Oracle：左连接非常大 table 并将连接的行限制为具有最大字段值的行

Question

我有两个 table。第二个通过 m_id.

引用第一个

主要table

M_ID | M_FIELD
  1  | 'main1'
  2  | 'main2'
  3  | 'main3'

亚table

S_ID | S_FIELD  | S_ORDER | M_ID
  1  | 'sub1-1' |    1    |  1
  2  | 'sub1-2' |    2    |  1
  3  | 'sub1-3' |    3    |  1
  4  | 'sub2-1' |    1    |  2
  5  | 'sub2-2' |    2    |  2
  6  | 'sub2-3' |    3    |  2
  7  | 'sub3-1' |    1    |  3
  8  | 'sub3-2' |    2    |  3
  9  | 'sub3-3' |    3    |  3

我需要加入这两个 tables（按 M_ID）但是从 Sub-table 我只需要具有最大值 S_ORDER 的行。

所以查询的预期结果是：

M_ID | M_FIELD | S_FIELD
  1  | 'main1' | 'sub1-3'
  2  | 'main2' | 'sub2-3'
  3  | 'main3' | 'sub3-3'

这个问题的答案中有解析函数的工作解：How do I limit the number of rows returned by this LEFT JOIN to one? （我会post它在底部）但问题是 Sub-Table 非常大（实际上是一个带有一些内部计算的视图）并且这种子查询工作时间太长。所以我想我需要先通过 m_id 过滤掉 table ，然后才找到最大 S_ORDER[=25= 的字段]

我需要像这样简单的东西（失败是因为二级子查询在外面看不到 M.M_ID 字段）：

SELECT m.*,
       (SELECT s_field
        FROM (SELECT s_field
              FROM t_sub s
              WHERE s.m_id = m.m_id
              ORDER BY s_order DESC)
        WHERE ROWNUM = 1) s_field
 FROM t_main m;

创建和填充测试架构的代码：

CREATE TABLE t_main (m_id NUMBER PRIMARY KEY,
                     m_field VARCHAR2(10));
CREATE TABLE t_sub  (s_id NUMBER PRIMARY KEY,
                     s_field VARCHAR2(10),
                     s_order NUMBER,
                     m_id NUMBER );
INSERT INTO t_main VALUES (1,'main1');
INSERT INTO t_main VALUES (2,'main2');
INSERT INTO t_main VALUES (3,'main3');
INSERT INTO t_sub VALUES (1,'sub1-1', 1, 1);
INSERT INTO t_sub VALUES (2,'sub1-2', 2, 1);
INSERT INTO t_sub VALUES (3,'sub1-3', 3, 1);
INSERT INTO t_sub VALUES (4,'sub2-1', 1, 2);
INSERT INTO t_sub VALUES (5,'sub2-2', 2, 2);
INSERT INTO t_sub VALUES (6,'sub2-3', 3, 2);
INSERT INTO t_sub VALUES (7,'sub3-1', 1, 3);
INSERT INTO t_sub VALUES (8,'sub3-2', 2, 3);
INSERT INTO t_sub VALUES (9,'sub3-3', 3, 3);
COMMIT;

上面提到的工作解决方案（大 T_SUB table 工作太慢）：

SELECT m.*,
       s.s_field
FROM t_main m
LEFT JOIN
  (SELECT *
   FROM
     (SELECT ts.*,
             ROW_NUMBER() OVER (PARTITION BY m_id
                                ORDER BY s_order DESC) AS seq
      FROM t_sub ts)
   WHERE seq = 1) s ON s.m_id = m.m_id;

我们使用的DB是Oracle 10g

非常感谢您的帮助

Answer 1

select t.*, s.s_field from t_main t
left join (select m_id, min(s_field) keep(dense_rank first order by s_order desc) as s_field 
       from t_sub group by m_id) s on (s.m_id = t.m_id)

Answer 2

试试这个

SELECT m.*,
       (select s.s_field 
          from t_sub s
         where s.m_id = m.m_id
           and s.s_order = (select max(s_order) from t_sub where t_sub.m_id = s.m_id)
           and rownum = 1)
FROM t_main m

或者你可以试试这个（这是你的代码，但有一些修改）

SELECT m.*,
      (select s.s_field from 
       (SELECT s_field, m_id
          FROM t_sub
         --where t_sub.m_id = m.m_id
         order by s_order DESC) s
        where s.m_id = m.m_id
          and rownum = 1)
FROM t_main m

Oracle：左连接非常大 table 并将连接的行限制为具有最大字段值的行

Oracle: Left join very big table and limit the joined rows to one with the largest field value

sql

oracle

performance

join

limit