExceptionFactory.wrapException("Error updating database.  Cause: " + e, e);

    } finally {

      ErrorContext.instance().reset();

    }

}

很明顯,這裏調用了CachingExecutor的update方法:

1

2

3

4

public int update(MappedStatement ms, Object parameterObject) throws SQLException {

    flushCacheIfRequired(ms);

    return delegate.update(ms, parameterObject);

}

這裏的flushCacheIfRequired方法清除的是二級緩存,咱們以後會分析。 CachingExecutor委託給了(以前已經分析過)SimpleExecutor的update方法,SimpleExecutor沒有Override父類BaseExecutor的update方法,所以咱們看BaseExecutor的update方法:

1

2

3

4

5

6

public int update(MappedStatement ms, Object parameter) throws SQLException {

    ErrorContext.instance().resource(ms.getResource()).activity("executing an update").object(ms.getId());

    if (closed) throw new ExecutorException("Executor was closed.");

    clearLocalCache();

    return doUpdate(ms, parameter);

}

咱們看到了關鍵的一句代碼: clearLocalCache(); 進去看看:

1

2

3

4

5

6

public void clearLocalCache() {

    if (!closed) {

      localCache.clear();

      localOutputParameterCache.clear();

    }

}

沒錯,就是這條,sqlsession沒有關閉的話,進行新增、刪除、修改操做的話就是清除一級緩存,也就是SqlSession的緩存。

二級緩存

二級緩存的做用域是全局,換句話說,二級緩存已經脫離SqlSession的控制了。

在測試二級緩存以前,我先把結論說一下:

二級緩存的做用域是全局的,二級緩存在SqlSession關閉或提交以後纔會生效。

在分析MyBatis的二級緩存以前,咱們先簡單看下MyBatis中一個關於二級緩存的類(其餘相關的類和接口以前已經分析過):

org.apache.ibatis.mapping.MappedStatement:

MappedStatement類在Mybatis框架中用於表示XML文件中一個sql語句節點,即一個<select />、<update />或者<insert />標籤。Mybatis框架在初始化階段會對XML配置文件進行讀取,將其中的sql語句節點對象化爲一個個MappedStatement對象。

配置

二級緩存跟一級緩存不一樣,一級緩存不須要配置任何東西,且默認打開。 二級緩存就須要配置一些東西。

本文就說下最簡單的配置,在mapper文件上加上這句配置便可:

1

<cache/>

其實二級緩存跟3個配置有關:

  1. mybatis全局配置文件中的setting中的cacheEnabled須要爲true(默認爲true,不設置也行)
  2. mapper配置文件中須要加入<cache>節點
  3. mapper配置文件中的select節點須要加上屬性useCache須要爲true(默認爲true,不設置也行)

測試

不一樣SqlSession,查詢相同語句,第一次查詢以後commit SqlSession:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

@Test

public void testCache2() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    SqlSession sqlSession2 = sqlSessionFactory.openSession();

    try {

        String sql = "org.format.mybatis.cache.UserMapper.getById";

        User user = (User)sqlSession.selectOne(sql, 1);

        log.debug(user);

        // 注意,這裏必定要提交。 不提交仍是會查詢兩次數據庫

        sqlSession.commit();

        User user2 = (User)sqlSession2.selectOne(sql, 1);

        log.debug(user2);

    } finally {

        sqlSession.close();

        sqlSession2.close();

    }

}

MyBatis僅進行了一次數據庫查詢:

1

2

3

4

5

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

不一樣SqlSession,查詢相同語句,第一次查詢以後close SqlSession:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

@Test

public void testCache2() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    SqlSession sqlSession2 = sqlSessionFactory.openSession();

    try {

        String sql = "org.format.mybatis.cache.UserMapper.getById";

        User user = (User)sqlSession.selectOne(sql, 1);

        log.debug(user);

        sqlSession.close();

        User user2 = (User)sqlSession2.selectOne(sql, 1);

        log.debug(user2);

    } finally {

        sqlSession2.close();

    }

}

MyBatis僅進行了一次數據庫查詢:

1

2

3

4

5

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

不一樣SqlSesson,查詢相同語句。 第一次查詢以後SqlSession不提交:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

@Test

public void testCache2() {

    SqlSession sqlSession = sqlSessionFactory.openSession();

    SqlSession sqlSession2 = sqlSessionFactory.openSession();

    try {

        String sql = "org.format.mybatis.cache.UserMapper.getById";

        User user = (User)sqlSession.selectOne(sql, 1);

        log.debug(user);

        User user2 = (User)sqlSession2.selectOne(sql, 1);

        log.debug(user2);

    } finally {

        sqlSession.close();

        sqlSession2.close();

    }

}

MyBatis執行了兩次數據庫查詢:

1

2

3

4

5

6

7

8

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

==>  Preparing: select * from USERS WHERE ID = ?

==> Parameters: 1(Integer)

<==      Total: 1

User{id=1, name='format', age=23, birthday=Sun Oct 12 23:20:13 CST 2014}

源碼分析

咱們從在mapper文件中加入的<cache/>中開始分析源碼,關於MyBatis的SQL解析請參考另一篇博客Mybatis解析動態sql原理分析。接下來咱們看下這個cache的解析:

XMLMappedBuilder(解析每一個mapper配置文件的解析類,每個mapper配置都會實例化一個XMLMapperBuilder類)的解析方法:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

private void configurationElement(XNode context) {

    try {

      String namespace = context.getStringAttribute("namespace");

      if (namespace.equals("")) {

          throw new BuilderException("Mapper's namespace cannot be empty");

      }

      builderAssistant.setCurrentNamespace(namespace);

      cacheRefElement(context.evalNode("cache-ref"));

      cacheElement(context.evalNode("cache"));

      parameterMapElement(context.evalNodes("/mapper/parameterMap"));

      resultMapElements(context.evalNodes("/mapper/resultMap"));

      sqlElement(context.evalNodes("/mapper/sql"));

      buildStatementFromContext(context.evalNodes("select|insert|update|delete"));

    } catch (Exception e) {

      throw new BuilderException("Error parsing Mapper XML. Cause: " + e, e);

    }

}

咱們看到了解析cache的那段代碼:

1

2

3

4

5

6

7

8

9

10

11

12

13

private void cacheElement(XNode context) throws Exception {

    if (context != null) {

      String type = context.getStringAttribute("type", "PERPETUAL");

      Class<? extends Cache> typeClass = typeAliasRegistry.resolveAlias(type);

      String eviction = context.getStringAttribute("eviction", "LRU");

      Class<? extends Cache> evictionClass = typeAliasRegistry.resolveAlias(eviction);

      Long flushInterval = context.getLongAttribute("flushInterval");

      Integer size = context.getIntAttribute("size");

      boolean readWrite = !context.getBooleanAttribute("readOnly", false);

      Properties props = context.getChildrenAsProperties();

      builderAssistant.useNewCache(typeClass, evictionClass, flushInterval, size, readWrite, props);

    }

}

解析完cache標籤以後會使用builderAssistant的userNewCache方法,這裏的builderAssistant是一個MapperBuilderAssistant類型的幫助類,每一個XMLMappedBuilder構造的時候都會實例化這個屬性,MapperBuilderAssistant類內部有個Cache類型的currentCache屬性,這個屬性也就是mapper配置文件中cache節點所表明的值:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

public Cache useNewCache(Class<? extends Cache> typeClass,

  Class<? extends Cache> evictionClass,

  Long flushInterval,

  Integer size,

  boolean readWrite,

  Properties props) {

    typeClass = valueOrDefault(typeClass, PerpetualCache.class);

    evictionClass = valueOrDefault(evictionClass, LruCache.class);

    Cache cache = new CacheBuilder(currentNamespace)

        .implementation(typeClass)

        .addDecorator(evictionClass)

        .clearInterval(flushInterval)

        .size(size)

        .readWrite(readWrite)

        .properties(props)

        .build();

    configuration.addCache(cache);

    currentCache = cache;

    return cache;

}

ok,如今mapper配置文件中的cache節點被解析到了XMLMapperBuilder實例中的builderAssistant屬性中的currentCache值裏。

接下來XMLMapperBuilder會解析select節點,解析select節點的時候使用XMLStatementBuilder進行解析(也包括其餘insert,update,delete節點):

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

public void parseStatementNode() {

    String id = context.getStringAttribute("id");

    String databaseId = context.getStringAttribute("databaseId");

 

    if (!databaseIdMatchesCurrent(id, databaseId, this.requiredDatabaseId)) return;

 

    Integer fetchSize = context.getIntAttribute("fetchSize");

    Integer timeout = context.getIntAttribute("timeout");

    String parameterMap = context.getStringAttribute("parameterMap");

    String parameterType = context.getStringAttribute("parameterType");

    Class<?> parameterTypeClass = resolveClass(parameterType);

    String resultMap = context.getStringAttribute("resultMap");

    String resultType = context.getStringAttribute("resultType");

    String lang = context.getStringAttribute("lang");

    LanguageDriver langDriver = getLanguageDriver(lang);

 

    Class<?> resultTypeClass = resolveClass(resultType);

    String resultSetType = context.getStringAttribute("resultSetType");

    StatementType statementType = StatementType.valueOf(context.getStringAttribute("statementType", StatementType.PREPARED.toString()));

    ResultSetType resultSetTypeEnum = resolveResultSetType(resultSetType);

 

    String nodeName = context.getNode().getNodeName();

    SqlCommandType sqlCommandType = SqlCommandType.valueOf(nodeName.toUpperCase(Locale.ENGLISH));

    boolean isSelect = sqlCommandType == SqlCommandType.SELECT;

    boolean flushCache = context.getBooleanAttribute("flushCache", !isSelect);

    boolean useCache = context.getBooleanAttribute("useCache", isSelect);

    boolean resultOrdered = context.getBooleanAttribute("resultOrdered", false);

 

    // Include Fragments before parsing

    XMLIncludeTransformer includeParser = new XMLIncludeTransformer(configuration, builderAssistant);

    includeParser.applyIncludes(context.getNode());

 

    // Parse selectKey after includes and remove them.

    processSelectKeyNodes(id, parameterTypeClass, langDriver);

 

    // Parse the SQL (pre: <selectKey> and <include> were parsed and removed)

    SqlSource sqlSource = langDriver.createSqlSource(configuration, context, parameterTypeClass);

    String resultSets = context.getStringAttribute("resultSets");

    String keyProperty = context.getStringAttribute("keyProperty");

    String keyColumn = context.getStringAttribute("keyColumn");

    KeyGenerator keyGenerator;

    String keyStatementId = id + SelectKeyGenerator.SELECT_KEY_SUFFIX;

    keyStatementId = builderAssistant.applyCurrentNamespace(keyStatementId, true);

    if (configuration.hasKeyGenerator(keyStatementId)) {

      keyGenerator = configuration.getKeyGenerator(keyStatementId);

    } else {

      keyGenerator = context.getBooleanAttribute("useGeneratedKeys",

          configuration.isUseGeneratedKeys() && SqlCommandType.INSERT.equals(sqlCommandType))

          ? new Jdbc3KeyGenerator() : new NoKeyGenerator();

    }

 

    builderAssistant.addMappedStatement(id, sqlSource, statementType, sqlCommandType,

        fetchSize, timeout, parameterMap, parameterTypeClass, resultMap, resultTypeClass,

        resultSetTypeEnum, flushCache, useCache, resultOrdered,

        keyGenerator, keyProperty, keyColumn, databaseId, langDriver, resultSets);

}

這段代碼前面都是解析一些標籤的屬性,咱們看到了最後一行使用builderAssistant添加MappedStatement,其中builderAssistant屬性是構造XMLStatementBuilder的時候經過XMLMappedBuilder傳入的,咱們繼續看builderAssistant的addMappedStatement方法:

進入setStatementCache:

1

2

3

4

5

6

7

8

9

10

11

12

private void setStatementCache(

  boolean isSelect,

  boolean flushCache,

  boolean useCache,

  Cache cache,

  MappedStatement.Builder statementBuilder) {

    flushCache = valueOrDefault(flushCache, !isSelect);

    useCache = valueOrDefault(useCache, isSelect);

    statementBuilder.flushCacheRequired(flushCache);

    statementBuilder.useCache(useCache);

    statementBuilder.cache(cache);

}

最終mapper配置文件中的<cache/>被設置到了XMLMapperBuilder的builderAssistant屬性中,XMLMapperBuilder中使用XMLStatementBuilder遍歷CRUD節點,遍歷CRUD節點的時候將這個cache節點設置到這些CRUD節點中,這個cache就是所謂的二級緩存!

接下來咱們回過頭來看查詢的源碼,CachingExecutor的query方法:

進入TransactionalCacheManager的putObject方法:

1

2

3

4

5

6

7

8

9

10

11

12

public void putObject(Cache cache, CacheKey key, Object value) {

    getTransactionalCache(cache).putObject(key, value);

}

 

private TransactionalCache getTransactionalCache(Cache cache) {

    TransactionalCache txCache = transactionalCaches.get(cache);

    if (txCache == null) {

      txCache = new TransactionalCache(cache);

      transactionalCaches.put(cache, txCache);

    }

    return txCache;

}

TransactionalCache的putObject方法:

1

2

3

4

public void putObject(Object key, Object object) {

    entriesToRemoveOnCommit.remove(key);

    entriesToAddOnCommit.put(key, new AddEntry(delegate, key, object));

}

咱們看到,數據被加入到了entriesToAddOnCommit中,這個entriesToAddOnCommit是什麼東西呢,它是TransactionalCache的一個Map屬性:

1

private Map<Object, AddEntry> entriesToAddOnCommit;

AddEntry是TransactionalCache內部的一個類:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

private static class AddEntry {

    private Cache cache;

    private Object key;

    private Object value;

 

    public AddEntry(Cache cache, Object key, Object value) {

      this.cache = cache;

      this.key = key;

      this.value = value;

    }

 

    public void commit() {

      cache.putObject(key, value);

    }

}

好了,如今咱們發現使用二級緩存以後:查詢數據的話,先從二級緩存中拿數據,若是沒有的話,去一級緩存中拿,一級緩存也沒有的話再查詢數據庫。有了數據以後在丟到TransactionalCache這個對象的entriesToAddOnCommit屬性中。

接下來咱們來驗證爲何SqlSession commit或close以後,二級緩存纔會生效這個問題。

DefaultSqlSession的commit方法:

1

2

3

4

5

6

7

8

9

10

public void commit(boolean force) {

    try {

      executor.commit(isCommitOrRollbackRequired(force));

      dirty = false;

    } catch (Exception e) {

      throw ExceptionFactory.wrapException("Error committing transaction.  Cause: " + e, e);

    } finally {

      ErrorContext.instance().reset();

    }

}

CachingExecutor的commit方法:

1

2

3

4

5

public void commit(boolean required) throws SQLException {

    delegate.commit(required);

    tcm.commit();

    dirty = false;

}

tcm.commit即 TransactionalCacheManager的commit方法:

1

2

3

4

5

public void commit() {

    for (TransactionalCache txCache : transactionalCaches.values()) {

      txCache.commit();

    }

}

TransactionalCache的commit方法:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

public void commit() {

    delegate.getReadWriteLock().writeLock().lock();

    try {

      if (clearOnCommit) {

        delegate.clear();

      } else {

        for (RemoveEntry entry : entriesToRemoveOnCommit.values()) {

          entry.commit();

        }

      }

      for (AddEntry entry : entriesToAddOnCommit.values()) {

        entry.commit();

      }

      reset();

    } finally {

      delegate.getReadWriteLock().writeLock().unlock();

    }

}

發現調用了AddEntry的commit方法:

1

2

3

public void commit() {

  cache.putObject(key, value);

}

發現了! AddEntry的commit方法會把數據丟到cache中,也就是丟到二級緩存中!

關於爲什麼調用close方法後,二級緩存纔會生效,由於close方法內部會調用commit方法。本文就不具體說了。 讀者有興趣的話看一看源碼就知道爲何了。

其餘

Cache接口簡介

org.apache.ibatis.cache.Cache是MyBatis的緩存接口,想要實現自定義的緩存須要實現這個接口。

MyBatis中關於Cache接口的實現類也使用了裝飾者設計模式。

咱們看下它的一些實現類

簡單說明:

LRU – 最近最少使用的:移除最長時間不被使用的對象。

FIFO – 先進先出:按對象進入緩存的順序來移除它們。

SOFT – 軟引用:移除基於垃圾回收器狀態和軟引用規則的對象。

WEAK – 弱引用:更積極地移除基於垃圾收集器狀態和弱引用規則的對象。

1

2

3

4

5

<cache

  eviction="FIFO"

  flushInterval="60000"

  size="512"

  readOnly="true"/>

能夠經過cache節點的eviction屬性設置,也能夠設置其餘的屬性。

cache-ref節點

mapper配置文件中還能夠加入cache-ref節點,它有個屬性namespace。

若是每一個mapper文件都是用cache-ref,且namespace都同樣,那麼就表明着真正意義上的全局緩存。

若是隻用了cache節點,那僅表明這個這個mapper內部的查詢被緩存了,其餘mapper文件的不起做用,這並非所謂的全局緩存。

總結

整體來講,MyBatis的源碼看起來仍是比較輕鬆的,本文從實踐和源碼方面深刻分析了MyBatis的緩存原理,但願對讀者有幫助

歡迎學Java和大數據的朋友們加入java架構交流: 855835163
加羣連接:https://jq.qq.com/?_wv=1027&k=5dPqXGI 羣內提供免費的架構資料還有:Java工程化、高性能及分佈式、高性能、深刻淺出。高架構。性能調優、Spring,MyBatis,Netty源碼分析和大數據等多個知識點高級進階乾貨的免費直播講解  能夠進來一塊兒學習交流哦

相關文章
相關標籤/搜索