SQL Oracle 摘要 table 查询每个组的前 n 个

Question

我有两个 table 这样描述的

交易Table 客户 ID | fk 给客户 table, 金额 |每笔交易花费的金额

客户Table 客户 ID | , 年龄 |客户年龄

我需要做按年龄范围对客户进行分组或分桶的查询我需要让每个年龄组的前 3 名金额最高，也是后 3 名或最低消费者到目前为止我所做的是 https://dbfiddle.uk/?rdbms=oracle_11.2&fiddle=ccf988ec525dc133cf7825ba052e8054

what i manage so far inner query

我已经尝试使用 CTE，但仍然不确定如何在每组中的每个 group/bottom 3 个中获得前 3 名 with cte

问题出在第一行的第一张图片上，return 超过 3/他们附加了所有记录，而我只需要前 3

感谢您的帮助

Answer 1

您可以使用 ROW_NUMBER() 分析函数找到最高和最低，然后您可以在 LISTAGG 中使用条件聚合仅显示相应列表中的最高和最低：

select age_bucket, 
       LISTAGG(CASE WHEN max_rn <= 3 THEN amount END, ',')
         WITHIN GROUP (ORDER BY amount DESC) AS TOP_3_AMOUNT,
       LISTAGG(CASE WHEN min_rn <= 3 THEN amount END, ',')
         WITHIN GROUP (ORDER BY amount ASC ) AS BOTTOM_3_AMOUNT
FROM   (
  SELECT amount,
         age_bucket,
         ROW_NUMBER() OVER (PARTITION BY age_bucket ORDER BY amount DESC) AS max_rn,
         ROW_NUMBER() OVER (PARTITION BY age_bucket ORDER BY amount ASC ) AS min_rn
  FROM   (
    select t.amount,
           case
           when c.age <= 20             THEN '<= 20'
           when c.age between 21 and 30 THEN '21-30'
           when c.age between 31 and 40 THEN '31-40'
           when c.age between 41 and 50 THEN '41-50'
           when c.age >= 51             THEN '>= 51'
           END as age_bucket 
    from   transaction t
           join customer c on t.cust_id=c.cust_id
  )
)
WHERE max_rn <= 3
OR    min_rn <= 3
group by age_bucket
order by age_bucket asc;

db<>fiddle here

I need to make it only top 3 or bottom 3 distinct by the customer id

select age_bucket, 
       LISTAGG(CASE WHEN max_rn <= 3 THEN amount END, ',')
         WITHIN GROUP (ORDER BY amount DESC) AS TOP_3_AMOUNT,
       LISTAGG(CASE WHEN min_rn <= 3 THEN amount END, ',')
         WITHIN GROUP (ORDER BY amount ASC ) AS BOTTOM_3_AMOUNT
FROM   (
  SELECT amount,
         age_bucket,
         CASE max_cust_rn
         WHEN 1
         THEN ROW_NUMBER() OVER (PARTITION BY age_bucket ORDER BY amount DESC)
         END AS max_rn,
         CASE min_cust_rn
         WHEN 1
         THEN ROW_NUMBER() OVER (PARTITION BY age_bucket ORDER BY amount ASC )
         END AS min_rn
  FROM   (
    SELECT amount,
           age_bucket,
           ROW_NUMBER() OVER (
             PARTITION BY age_bucket, cust_id ORDER BY amount DESC) AS max_cust_rn,
           ROW_NUMBER() OVER (
             PARTITION BY age_bucket, cust_id ORDER BY amount ASC ) AS min_cust_rn
    FROM   (
      select t.amount,
             t.cust_id,
             case
             when c.age <= 20             THEN '<= 20'
             when c.age between 21 and 30 THEN '21-30'
             when c.age between 31 and 40 THEN '31-40'
             when c.age between 41 and 50 THEN '41-50'
             when c.age >= 51             THEN '>= 51'
             END as age_bucket 
      from   transaction t
             join customer c on t.cust_id=c.cust_id
    )
  )
  WHERE max_cust_rn = 1
  OR    min_cust_rn = 1
)
WHERE max_rn <= 3
OR    min_rn <= 3
group by age_bucket
order by age_bucket asc;

db<>fiddle here

Answer 2

因为没有表的说明，所以就试着改了一点你能试试吗：

SELECT   age_bucket,
         Listagg(
         CASE
                  WHEN seqnumbyeachcusttop = 1 THEN amount
         END, ',') within GROUP (ORDER BY amount DESC)            AS test_top_3_amouny,
         listagg(amount, ',') within GROUP (ORDER BY amount DESC) AS top_3_principal
FROM     (
                  SELECT   age_bucket,
                           seqnumbyeachcusttop,
                           amount,
                           row_number() OVER (partition BY amount ORDER BY amount DESC) rn
                  FROM     (
                                    SELECT   t.amount                                                          AS amount,
                                             t.cust_id                                                         AS cust_id,
                                             row_number() OVER (partition BY t.cust_id ORDER BY t.amount DESC) AS seqnumbyeachcusttop,
                                             CASE
                                                      WHEN c.age <= 20 THEN ' <= 20'
                                                      WHEN c.age BETWEEN 21 AND      30 THEN '21-30'
                                                      WHEN c.age BETWEEN 31 AND      40 THEN '31-40'
                                                      WHEN c.age BETWEEN 41 AND      50 THEN '41-50'
                                                      WHEN c.age >= 51 THEN '>= 51'
                                             END AS age_bucket
                                    FROM     TRANSACTION t
                                    JOIN     customer c
                                    ON       t.cust_id=c.cust_id ))
WHERE    rn <= 3) GROUP BY age_bucket ORDER BY age_bucket ASC;

SQL Oracle 摘要 table 查询每个组的前 n 个

SQL Oracle summary table query with top-n on each group

sql

oracle