显示 SQL 行到行数可能不同的列

Question

我有一个 main table 如下

ID    Latest_rating
---  --------------
1      3
2      5

然后我有一个 main_audit table 显示评级变化历史如下

ID   rating  rating_change_date
--   ------  ------------------
1      2     26-Sep-15
1      4     16-Apr-14
1      3     10-Mar-13
2      3     04-Apr-15
2      2     01-Oct-13

所以我必须同时加入 tables 并且应该显示如下

ID  Latest_Rating  1st_Rating rating_change_date 2nd_rating rating_change_date
--  -------------  ---------- ------------------ ---------- ------------------
1         3            3          10-Mar-13          4          16-Apr-14
2         5            2          01-Oct-13          3          04-Apr-15

我应该在上面的列中显示为 1st rating, 2nd rating, 3rd rating, 4 rating

Answer 1

您可以使用 row_number() 和聚合来做到这一点。棘手的部分是：

select ma.id,
       max(case when seqnum = 1 then rating end) as first_rating,
       max(case when seqnum = 1 then rating_change_date end) as first_rating_change_date,
       max(case when seqnum = 2 then ma.rating end) as second_rating,
       max(case when seqnum = 2 then ma.rating_change_date end) as second_rating_change_date
from (select ma.*,
             row_number() over (partition by id order by rating_change_date desc) as seqnum
      from main_audit ma
     ) ma
group by id;

然后您可以组合它们：

with ma as (
      select ma.id,
             max(case when seqnum = 1 then rating end) as first_rating,
             max(case when seqnum = 1 then rating_change_date end) as first_rating_change_date,
             max(case when seqnum = 2 then ma.rating end) as second_rating,
             max(case when seqnum = 2 then ma.rating_change_date end) as second_rating_change_date
      from (select ma.*,
                   row_number() over (partition by id order by rating_change_date desc) as seqnum
            from main_audit ma
           ) ma
      group by id
     )
select m.id, m.latest_rating,
       ma.first_rating, ma.first_rating_change_date,
       ma.second_rating, ma.second_rating_change_date
from main m join
     ma
     on m.id = ma.id;

Answer 2

根据您的评论，如果评级相同，您可以合并，如果有 3 次更改 1 和 2 相同，则可以采用这种方式的最短日期 1 和 2 合并，第三次更改上升到第 2 位. 也不需要 3 个查询，你可以在 1 个嵌套的 select 中使用连接和条件聚合来完成：

SELECT
    m.Id
    ,m.Latest_rating
    ,MAX(CASE WHEN t.RateChangeNum = 1 THEN t.rating END) as first_rating
    ,MAX(CASE WHEN t.RateChangeNum = 1 THEN t.MinRatingChangeDate END) as first_rating_change_date
    ,MAX(CASE WHEN t.RateChangeNum = 2 THEN t.rating END) as first_rating
    ,MAX(CASE WHEN t.RateChangeNum = 2 THEN t.MinRatingChangeDate END) as first_rating_change_date
FROM
    main m
    INNER JOIN
    (
       SELECT
          ID
          ,rating
          ,MIN(rating_change_date) as MinRatingChangeDate
          ,ROW_NUMBER() OVER (PARTITION BY ID ORDER BY MIN(rating_change_date)) as RateChangeNum
       FROM
          main_audit a
       GROUP BY
          ID
          ,rating
    ) t
    ON m.Id = t.ID
GROUP BY
    m.Id
    ,m.Latest_rating

至于您对变化数量的评论。您可以硬编码到您认为可能存在或可能不存在的完整数字，如果没有那么多更改，或者您需要转向动态 sql.

，则字段将为空

条件聚合而不是 PIVOT 的原因是因为你需要聚合多个列，如果你只想要评级或日期那么你可以使用 PIVOT 并且会少一些代码编写（http://www.oracle.com/technetwork/articles/sql/11g-pivot-097235.html)

显示 SQL 行到行数可能不同的列

Displaying SQL rows to columns where the row count may vary

sql

oracle

pivot