0

我需要在过去 1 年中每天计算 MTD 唯一用户数。前任。如果我的数据像

dt  customername
1   a
1   b
2   a
2   c
3   b
3   a
4   c
4   d
4   e

预期输出 - 第一次只有 2 个唯一用户。在第 2 (1&2nd) 3 个用户。3日（1日至3日）3个用户。4日（1日至4日）5个用户。我在过去 1 年的每个日期每个月都需要这个

dt uniquecustcount
1  2
2  3
3  3
4  5

我的数据格式如下，但 MTD 的核心逻辑高于一个。感谢任何帮助

CREATE MULTISET TABLE GK_DAILY_USERS ,NO FALLBACK ,
     NO BEFORE JOURNAL,
     NO AFTER JOURNAL,
     CHECKSUM = DEFAULT,
     DEFAULT MERGEBLOCKRATIO
     (
      EVENT_DATE DATE FORMAT 'yyyy-mm-dd',
      Subs_Id INTEGER,
      PAYMENT_METHOD_CD VARCHAR(8) CHARACTER SET UNICODE NOT CASESPECIFIC,
      Data_User_Flag VARCHAR(1) CHARACTER SET UNICODE NOT CASESPECIFIC,
      BUS_UNIT VARCHAR(20) CHARACTER SET LATIN NOT CASESPECIFIC,
      Event_Count INTEGER)
PRIMARY INDEX ( EVENT_DATE ,Subs_Id );

Insert into GK_DAILY_USERS values (?,?,?,?,?,?);

我正在尝试下面的sql。我知道这个 sql 没有意义，因此正在寻找解决方案。

SELECT EVENT_DATE, MAX(R2) FROM (
SELECT EVENT_DATE 
,COUNT(A.SUBS_ID)OVER(ORDER BY ROWNUM,A.SUBS_ID) AS R2
FROM (
SELECT A.EVENT_DATE,A.SUBS_ID
,Row_Number() Over (Order by EVENT_DATE, SUBS_ID) rownum
,COUNT(A.SUBS_ID)OVER(PARTITION BY A.SUBS_ID ORDER BY A.EVENT_DATE) AS RUN_TOTAL3
FROM  GK_DAILY_USERS A 
WHERE Data_User_Flag='Y' 
AND subs_id in (566875703,289813839, 151153086,279262050,18378517)
ORDER BY EVENT_DATE, SUBS_ID
) A 
WHERE RUN_TOTAL3 = 1
) A GROUP BY EVENT_DATE ORDER BY 1
;
4

1 回答 1

0 
SELECT EVENT_DATE 
   -- new customers per date
  ,Sum(SUBS_ID) AS daily_new_cust
  -- running total of new customers per date
  ,Sum(Sum(SUBS_ID))
   Over(ORDER BY EVENT_DATE
        ROWS Unbounded Preceding) AS unique_cust
FROM
 (
   SELECT *
   FROM  GK_DAILY_USERS
   WHERE Data_User_Flag='Y' 
   AND subs_id IN (566875703,289813839, 151153086,279262050,18378517)
   QUALIFY
      -- return a single row with the earliest date for each subs_id
      Row_Number()
      Over (PARTITION BY A.SUBS_ID
            ORDER BY EVENT_DATE) = 1
 ) AS dt
GROUP BY EVENT_DATE 
ORDER BY 1
;
于 2020-03-24T08:34:01.377 回答