Count unique ids in a rolling time frame

与世无争的帅哥 提交于 2020-01-10 04:01:04

问题


I have a simple table as bellow with lots of IDs and dates.

ID      Date
10R46   2014-11-23  
10R46   2016-04-11  
100R9   2016-12-21
10R91   2013-05-03 
...     ...

I want to formulate a query which counts the unique IDs for a rolling time frame of dates, for example ten days. Meaning that for each date it should give me the number of unique IDs between that date and 10 days back. Result should look something like this.

UniqueTenDays    Date
200              2014-11-23 
324              2014-11-24 
522              2014-11-25
532              2014-11-26 
...              ...

Something along the lines of the bellow but I realise I need to apply the WHERE clause and count the IDs for each Date somehow.

SELECT Date, COUNT(DISTINCT ID)
FROM T 
WHERE Date BETWEEN DATE_SUB(Date, INTERVAL 10 DAY) AND Date
GROUP BY Date
ORDER BY Date

Thanks in advance.


回答1:


Below is for BigQuery Standard SQL

#standardSQL
WITH temp1 AS (
  SELECT dt, STRING_AGG(DISTINCT id) AS users
  FROM `project.dataset.yourtable`
  GROUP BY dt
), temp2 AS (
  SELECT
    dt, 
    STRING_AGG(users) OVER(ORDER BY UNIX_DATE(dt) RANGE BETWEEN 10 PRECEDING AND CURRENT ROW) users
  FROM temp1
)
SELECT dt, 
  (SELECT COUNT(DISTINCT id) FROM UNNEST(SPLIT(users)) AS id) UniqueTenDays
FROM temp2

you can test / play with it using dummy data as below

#standardSQL
WITH `project.dataset.yourtable` AS (
  SELECT '10R46' id,  DATE '2014-11-23' dt UNION ALL  
  SELECT '10R46',     DATE '2016-04-11' UNION ALL  
  SELECT '10R46',     DATE '2016-04-12' UNION ALL  
  SELECT '10R47',     DATE '2016-04-13' UNION ALL  
  SELECT '10R48',     DATE '2016-04-14' UNION ALL  
  SELECT '100R9',     DATE '2016-12-21' UNION ALL
  SELECT '10R91',     DATE '2013-05-03'
), temp1 AS (
  SELECT dt, STRING_AGG(DISTINCT id) AS users
  FROM `project.dataset.yourtable`
  GROUP BY dt
), temp2 AS (
  SELECT
    dt, 
    STRING_AGG(users) OVER(ORDER BY UNIX_DATE(dt) RANGE BETWEEN 10 PRECEDING AND CURRENT ROW) users
  FROM temp1
)
SELECT dt,  
  (SELECT COUNT(DISTINCT id) FROM UNNEST(SPLIT(users)) AS id) UniqueTenDays
FROM temp2


来源:https://stackoverflow.com/questions/47658874/count-unique-ids-in-a-rolling-time-frame

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!