如何选择两列的最大值并按工位分组

Question

I have a simple table like this

 station    num     tim
   1        150    10
   1       200     222
   1       100     5000
   1       200     555
   2       100     500
   2       120     200
   3        1      2

The output desired is like this

station num     tim
   1    200     555
   2    120     200
   3    1       2

i wrote this code but for station=1 and num=200 return two rows

(SELECT a.station , a.num ,a.tim
FROM test.dbo.tst a
JOIN (
SELECT station, MAX(num) num
FROM test.dbo.tst
GROUP BY station
) b ON a.station = b.station and a.num=b.num ) order by station 

Answer 1

One possible approach is to use ROW_NUMBER() to number rows grouped by station and ordered by num and tim descending and then select the rows with number equal to 1.

Input:

CREATE TABLE #Stations (
   station int,
   num int,
   tim int
)
INSERT INTO #Stations
   (station, num, tim)
VALUES
   (1, 150, 10),
   (1, 200, 222),
   (1, 100, 5000),
   (1, 200, 555),
   (2, 100, 500),
   (2, 120, 200),
   (3, 1,   2)

Statement:

;WITH cte AS (   
   SELECT 
      station, 
      num, 
      tim,
      ROW_NUMBER() OVER (PARTITION BY station ORDER BY num DESC, tim DESC) AS Rn
   FROM #Stations
)
SELECT 
   station, 
   num, 
   tim
FROM cte
WHERE Rn = 1

Output:

station num tim
1       200 555
2       120 200
3       1   2

Answer 2

Update 1

The recommended method is the one mentioned by @Zhorov, but i will leave my answer to give all possible solutions

---

Initial answer

Try using a similar query:

SELECT T.station, T.maxnum, MAX(T.tim) as maxtim 
FROM (SELECT station, tim, MAX(num) as maxnum
      GROUP BY station, tim ) T
GROUP BY station, maxnum

OR

WITH CTE_1 as (SELECT station, tim, MAX(num) as maxnum
               GROUP BY station, tim)
SELECT station, maxnum, Max(tim)
FROM CTE 
GROUP BY station, tim

Answer 3

You could use the ROW_NUMBER partition function, something along these lines should do the trick:

SELECT  [Station], [Num], [Tim]
FROM
(
    SELECT  Station,
            Num,
            Tim,
            ROW_NUMBER() OVER(PARTITION BY Station ORDER BY Num DESC, Tim DESC) [N]
    FROM    tst
) SQ
WHERE    [N] = 1

This works on the basis of taking each group of rows, as grouped by the Station column (this is what the PARTITION BY part of the ROW_NUMBER() statement refers to), then ordering them first by Num and then by Tim and assigning each row in the result set a row number. The outer query then takes only the first row of each group, ie for each value of Station it takes the one with the highest value for Num and Tim .