SQL finding number of users in consecutive months in a table

Question

I have an table with the following columns

email        ----        created at
abc@gmail.com        2019-12-12 16:03:34
rp@gamil.com         2019-11-12 16:03:34
abc@gmail.com        2020-1-12 16:03:34
er@gmail.com         2020-1-12 16:03:34

I want to design a query that return the back number of emails that registered in consecutive 2 months. I am no novice with queries and have been struggling to come up with a query for this.

For the above the query abc@gmail.com was registered twice in consecutive months

Answer 1

You can use an EXISTS query to check if an email exists that also had a registration in the previous month:

SELECT DISTINCT email
FROM yourtable t1
WHERE EXISTS (SELECT *
              FROM yourtable t2
              WHERE t2.email = t1.email 
                AND DATE_FORMAT(t2.createdat, '%Y%m') = DATE_FORMAT(t1.createdat - INTERVAL 1 MONTH, '%Y%m'))

Output for your sample data

abc@gmail.com

Demo on dbfiddle

We use DISTINCT so we don't get multiple copies of the same email if an email address is registered in more than one consecutive month.

Answer 2

By doing a self-join for Month + 1 and email (and also taking December-to-January transitions into account) this should work:

SELECT
    *
FROM
    (
        SELECT
            email,
            YEAR( created ) AS createdYear,
            MONTH( created ) AS createdMonth
        FROM
            table
    ) AS t

    INNER JOIN
    (
        SELECT
            email,
            YEAR( created ) AS createdYear,
            MONTH( created ) AS createdMonth
        FROM
            table
    ) AS monthPlus1 ON 
        t.email = monthPlus1.email
        AND
        (
            (
                t.createdMonth = monthPlus1.createdMonth + 1
                AND
                t.createdYear = monthPlus1.createdYear
            )
            OR
            (
                t.createdMonth = 12
                AND
                monthPlus1.createdMonth = 1
                AND
                t.createdYear + 1 = monthPlus1.createdYear
            )
        )

The date logic in this query is a bit gnarly - it can probably be improved by representing the month as a single date value or integer months-since-epoc rather than a year + month tuple.

Answer 3

You can use lag() . If this occurs, then lag() will be in two adjacent months.

select t.email
from (select t.*
             lag(created_at) over (partition by t.email order by created_at) as prev_created_at
      from t
     ) t
where extract(year_month from created_at) = extract(year_month from (prev_created_at + interval 1 month));

You may need select distinct , if this can occur multiple times.