在 Oracle 中将 EXISTS 重写为 JOIN 或子查询

Question

我有一个查询成本非常高，需要一个多小时才能执行。我尝试将 EXISTS 子句转换为加入，但我卡住了，有人能帮忙吗？

目的是在唯一的 space ID 中查找重复的产品。 FLAT_TABLE 包含 500 万条记录。

查询：

 select
    tbl1.product,
    tbl1.status,
    tbl1.reservation,
    tbl1.unique_space_id
FROM
    schema1.flat_table tbl1
WHERE
    tbl1.status = 'Active'
    AND tbl1.product = 'Cage'
    AND EXISTS
        (SELECT 1
        FROM schema1.flat_table tbl2
        WHERE tbl2.product = 'Cage'
        AND tbl2.status = 'Active'
        AND tbl2.reservation <> 'Space Reserved'
        AND tbl1.unique_space_id = tbl2.unique_space_id
        GROUP BY tbl2.unique_space_id
        HAVING COUNT (1) > 1
        );

Answer 1

您可以将查询重写为当前存在的子查询的内部联接。加入将以与 exists 子句相同的方式产生过滤效果。

SELECT DISTINCT
    tbl1.product,
    tbl1.status,
    tbl1.reservation,
    tbl1.unique_space_id
FROM schema1.flat_table tbl1
INNER JOIN
(
    SELECT unique_space_id
    FROM schema1.flat_table
    WHERE product = 'Cage' AND
          status = 'Active' AND
          reservation <> 'Space Reserved'
    GROUP BY unique_space_id
    HAVING COUNT(*) > 1
) tbl2
    ON tbl2.unique_space_id = tbl1.unique_space_id
WHERE
    tbl1.status = 'Active' AND
    tbl1.product = 'Cage';

这是一个更简洁的版本，使用 COUNT 作为分析函数，以及一个 QUALIFY 子句；

SELECT DISTINCT product, status, reservation, unique_space_id
FROM schema1.flat_table
WHERE status = 'Active' AND product = 'Cage'
QUALIFY COUNT(CASE WHEN reservation <> 'Space Reserved' THEN 1 END) 
        OVER (PARTITION BY unique_space_id) > 1;

Answer 2

您可以使用解析函数count，如下所示：

select * from
(select tbl1.product, tbl1.status, tbl1.reservation, tbl1.unique_space_id,
        count(case when tbl1.reservation <> 'Space Reserved' then 1 end) 
         over(partition by tbl1.unique_space_id) as cnt 
  FROM schema1.flat_table tbl1  
 WHERE tbl1.status = 'Active' AND tbl1.product = 'Cage')
 where cnt > 1

在 Oracle 中将 EXISTS 重写为 JOIN 或子查询

Re-writing EXISTS as JOIN or a subquery in Oracle

sql

oracle

performance

exists