SQL/Bigquery 中的多重聚合

Question

I have a table with called user_state with columns user_ID, account_id, Balance, Date. Each user_id can have multiple account. Sample table looks like this: https://i.stack.imgur.com/w4odf.png you can create the table using script:

    CREATE TABLE USER_STATE (USER_ID int, ACCOUNT_ID int, SNAPSHOT_DATE DATE, BALANCE float);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,1 ,'2016-07-01', 50);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,2 ,'2016-07-01', 50);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,2 ,'2016-07-05', 80);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,1, '2016-07-27', 150);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,1,'2016-07-31', 200);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,1, '2016-08-18', 150);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,2,'2016-08-21', 250);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(23,1,'2016-08-21', 250);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,1, '2016-06-01', 10);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,2, '2016-06-01', 20);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44, 1,'2016-06-05', 40);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,1, '2016-06-27', 90);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,1, '2016-06-31', 300);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,1, '2016-09-18', 400);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,1, '2016-09-21', 200);
    INSERT INTO USER_STATE(USER_ID,ACCOUNT_ID,SNAPSHOT_DATE,BALANCE) VALUES(44,2, '2016-09-21', 200);

I would like to get this values per month per user:

• first_date of the month and the sum_balance for the first of the month per user.
• last_date of the month and sum_balance for the last date of that month per user.
• ave_balance for the each month per user.

Final Result should look like this: https://i.stack.imgur.com/Fyjes.png

One solution is to get the item I listed and I need per month per user separately and join the tables, Is there more efficient and easier solution for it?

Thanks, Bikram

Answer 1

Hmmm. . . aggregate by snapshot date and user and then by user and month:

select user_id, date_trunc(snapshot_date, month) as yyyymm,
       min(snapshot_date) as first_date,
       array_agg(balance order by snapshot_date limit 1)[ordinal(1)] as first_balance,
       max(snapshot_date) as first_date,
       array_agg(balance order by snapshot_date desc limit 1)[ordinal(1)] as last_balance,
       avg(balance)
from (select user_id, snapshot_date, sum(balance) as balance
      from user_state u
      group by 1, 2
     ) u
group by 1, 2;