从一列获取MAX值，从另一列获取MIN

Question

I've been building a game based on Candy Crush. The Score table has the three following columns:

stage_level_id                        | value | moves
------------------------------------------------------
9f7678f0-fc8f-11e3-a398-b2227cce2b53  | 35000 | 350
9f7678f0-fc8f-11e3-a398-b2227cce2b53  | 35000 | 500
9f7678f0-fc8f-11e3-a398-b2227cce2b54  | 15000 | 125
9f7678f0-fc8f-11e3-a398-b2227cce2b54  | 13500 | 100
9f7678f0-fc8f-11e3-a398-b2227cce2b55  | 12500 | 350
9f7678f0-fc8f-11e3-a398-b2227cce2b55  | 7500  | 25

I need to get the top Score grouped by stage_level_id . If an stage_level_id have the same Value (as the one ending with 53), it must return the row with the smallest number of Moves .

I'm trying the following but it's not working as expected:

SELECT a.stage_level_id, MAX(a.value) as max_value, a.moves
FROM scores a
LEFT JOIN scores b ON (
  a.stage_level_id = b.stage_level_id
)
RIGHT JOIN scores c ON (
  c.moves = ( SELECT MIN(moves) as moves FROM scores WHERE c.stage_level_id =         a.stage_level_id )
)
WHERE a.player_id = 1475332386040815
GROUP BY a.stage_level_id

The expected result is:

stage_level_id                        | value | moves
------------------------------------------------------
9f7678f0-fc8f-11e3-a398-b2227cce2b53  | 35000 | 350
9f7678f0-fc8f-11e3-a398-b2227cce2b54  | 15000 | 125
9f7678f0-fc8f-11e3-a398-b2227cce2b55  | 12500 | 350

What I'm doing wrong?

Answer 1

Your attempt wasn't that far off. You were missing a necessary part of the first JOIN ... ON clause though, and the second JOIN isn't necessary.

SELECT tbl1.stage_level_id, tbl1.max_value, MIN(s.moves) AS moves
FROM 
(
  SELECT stage_level_id, MAX(value) AS max_value
  FROM scores
  GROUP BY stage_level_id
) tbl1
LEFT JOIN scores s ON tbl1.stage_level_id = s.stage_level_id AND tbl1.max_value = s.value
GROUP BY stage_level_id

DEMO

Answer 2

You can get the minimum number of moves for each (stage_id,max score) group using NOT EXISTS

select stage_level_id, max(value), min(moves)
from scores s1
where not exists (
    select 1 from scores s2
    where s2.stage_level_id = s1.stage_level_id
    and s2.value > s1.value
)
group by stage_level_id;

The not exists part limits the results to only those rows that have the maximum score within each group (in other words no other row with a higher score within the group exists).

This query can take advantage of a composite index on (stage_id,value)

http://sqlfiddle.com/#!2/88ee6/8

Answer 3

In fact, this can be done simply by one select with a left join and a group by :

select s1.stage_level_id, s1.score, min(s1.moves) as moves
from scores s1
  left join scores s2
    on s2.stage_level_id = s1.stage_level_id
    and s2.score > s1.score
where s2.score IS NULL
group by s1.stage_level_id;

The left join gives:

| s1                                                   | s2                                                     |
|--------------------------------------|-------|-------|--------------------------------------|--------|--------|
| stage_level_id                       | score | moves | stage_level_id                       | score  | moves  |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b53 | 35000 | 350   | (null)                               | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b53 | 35000 | 500   | (null)                               | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b54 | 15000 | 125   | (null)                               | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b54 | 13500 | 100   | 9f7678f0-fc8f-11e3-a398-b2227cce2b54 | 15000  | 125    |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b55 | 7500  | 25    | 9f7678f0-fc8f-11e3-a398-b2227cce2b55 | 12500  | 350    |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b55 | 12500 | 350   | (null)                               | (null) | (null) |

The where clause selects the rows that contain the highest score(s) per stage_level_id :

| s1                                                   | s2                               |
|--------------------------------------|-------|-------|----------------|--------|--------|
| stage_level_id                       | score | moves | stage_level_id | score  | moves  |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b53 | 35000 | 350   | (null)         | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b53 | 35000 | 500   | (null)         | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b54 | 15000 | 125   | (null)         | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b55 | 12500 | 350   | (null)         | (null) | (null) |

so that we can run min(s1.moves) on each stage_level_id group formulated by group by to get the final result:

| s1                                                   | s2                               |
|--------------------------------------|-------|-------|----------------|--------|--------|
| stage_level_id                       | score | moves | stage_level_id | score  | moves  |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b53 | 35000 | 350   | (null)         | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b54 | 15000 | 125   | (null)         | (null) | (null) |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b55 | 12500 | 350   | (null)         | (null) | (null) |

or

| stage_level_id                       | score | moves |
|--------------------------------------|-------|-------|
| 9f7678f0-fc8f-11e3-a398-b2227cce2b53 | 35000 | 350   |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b54 | 15000 | 125   |
| 9f7678f0-fc8f-11e3-a398-b2227cce2b55 | 12500 | 350   |

Test it out at http://sqlfiddle.com/#!9/50ed8/6/0 .

Answer 4

Not sure if I misunderstood the question but isnt it as simple as this:

SELECT column1, max(column2), IF(column1 like "%53", min(column3), max(column3))
FROM test
GROUP BY column1
;

Check this out, SQLFiddle link