如何针对这种情况编写更好的SQL

Question

I have a table called " DEVICE " which contains these columns:

------------------------------------------
SN | User_ID |  State    | ... 
------------------------------------------
1 |  1001   | deployed  | ...
2 |  1001   | deploying | ...
3 |  1002   | inventory | ...
4 |  1003   | deploying | ...
5 |  1001   | deploying | ...
6 |  1002   | synced    | ...
7 |  1002   | synced    | ...
8 |  1010   | synced    | ...
9 |  1008   | unsynced  | ...
---------------------------------------

As you can see, this table lists all the devices, which are belong to different user by their user_id. Also for all these devices, their state is one of these 5 types:

"inventory", 
"deployed", 
"deploying", 
"synced"
"unsynced". 

I need a SQL Query that returns back the user_id that holds the maximum devices of each individual State. For example, for the above set of data in the first table, the return would be:

-----------------------------------------------------------
User_ID     |    State    |   Maximum 
-----------------------------------------------------------
1001        |    deployed |     1 
1001        |    deploying|     2
1002        |    inventory|     1
1002        |    synced   |     2 
1008        |    unsynced |     1 
------------------------------------------------------------

The SQL I have created so far is like this:

SELECT 
    STATE, ACCOUNT, CNT 
from 
( select account_id as ACCOUNT, state as STATE, count(*) as CNT 
  from DEVICE group by account_id, state ) T1
where  
    ( STATE, CNT ) 
IN 
    ( select STATE, MAX(CNT) 
from 
( select account_id as ACCOUNT, state as STATE, count(*) as CNT 
   from DEVICE group by account_id, state ) T2 
Group by STATE
);

which is so ugly. Is there a better way to do that ?

Thanks,

Jack

Answer 1

You can create an intermediate summary table or a view like so:

drop table if exists device_summary;

create table device_summary as
    select user_id, state, count(*) as counter
    from device
    group by user_id, state;

Then, you can run a query like so to get your desired results:

select d.*
from device_summary d
inner join (
    select state, max(counter) as maxcounter
    from device_summary
    group by state
) d1 on d.state = d1.state and d.counter = d1.maxcounter;

Example: https://rextester.com/ZMZH93648

Answer 2

You can make use of ROW_NUMBER analytical function as follows:

SELECT
    B.USER_ID,
    B.STATE,
    B.CNT
FROM
    (
        SELECT
            ROW_NUMBER() OVER(
                PARTITION BY A.STATE
                ORDER BY
                    A.CNT DESC
            ) AS RN,
            A.USER_ID,
            A.STATE,
            A.CNT
        FROM
            (
                SELECT
                    USER_ID,
                    STATE,
                    COUNT(1) CNT
                FROM
                    DEVICE
                GROUP BY
                    USER_ID,
                    STATE
            ) A
    ) B
WHERE
    B.RN = 1

db<>fiddle demo

Cheers!!

Answer 3

Try this:

SELECT User_ID, State, count(*) AS Maximum
FROM DEVICE
GROUP BY User_ID, State
ORDER BY User_ID, State

Answer 4

You can use window functions, but only need one level of subquery is needed:

SELECT USER_ID, STATE, CNT
FROM (SELECT USER_ID, STATE, COUNT(*) as CNT,
             ROW_NUMBER() OVER (PARTITION BY state ORDER BY COUNT(*) DESC) as seqnum
      FROM DEVICE
      GROUP BY USER_ID, STATE
     ) us
WHERE seqnum = 1;

Note: If there are ties, this returns an arbitrary user. If you want all of them, use RANK() instead of ROW_NUMBER() .

Note: Window functions are only available in MySQL starting with version 8.