PS：該篇內(nèi)容基于 mybatis 3.4.4 版本 趟妥， 數(shù)據(jù)庫(kù)基于 mysql 5.6

Executor 類圖

主要分析 SimpleExecutor , ReuseExecutor , BatchExecutor , CachingExecutor 的不同點(diǎn)朽基，以及實(shí)際使用中該如何選擇疯搅。

image.png

BaseExecutor

BaseExecutor 主要是使用了模板設(shè)計(jì)模式（template）, 共性被封裝在 BaseExecutor 中 , 容易變化的內(nèi)容被分離到了子類中 。

SimpleExecutor

MyBatis 的官方文檔中對(duì) SimpleExecutor 的說(shuō)明是 "普通的執(zhí)行器" , 普通就在于每一次執(zhí)行都會(huì)創(chuàng)建一個(gè)新的 Statement 對(duì)象 。下面看一下 mybatis 創(chuàng)建 Statement 對(duì)象的代碼 ：

// BaseStatementHandler 中的方法
  @Override
  public Statement prepare(Connection connection, Integer transactionTimeout) throws SQLException {
    ErrorContext.instance().sql(boundSql.getSql());
    Statement statement = null;
    try {
      statement = instantiateStatement(connection);
      setStatementTimeout(statement, transactionTimeout);
      setFetchSize(statement);
      return statement;
    } catch (SQLException e) {
      closeStatement(statement);
      throw e;
    } catch (Exception e) {
      closeStatement(statement);
      throw new ExecutorException("Error preparing statement.  Cause: " + e, e);
    }
  }
  
// PreparedStatementHandler 中的方法
  @Override
  protected Statement instantiateStatement(Connection connection) throws SQLException {
    String sql = boundSql.getSql();
    if (mappedStatement.getKeyGenerator() instanceof Jdbc3KeyGenerator) {
      String[] keyColumnNames = mappedStatement.getKeyColumns();
      if (keyColumnNames == null) {
        return connection.prepareStatement(sql, PreparedStatement.RETURN_GENERATED_KEYS);
      } else {
        return connection.prepareStatement(sql, keyColumnNames);
      }
    } else if (mappedStatement.getResultSetType() != null) {
      return connection.prepareStatement(sql, mappedStatement.getResultSetType().getValue(), ResultSet.CONCUR_READ_ONLY);
    } else {
      return connection.prepareStatement(sql);
    }
  }

在眼見為實(shí)了 mybatis 創(chuàng)建 Statement 對(duì)象的代碼后我們可以看SimpleExecutor 的代碼 ， 每次調(diào)用時(shí)都創(chuàng)建了一個(gè)新的 Statement 對(duì)象：

 // 實(shí)現(xiàn) BaseExecutor 中的抽象方法
  @Override
  public <E> List<E> doQuery(MappedStatement ms, Object parameter, RowBounds rowBounds, ResultHandler resultHandler, BoundSql boundSql) throws SQLException {
    Statement stmt = null;
    try {
      Configuration configuration = ms.getConfiguration();
      StatementHandler handler = configuration.newStatementHandler(wrapper, ms, parameter, rowBounds, resultHandler, boundSql);
      stmt = prepareStatement(handler, ms.getStatementLog());
      return handler.<E>query(stmt, resultHandler);
    } finally {
      closeStatement(stmt);
    }
  }

  // 獲取一個(gè) Statement ， 這里的 handler 默認(rèn)使用的是 
 // org.apache.ibatis.executor.statement.PreparedStatementHandler
  private Statement prepareStatement(StatementHandler handler, Log statementLog) throws SQLException {
      Statement stmt;
      Connection connection = getConnection(statementLog);
      stmt = handler.prepare(connection, transaction.getTimeout());
      handler.parameterize(stmt);
      return stmt;
    }

ReuseExecutor

官方文檔中的解釋是“執(zhí)行器會(huì)重用預(yù)處理語(yǔ)句（prepared statements）” 涛救， 這次倒是解釋的很詳細(xì)。也就是說(shuō)不會(huì)每一次調(diào)用都去創(chuàng)建一個(gè) Statement 對(duì)象 业扒， 而是會(huì)重復(fù)利用以前創(chuàng)建好的（如果SQL相同的話）检吆，這也就是在很多數(shù)據(jù)連接池庫(kù)中常見的 PSCache 概念 。但是ReuseExecutor的PSCache 范圍只能存在于一次回話中 程储， 因?yàn)槊恳淮位卦拑?nèi)部都會(huì)使用一個(gè)新的 ReuseExecutor 對(duì)象 蹭沛， 所以 mybatis 的 PSCache 作用十分有限。

  // 存儲(chǔ) SQL 語(yǔ)句對(duì)應(yīng)的 Statement 對(duì)象
  private final Map<String, Statement> statementMap = new HashMap<String, Statement>();

  // .....

  private Statement prepareStatement(StatementHandler handler, Log statementLog) throws SQLException {
    Statement stmt;
    BoundSql boundSql = handler.getBoundSql();
    String sql = boundSql.getSql();
    // 檢查該 sql 是否有可用的 Statement 章鲤， 如果有的話直接從緩存中獲取 摊灭， 沒有的話創(chuàng)建新的 Statement 并緩存
    if (hasStatementFor(sql)) {
      stmt = getStatement(sql);
      applyTransactionTimeout(stmt);
    } else {
      Connection connection = getConnection(statementLog);
      stmt = handler.prepare(connection, transaction.getTimeout());
      putStatement(sql, stmt);
    }
    handler.parameterize(stmt);
    return stmt;
  }

SimpleExecutor & ReuseExecutor 對(duì)比總結(jié)

SimpleExecutor 比 ReuseExecutor 的性能要差 ， 因?yàn)?SimpleExecutor 沒有做 PSCache败徊。為什么做了 PSCache 性能就會(huì)高呢 帚呼， 因?yàn)楫?dāng)SQL越復(fù)雜占位符越多的時(shí)候預(yù)編譯的時(shí)間也就越長(zhǎng)，創(chuàng)建一個(gè) PreparedStatement 對(duì)象的時(shí)間也就越長(zhǎng)皱蹦。測(cè)試代碼 ：

public class PrepareStatementTest {

    private static final Logger LOG = LoggerFactory.getLogger(PrepareStatementTest.class);
    private static final Properties PROP = new Properties();
    private static final int COUNT = 2000000;

    private String driverClass;
    private String url;
    private String user;
    private String password;

    private static final StringBuilder TEST_PREPAREED_STATEMENT_SQL = new StringBuilder();
    private long startTime;

    @Before
    public void init() throws Exception {

        PROP.load(PrepareStatementTest.class.getClassLoader().getResourceAsStream("jdbc.properties"));
        this.driverClass = PROP.getProperty("jdbc.driver");
        this.url = PROP.getProperty("jdbc.url");
        this.user = PROP.getProperty("jdbc.user");
        this.password = PROP.getProperty("jdbc.password");
        DriverManager.registerDriver((Driver) Class.forName(this.driverClass).newInstance());

        TEST_PREPAREED_STATEMENT_SQL.append("INSERT INTO `user` (nickname,realname,phone,login_password,pay_password,create_time,update_time) VALUES");
        for (int i = 0 ; i < COUNT ; i++ ) {
            if (i == COUNT - 1) {
                TEST_PREPAREED_STATEMENT_SQL.append(" (?,?,?,?,?,?,?)");
            } else {
                TEST_PREPAREED_STATEMENT_SQL.append(" (?,?,?,?,?,?,?)").append(", ");
            }
        }

        this.startTime = System.currentTimeMillis();
    }

    @After
    public void after() throws Exception {
        LOG.debug("======> cost time {} ms" , (System.currentTimeMillis() - this.startTime));
    }

    @Test
    public void costTime() throws Exception {
        Connection connection = DriverManager.getConnection(url, user, password);
        PreparedStatement ps = connection.prepareStatement(TEST_PREPAREED_STATEMENT_SQL.toString());

    }
}

BatchExecutor

BatchExecutor 的特性其實(shí)非常簡(jiǎn)單煤杀，其實(shí)就是調(diào)用了 Statement 的 addBatch 方法。另外千萬(wàn)不要認(rèn)為 BatchExecutor 比 ReuseExecutor 功能強(qiáng)大性能高 沪哺， 實(shí)際上不是的 BatchExecutor 是沒有做 PSCache 的怜珍。

  @Override
  public int doUpdate(MappedStatement ms, Object parameterObject) throws SQLException {
    final Configuration configuration = ms.getConfiguration();
    final StatementHandler handler = configuration.newStatementHandler(this, ms, parameterObject, RowBounds.DEFAULT, null, null);
    final BoundSql boundSql = handler.getBoundSql();
    final String sql = boundSql.getSql();
    final Statement stmt;
    if (sql.equals(currentSql) && ms.equals(currentStatement)) {
      int last = statementList.size() - 1;
      stmt = statementList.get(last);
      applyTransactionTimeout(stmt);
     handler.parameterize(stmt);//fix Issues 322
      BatchResult batchResult = batchResultList.get(last);
      batchResult.addParameterObject(parameterObject);
    } else {
      Connection connection = getConnection(ms.getStatementLog());
      stmt = handler.prepare(connection, transaction.getTimeout());
      handler.parameterize(stmt);    //fix Issues 322
      currentSql = sql;
      currentStatement = ms;
      statementList.add(stmt);
      batchResultList.add(new BatchResult(ms, sql, parameterObject));
    }
  // handler.parameterize(stmt);
    handler.batch(stmt);
    return BATCH_UPDATE_RETURN_VALUE;
  }

BatchExecutor 與 SimpleExecutor 和 ReuseExecutor 還有一個(gè)區(qū)別就是 ， BatchExecutor 的事務(wù)是沒法自動(dòng)提交的凤粗。因?yàn)?BatchExecutor 只有在調(diào)用了 SqlSession 的 commit 方法的時(shí)候 ， 它才會(huì)去執(zhí)行 executeBatch 方法。

  // BaseExecutor 的 commit 方法
  public void commit(boolean required) throws SQLException {
    if (closed) throw new ExecutorException("Cannot commit, transaction is already closed");
    clearLocalCache();
    flushStatements();
    if (required) {
      transaction.commit();
    }
  }

// BatchExecutor 中的 doFlushStatements 方法
@Override
  public List<BatchResult> doFlushStatements(boolean isRollback) throws SQLException {
    try {
      List<BatchResult> results = new ArrayList<BatchResult>();
      if (isRollback) {
        return Collections.emptyList();
      }
      for (int i = 0, n = statementList.size(); i < n; i++) {
        Statement stmt = statementList.get(i);
        applyTransactionTimeout(stmt);
        BatchResult batchResult = batchResultList.get(i);
        try {
          // 這里才調(diào)用了 executeBatch 方法
          batchResult.setUpdateCounts(stmt.executeBatch());
          MappedStatement ms = batchResult.getMappedStatement();
          List<Object> parameterObjects = batchResult.getParameterObjects();
          KeyGenerator keyGenerator = ms.getKeyGenerator();
          if (Jdbc3KeyGenerator.class.equals(keyGenerator.getClass())) {
            Jdbc3KeyGenerator jdbc3KeyGenerator = (Jdbc3KeyGenerator) keyGenerator;
            jdbc3KeyGenerator.processBatch(ms, stmt, parameterObjects);
          } else if (!NoKeyGenerator.class.equals(keyGenerator.getClass())) { //issue #141
            for (Object parameter : parameterObjects) {
              keyGenerator.processAfter(this, ms, stmt, parameter);
            }
          }
        } catch (BatchUpdateException e) {
          StringBuilder message = new StringBuilder();
          message.append(batchResult.getMappedStatement().getId())
              .append(" (batch index #")
              .append(i + 1)
              .append(")")
              .append(" failed.");
          if (i > 0) {
            message.append(" ")
                .append(i)
                .append(" prior sub executor(s) completed successfully, but will be rolled back.");
          }
          throw new BatchExecutorException(message.toString(), e, results, batchResult);
        }
        results.add(batchResult);
      }
      return results;
    } finally {
      for (Statement stmt : statementList) {
        closeStatement(stmt);
      }
      currentSql = null;
      statementList.clear();
      batchResultList.clear();
    }
  }

CachingExecutor

再回頭去看一下 Executor 組件的類結(jié)構(gòu)圖 嫌拣， 發(fā)現(xiàn) CachingExecutor 沒有 extends BaseExecutor 柔袁， 為什么？一定有原因這個(gè)世界上沒有無(wú)緣無(wú)故的事情异逐〈匪鳎看看 CachingExecutor 的實(shí)現(xiàn)，發(fā)現(xiàn) CachingExecutor 其實(shí)是一個(gè)裝飾者對(duì)象 灰瞻， mybatis 這里對(duì) Executor 的設(shè)計(jì)使用了 Decorator （裝飾者） 設(shè)計(jì)模式腥例。

public class CachingExecutor implements Executor {

  private Executor delegate;
  private TransactionalCacheManager tcm = new TransactionalCacheManager();

  public CachingExecutor(Executor delegate) {
    this.delegate = delegate;
    delegate.setExecutorWrapper(this);
  }

接下來(lái)說(shuō)明 CachingExecutor 這個(gè)裝飾者對(duì)象的作用 ，看名字的話也能猜的差不多了 酝润， 這個(gè)裝飾對(duì)象是用來(lái)處理二級(jí)緩存的燎竖。 當(dāng)全局設(shè)置開啟了二級(jí)緩存時(shí)會(huì)初始化一個(gè) CachingExecutor 。

  // org.apache.ibatis.session.Configuration 中的方法
  public Executor newExecutor(Transaction transaction, ExecutorType executorType) {
    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);
    }
    
    // 如果開啟了二級(jí)緩存 要销， 實(shí)例化 CachingExecutor 對(duì)象
    if (cacheEnabled) {
      executor = new CachingExecutor(executor);
    }
    executor = (Executor) interceptorChain.pluginAll(executor);
    return executor;
  }

// CachingExecutor
@Override
  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);
        @SuppressWarnings("unchecked")
        List<E> list = (List<E>) tcm.getObject(cache, key);
        if (list == null) {
          list = delegate.<E> query(ms, parameterObject, rowBounds, resultHandler, key, boundSql);
          tcm.putObject(cache, key, list); // issue #578 and #116
        }
        return list;
      }
    }
    return delegate.<E> query(ms, parameterObject, rowBounds, resultHandler, key, boundSql);
  }

總結(jié)

實(shí)際生產(chǎn)環(huán)境中建議使用 ReuseExecutor 构回， 另外在實(shí)際應(yīng)用中涉及到大量數(shù)據(jù)的更新，插入操作不建議使用 mybatis 而應(yīng)該使用原生的 JDBC 操作 疏咐， 因?yàn)閿?shù)據(jù)量很大的時(shí)候進(jìn)行一次 executeBatch 也是很耗時(shí)的 纤掸， 使用原生 JDBC 操作可以 clearBatch 和 executeBatch 結(jié)合使用提高性能 ; 通過對(duì)源碼的分析，了解到了各個(gè)Executor實(shí)現(xiàn)的優(yōu)劣 浑塞， 它們之間的組織與協(xié)同關(guān)系 借跪， 希望閱讀源碼不僅僅是熟悉了框架的運(yùn)行流程，實(shí)現(xiàn)原理酌壕，更能體會(huì)作者的思想掏愁，分析出框架的優(yōu)點(diǎn)缺點(diǎn) ， 結(jié)合實(shí)際為自己所用仅孩。站在巨人的肩膀上能看的更遠(yuǎn)托猩。