DataFusion SQL查询规划器

此crate提供了一款通用的SQL查询规划器，可以解析SQL并将查询转换为逻辑计划。尽管此crate被DataFusion查询引擎使用，但它被设计成易于从任何需要SQL查询规划器的项目中使用，并且不对结果逻辑计划如何转换为物理计划做任何假设。例如，在逻辑计划中没有基于行与基于列执行的概念。

使用示例

请参阅示例目录以获取完全工作的示例。

以下是从SQL字符串生成逻辑计划的示例。

fn main() {
    let sql = "SELECT \
            c.id, c.first_name, c.last_name, \
            COUNT(*) as num_orders, \
            SUM(o.price) AS total_price, \
            SUM(o.price * s.sales_tax) AS state_tax \
        FROM customer c \
        JOIN state s ON c.state = s.id \
        JOIN orders o ON c.id = o.customer_id \
        WHERE o.price > 0 \
        AND c.last_name LIKE 'G%' \
        GROUP BY 1, 2, 3 \
        ORDER BY state_tax DESC";

    // parse the SQL
    let dialect = GenericDialect {}; // or AnsiDialect, or your own dialect ...
    let ast = Parser::parse_sql(&dialect, sql).unwrap();
    let statement = &ast[0];

    // create a logical query plan
    let schema_provider = MySchemaProvider::new();
    let sql_to_rel = SqlToRel::new(&schema_provider);
    let plan = sql_to_rel.sql_statement_to_plan(statement.clone()).unwrap();

    // show the plan
    println!("{:?}", plan);
}

这是从示例生成的逻辑计划。请注意，这是一个未优化的逻辑计划。datafusion-optimizer crate提供了一款查询优化器，可以将此crate生成的计划进行优化。

Sort: state_tax DESC NULLS FIRST
  Projection: c.id, c.first_name, c.last_name, COUNT(Int64(1)) AS num_orders, SUM(o.price) AS total_price, SUM(o.price * s.sales_tax) AS state_tax
    Aggregate: groupBy=[[c.id, c.first_name, c.last_name]], aggr=[[COUNT(Int64(1)), SUM(o.price), SUM(o.price * s.sales_tax)]]
      Filter: o.price > Int64(0) AND c.last_name LIKE Utf8("G%")
        Inner Join: c.id = o.customer_id
          Inner Join: c.state = s.id
            SubqueryAlias: c
              TableScan: customer
            SubqueryAlias: s
              TableScan: state
          SubqueryAlias: o
            TableScan: orders