Sharding-JDBC源码解析

1. Sharding-JDBC初始化

Sharding-JDBC的初始化主要包括两个方面：

1. 数据源元数据信息和表元数据信息的收集。
2. 表分库分表策略和算法的配置信息收集。

工厂类ShardingDataSourceFactory.createDataSource()方法在创建Sharding-JDBC的数据源实现类ShardingDataSource的同时还创建了ShardingRule、ShardingContext两个核心类的对象，如下相关源码：

public final class ShardingDataSourceFactory {
    
    public static DataSource createDataSource(
            final Map dataSourceMap, final ShardingRuleConfiguration shardingRuleConfig, final Map configMap, final Properties props) throws SQLException {
        return new ShardingDataSource(dataSourceMap, new ShardingRule(shardingRuleConfig, dataSourceMap.keySet()), configMap, props);
    }
}

public class ShardingDataSource extends AbstractDataSourceAdapter {
    
    private final ShardingContext shardingContext;
    
    public ShardingDataSource(final Map dataSourceMap, final ShardingRule shardingRule, final Map configMap, final Properties props) throws SQLException {
        super(dataSourceMap);
        checkDataSourceType(dataSourceMap);
        if (!configMap.isEmpty()) {
            ConfigMapContext.getInstance().getConfigMap().putAll(configMap);
        }
        shardingContext = new ShardingContext(getDataSourceMap(), shardingRule, getDatabaseType(), props);
    }
}

ShardingContext持有两个属性ShardingRule、ShardingMetaData，ShardingRule保存了表的分库分表配置，这些配置包括分库策略以及算法、分表策略以及算法，也就是说根据一个表以及这个表的列可以从ShardingRule中获取这个表的分库分表策略和算法。ShardingMetaData则维护了数据源和表的元数据信息，其有两个属性：ShardingDataSourceMetaData和ShardingTableMetaData，分表表示数据源的元数据信息和表的元数据信息，这两个属性在ShardingMetaData的构造函数中被创建，源码如下：

public final class ShardingMetaData {
    
    private final ShardingDataSourceMetaData dataSource;
    
    private final ShardingTableMetaData table;
    
    public ShardingMetaData(final Map dataSourceURLs, final ShardingRule shardingRule, final DatabaseType databaseType, final ShardingExecuteEngine executeEngine, 
                            final TableMetaDataConnectionManager connectionManager, final int maxConnectionsSizePerQuery, final boolean isCheckingMetaData) {
        dataSource = new ShardingDataSourceMetaData(dataSourceURLs, shardingRule, databaseType);
        table = new ShardingTableMetaData(new TableMetaDataInitializer(dataSource, executeEngine, connectionManager, maxConnectionsSizePerQuery, isCheckingMetaData).load(shardingRule));
    }
}

ShardingRuleConfiguration是分库分表配置的核心和入口，它可以包含多个TableRuleConfiguration和MasterSlaveRuleConfiguration。每一组相同规则分片的表配置一个TableRuleConfiguration。一个TableRuleConfiguration表示一个表的分库分表策略配置，其持有两个类型为ShardingStrategyConfiguration的属性：databaseShardingStrategyConfig和tableShardingStrategyConfig，分别表示分库策略配置和分表策略配置。ShardingStrategyConfiguration有如下四种实现：

• StandardShardingStrategyConfiguration 支持精确分片和范围分片
• ComplexShardingStrategyConfiguration 支持复杂分表
• HintShardingStrategyConfiguration 强制某种策略分片
• InlineShardingStrategyConfiguration 支持表达式分片
  以上每种分片策略配置都关联一到两个对应的分片算法，分片算法由接口ShardingAlgorithm表示，其抽象子类有：
• PreciseShardingAlgorithm 精确分片算法
• RangeShardingAlgorithm 范围分片算法
• HintShardingAlgorithm 强制分片算法
• ComplexKeysShardingAlgorithm 复杂分片算法

Sharding-JDBC会使用ShardingRuleConfiguration实力化TableRule对象，源码如下：

public class ShardingRule {
    
    private final ShardingRuleConfiguration shardingRuleConfig;
    
    private final ShardingDataSourceNames shardingDataSourceNames;
    
    private final Collection tableRules = new LinkedList<>();
    
    private final ShardingStrategy defaultDatabaseShardingStrategy;
    
    private final ShardingStrategy defaultTableShardingStrategy;
    
    private final KeyGenerator defaultKeyGenerator;
    
    private final Collection masterSlaveRules = new LinkedList<>();
    
    public ShardingRule(final ShardingRuleConfiguration shardingRuleConfig, final Collection dataSourceNames) {
        Preconditions.checkNotNull(dataSourceNames, "Data sources cannot be null.");
        Preconditions.checkArgument(!dataSourceNames.isEmpty(), "Data sources cannot be empty.");
        this.shardingRuleConfig = shardingRuleConfig;
        shardingDataSourceNames = new ShardingDataSourceNames(shardingRuleConfig, dataSourceNames);
        for (TableRuleConfiguration each : shardingRuleConfig.getTableRuleConfigs()) {
            // 对于每一个TableRuleConfiguration都生成一个TableRule对象，TableRule构造函数会排列组合表和数据源实力化DataNode集合
            tableRules.add(new TableRule(each, shardingDataSourceNames));
        }
        for (String group : shardingRuleConfig.getBindingTableGroups()) {
            List tableRulesForBinding = new LinkedList<>();
            for (String logicTableNameForBindingTable : StringUtil.splitWithComma(group)) {
                tableRulesForBinding.add(getTableRuleByLogicTableName(logicTableNameForBindingTable));
            }
            bindingTableRules.add(new BindingTableRule(tableRulesForBinding));
        }
        broadcastTables.addAll(shardingRuleConfig.getBroadcastTables());
        defaultDatabaseShardingStrategy = null == shardingRuleConfig.getDefaultDatabaseShardingStrategyConfig()
                ? new NoneShardingStrategy() : ShardingStrategyFactory.newInstance(shardingRuleConfig.getDefaultDatabaseShardingStrategyConfig());
        defaultTableShardingStrategy = null == shardingRuleConfig.getDefaultTableShardingStrategyConfig()
                ? new NoneShardingStrategy() : ShardingStrategyFactory.newInstance(shardingRuleConfig.getDefaultTableShardingStrategyConfig());
        defaultKeyGenerator = null == shardingRuleConfig.getDefaultKeyGenerator() ? new DefaultKeyGenerator() : shardingRuleConfig.getDefaultKeyGenerator();
        for (MasterSlaveRuleConfiguration each : shardingRuleConfig.getMasterSlaveRuleConfigs()) {
            masterSlaveRules.add(new MasterSlaveRule(each));
        }
    }
    // 略
}

一个TableRule对象表示一个逻辑表的库表资源，其维护一个类型为DataNode的集合属性actualDataNodes，这个DataNode集合表示此逻辑表对应的实际库表的集合，例如：现在有两个库db0、db1，每个库有三个表，逻辑表名为t_order，那么TableRule对象的属性actualDataNodes则有6个元素：

db0 t_order0
db0 t_order1
db0 t_order2
db1 t_order0
db1 t_order1
db1 t_order2

Sharding-JDBC做路由时即是根据此集合使用相应的算法进行实际的库表选取的。

相关类关系图如下：

image

2. sql的解析与路由

ShardingPreparedStatement实现了PreparedStatement的三个核心方法：executeQuery()、executeUpdate()、execute()，相关源码如下：

public final class ShardingPreparedStatement extends AbstractShardingPreparedStatementAdapter {
    
    @Getter
    private final ShardingConnection connection;
    
    private final PreparedStatementRoutingEngine routingEngine;
    
    private final PreparedStatementExecutor preparedStatementExecutor;
    
    private SQLRouteResult routeResult;
    
    private ResultSet currentResultSet;
    
    // 略
    
    @Override
    public ResultSet executeQuery() throws SQLException {
        ResultSet result;
        try {
            clearPrevious();
            sqlRoute();
            initPreparedStatementExecutor();
            MergeEngine mergeEngine = MergeEngineFactory.newInstance(connection.getShardingContext().getDatabaseType(), connection.getShardingContext().getShardingRule(), 
                    routeResult.getSqlStatement(), connection.getShardingContext().getMetaData().getTable(), preparedStatementExecutor.executeQuery());
            result = new ShardingResultSet(preparedStatementExecutor.getResultSets(), mergeEngine.merge(), this);
        } finally {
            clearBatch();
        }
        currentResultSet = result;
        return result;
    }
    
    @Override
    public int executeUpdate() throws SQLException {
        try {
            clearPrevious();
            sqlRoute();
            initPreparedStatementExecutor();
            return preparedStatementExecutor.executeUpdate();
        } finally {
            refreshTableMetaData(connection.getShardingContext(), routeResult.getSqlStatement());
            clearBatch();
        }
    }
    
    @Override
    public boolean execute() throws SQLException {
        try {
            clearPrevious();
            sqlRoute();
            initPreparedStatementExecutor();
            return preparedStatementExecutor.execute();
        } finally {
            refreshTableMetaData(connection.getShardingContext(), routeResult.getSqlStatement());
            clearBatch();
        }
    }
    
    private void sqlRoute() {
        routeResult = routingEngine.route(new ArrayList<>(getParameters()));
    }
    
    // 略

从上面的源码可以看到，三个核心方法都调用到了sqlRoute()方法，此方法是sql的解析与路由的入口，sqlRoute()方法直接调用PreparedStatementRoutingEngine的route()方法，此方法首先调用shardingRouter的parse()方法sql进行解析，接着调用route()方法进行路由，相关源码如下：

public final class PreparedStatementRoutingEngine {
    
    private final String logicSQL;
    
    private final ShardingRouter shardingRouter;
    
    private final ShardingMasterSlaveRouter masterSlaveRouter;
    
    private SQLStatement sqlStatement;
    
    public PreparedStatementRoutingEngine(final String logicSQL, final ShardingRule shardingRule, final ShardingMetaData shardingMetaData, final DatabaseType databaseType, final boolean showSQL) {
        this.logicSQL = logicSQL;
        shardingRouter = ShardingRouterFactory.newInstance(shardingRule, shardingMetaData, databaseType, showSQL);
        masterSlaveRouter = new ShardingMasterSlaveRouter(shardingRule.getMasterSlaveRules());
    }
    
    public SQLRouteResult route(final List