Count Common UserIDs using MySQL between Month X and Month X+1, Month X+2, etc

askcode23
April 25, 2023
190 views
0 votes
2 Answers

I have the following table named Transactions where the timestamp column is in the Unix timestamp format with the minimum timestamp value being January 2008 and the maximum timestamp value is December 2008 (I’m only concerned with months for this problem):

+---------+--------+----------+------------+
| orderID | userID | cityName | timestamp  |
+---------+--------+----------+------------+
|       1 |      1 | city1    | 1199590481 |
|       2 |      2 | city2    | 1199590481 |
|       3 |      1 | city1    | 1199590481 |
|       4 |      2 | city2    | 1199590481 |
|       5 |      1 | city1    | 1202268881 |
|       6 |      2 | city2    | 1202268881 |
|       7 |      1 | city1    | 1202268881 |
|       8 |      2 | city2    | 1202268881 |
|       9 |      1 | city1    | 1199590481 |
|      10 |      2 | city2    | 1202268881 |
|      11 |      1 | city1    | 1205549051 |
|      12 |      2 | city2    | 1208227451 |
|      13 |      1 | city1    | 1210819451 |
|      14 |      2 | city2    | 1213497851 |
|      15 |      1 | city1    | 1216089851 |
|      16 |      2 | city1    | 1218768251 |
|      17 |      1 | city2    | 1221446651 |
|      18 |      2 | city2    | 1224038651 |
|      19 |      1 | city1    | 1226717051 |
|      20 |      2 | city2    | 1229309051 |
+---------+--------+----------+------------+
20 rows in set (0.01 sec)

Here’s the code to generate the Transactions table I have:

CREATE TABLE Transactions (orderID SMALLINT UNSIGNED,userID SMALLINT UNSIGNED,cityName VARCHAR(40),timestamp int(11),constraint pk_Transactions PRIMARY KEY (orderID));

INSERT INTO Transactions (orderID,userID,cityName,timestamp) VALUES(1,1,'city1',1199590481),(2,2,'city2',1199590481),(3,1,'city1',1199590481),(4,2,'city2',1199590481),(5,1,'city1',1202268881),(6,2,'city2',1202268881),(7,1,'city1',1202268881),(8,2,'city2',1202268881),(9,1,'city1',1200375000),(10,2,'city2',1203053400),(11,1,'city1',1205559000),(12,2,'city2',1208237400),(13,1,'city1',1210829400),(14,2,'city2',1213507800),(15,1,'city1',1216099800),(16,2,'city1',1218778200),(17,1,'city2',1221456600),         (18,2,'city2',1224048600),(19,1,'city1',1226727000),(20,2,'city2',1229319000);

My task is to count the common userIDs between each month X and month X+1, month X+2 etc, as in the first two columns of the table posted below (with the same format) to see how users are retained between each month of the year and every other later month.

Basically, If the user has ordered during the months written in the row values for the column Month 1 and column Month 2 in the below table, then that user is included in the count. If he hasn’t ordered in both Month 1 and Month 2, then the user isn’t counted.

The Month 1 and Month 2 columns in the table below should stay as they are, with only the counts being corrected. The number of rows in the table below should stay the same as well, this means including 0 count values in cases where no user has ordered in both the respective Month 1 and Month 2 row values.

In my Transactions table I only have two users, so the count should never be above 2 in any row which isn’t the case in the table below that my query generated:

+---------+---------+------------------------+
| Month 1 | Month 2 | Transacting user count |
+---------+---------+------------------------+
| 200801  | 200802  |                     25 |
| 200801  | 200803  |                      5 |
| 200801  | 200804  |                      5 |
| 200801  | 200805  |                      5 |
| 200801  | 200806  |                      5 |
| 200801  | 200807  |                      5 |
| 200801  | 200808  |                      5 |
| 200801  | 200809  |                      5 |
| 200801  | 200810  |                      5 |
| 200801  | 200811  |                      5 |
| 200801  | 200812  |                      5 |
| 200802  | 200803  |                      5 |
| 200802  | 200804  |                      5 |
| 200802  | 200805  |                      5 |
| 200802  | 200806  |                      5 |
| 200802  | 200807  |                      5 |
| 200802  | 200808  |                      5 |
| 200802  | 200809  |                      5 |
| 200802  | 200810  |                      5 |
| 200802  | 200811  |                      5 |
| 200802  | 200812  |                      5 |
| 200803  | 200804  |                      1 |
| 200803  | 200805  |                      1 |
| 200803  | 200806  |                      1 |
| 200803  | 200807  |                      1 |
| 200803  | 200808  |                      1 |
| 200803  | 200809  |                      1 |
| 200803  | 200810  |                      1 |
| 200803  | 200811  |                      1 |
| 200803  | 200812  |                      1 |
| 200804  | 200805  |                      1 |
| 200804  | 200806  |                      1 |
| 200804  | 200807  |                      1 |
| 200804  | 200808  |                      1 |
| 200804  | 200809  |                      1 |
| 200804  | 200810  |                      1 |
| 200804  | 200811  |                      1 |
| 200804  | 200812  |                      1 |
| 200805  | 200806  |                      1 |
| 200805  | 200807  |                      1 |
| 200805  | 200808  |                      1 |
| 200805  | 200809  |                      1 |
| 200805  | 200810  |                      1 |
| 200805  | 200811  |                      1 |
| 200805  | 200812  |                      1 |
| 200806  | 200807  |                      1 |
| 200806  | 200808  |                      1 |
| 200806  | 200809  |                      1 |
| 200806  | 200810  |                      1 |
| 200806  | 200811  |                      1 |
| 200806  | 200812  |                      1 |
| 200807  | 200808  |                      1 |
| 200807  | 200809  |                      1 |
| 200807  | 200810  |                      1 |
| 200807  | 200811  |                      1 |
| 200807  | 200812  |                      1 |
| 200808  | 200809  |                      1 |
| 200808  | 200810  |                      1 |
| 200808  | 200811  |                      1 |
| 200808  | 200812  |                      1 |
| 200809  | 200810  |                      1 |
| 200809  | 200811  |                      1 |
| 200809  | 200812  |                      1 |
| 200810  | 200811  |                      1 |
| 200810  | 200812  |                      1 |
| 200811  | 200812  |                      1 |
+---------+---------+------------------------+
66 rows in set (0.00 sec)

As you can see, the Transacting user count values don’t make any sense at all, especially given that I have only 2 userIDs in my Transactions table.

I’d really appreciate some help in writing a query for this. Here is the query I used to unsuccessfully generate the table above:

SELECT DISTINCT FROM_UNIXTIME(T1.timestamp, "%Y%m") 'Month 1',
FROM_UNIXTIME(T2.timestamp, "%Y%m") 'Month 2', 
count(*) 'Transacting user count'
FROM Transactions T1
JOIN Transactions T2
ON FROM_UNIXTIME(T1.timestamp, "%Y%m") < FROM_UNIXTIME(T2.timestamp, "%Y%m")
JOIN (
  SELECT count(*) FROM(
    SELECT userID
    FROM Transactions T1
    WHERE FROM_UNIXTIME(timestamp, "%Y%m") = (FROM_UNIXTIME(T1.timestamp, "%Y%m"))
    INTERSECT
    SELECT userID
    FROM Transactions T2
    WHERE FROM_UNIXTIME(timestamp, "%Y%m") = (FROM_UNIXTIME(T2.timestamp, "%Y%m"))
    ) union_table
  ) count_orders
GROUP BY FROM_UNIXTIME(T1.timestamp, "%Y%m"), FROM_UNIXTIME(T2.timestamp, "%Y%m")
ORDER BY 1, 2;

If anything is unclear then please let me know and I’ll try my best to clarify.

Tags: mysql sql

Answers

This SQL should give you what you want. If it doesn’t please update your question with the result you are expecting.

with DATA1 as (
  SELECT userid
  , FROM_UNIXTIME(timestamp, "%Y%m")  YEAR_MONTH1
  , date_format(FROM_UNIXTIME(timestamp) + interval 1 month, "%Y%m")  YEAR_MONTH2
  FROM Transactions
)
SELECT A.YEAR_MONTH1, A.YEAR_MONTH2, COUNT(A.USERID)
FROM DATA1 A
INNER JOIN DATA1 B ON A.USERID = B.USERID AND A.YEAR_MONTH2 = B.YEAR_MONTH2
GROUP BY A.YEAR_MONTH1, A.YEAR_MONTH2
ORDER BY A.YEAR_MONTH1, A.YEAR_MONTH2

SECOND ANSWER

This may give you what you want. If not, please update the table in your question to show the actual values you are expecting.

with dates1 as (
  select distinct FROM_UNIXTIME(timestamp, "%Y%m") month_val
  FROM Transactions
),
date_combos as (
  select a.month_val as YEAR_MONTH1
  , b.month_val as YEAR_MONTH2
  from dates1 a
  cross join dates1 b
  where  a.month_val <  b.month_val

),
txn_month as (
  select distinct userID, FROM_UNIXTIME(timestamp, "%Y%m") order_month
  from Transactions
),
usr_month1 as (
  select dc.YEAR_MONTH1, dc.YEAR_MONTH2
  ,tm1.userId, tm1.order_month
  from date_combos dc
  inner join txn_month tm1 on dc.year_month1 = tm1.order_month
  inner join txn_month tm2 on dc.year_month2 = tm2.order_month
    and tm1.userId = tm2.userId
)
select um1.year_month1, year_month2, count(um1.userId)
from usr_month1 um1
group by  um1.year_month1, year_month2
order by  um1.year_month1, year_month2
;

Try also this

with tranmon as(
  SELECT userid
      ,FROM_UNIXTIME(timestamp, "%Y%m") mon
  FROM Transactions T1
  GROUP BY userid, FROM_UNIXTIME(timestamp, "%Y%m")
)
select from_mon,to_mon
   ,case when to_mon is null then 0 else count(*) end usersqty
from(
   select t1.userid,t1.mon from_mon,t2.mon to_mon
   from tranmon t1
   left join tranmon t2
      on t2.userid=t1.userid and t2.mon>t1.mon
   group by t1.userid,t1.mon,t2.mon
) t
group by from_mon,to_mon;

Result (part)

from_mon	to_mon	usersqty
200801	200802	2
200801	200803	1
200801	200805	1
200801	200807	1
200801	200809	1
200801	200811	1
200801	200804	1
200801	200806	1
200801	200808	1
200801	200810	1
200801	200812	1
200802	200803	1
200802	200805	1
200802	200807	1

…
Updated2:
New query example. All month from first to last month (YYYYMM) of table Transactions included, regardless of whether the user had orders this month

with recursive allmonths as(
  select FROM_UNIXTIME(min(timestamp), "%Y%m") mn
      ,FROM_UNIXTIME(max(timestamp), "%Y%m") maxmn
  from Transactions
  union all
  select mn+1, maxmn from allmonths where mn<maxmn
)
,allmontomon as(
  select t1.mn from_mon,t2.mn to_mon
  from allmonths t1
  left join allmonths t2 on t2.mn>t1.mn
)
,tranmon as(
  SELECT userid
      ,FROM_UNIXTIME(timestamp, "%Y%m") mon
  FROM Transactions T1
  GROUP BY userid, FROM_UNIXTIME(timestamp, "%Y%m")
)
,um as(
select from_mon,to_mon
  ,case when to_mon is null then 0 else count(*) end usersqty
  ,group_concat(cast(t.userid as char)) lst
from(
   select t1.userid,t1.mon from_mon,t2.mon to_mon
   from tranmon t1
   left join tranmon t2
      on t2.userid=t1.userid and t2.mon>t1.mon
   group by t1.userid,t1.mon,t2.mon
) t
  group by from_mon,to_mon
)
select * 
from allmontomon c
left join um on c.from_mon=um.from_mon and c.to_mon=um.to_mon
order by c.from_mon,c.to_mon

Example here

Please signup or login to give your own answer.

Click here to cancel reply.