Find all rows that has a changed column value over time in SQL [closed]

Question

So I have three columns: Time_id, accounts and code. See attached picture for a sn-p of the data.

I want to find all account that have changed the code from 7100 to 7000 for that account, with ordered by the most recent ones.

• Time_id: is the date generated once every month for each account with updated fields. In the format of yyyymmdd
• account: is an unique account id for this customer
• code: is a four letter digit describing

I have tried a LAG of 'code' over a partition by time_id. However this returned a LAG code of from a different account. Not sure how to transform the query to only return changed code from LAG based on the same accounts. This was the query I tried:

SELECT time_id, account, code
    ,LAG(code, 1) OVER (partition by time_id order by time_id) LAG_1
  FROM my_table
  group by time_id, account, code

I was hoping to get all the accounts and the rows where code went from 7100 to 7000 and when it happened. For instance I want to get account 12500 and 15500 returned from the table below with the row when it changed from 7100:

time_id	account	code
20220510	12500	7100
20221101	12500	7000
20221120	12500	7000
20221201	17500	7100
20221202	12500	7100
20221203	15500	7100
20221204	15500	7000
20221205	15500	7000

I appreciate any new suggestions. Or improvements to my own query

Answer 1

Recall that PARTITION BY is the rolling window equivalent of a GROUP BY. Since you want to look at the changes in codefor a particular account, you need to partition by account:

    SELECT
        time_id,
        account,
        code,
        LAG(code) OVER (
          PARTITION BY account
          ORDER BY time_id
        ) AS previous_code
      FROM
        my_table

This query will return your original (time_id, account, code) records with a new column, previous_code, which denotes the value of code for the most recent preceding record for that account.

Now we can make progress:

WITH lagged_table AS (
    SELECT
        time_id,
        account,
        code,
        LAG(code) OVER (
          PARTITION BY account
          ORDER BY time_id
        ) AS previous_code
      FROM
        my_table
)

    SELECT
        time_id,
        account
      FROM
        lagged_table
     WHERE
           previous_code = 7100
       AND code = 7000

This should get you exactly what you want: an account ID and a time for all cases when the code switched from 7100 to 7000.