如何获得每个组的最大行数？

Question

I have a database in which people submit votes for people in different places. At a given time, I want to find out who has the most votes at each place. (A person can be voted at two different places)

This is the SQL I have so far:

SELECT placeId, userVotedId, cnt 
FROM 
    (SELECT uvo.userVotedId, p.placeId, count(*) as cnt 
     FROM users as u, users_votes as uvo, places as p 
     WHERE u.userId = uvo.userVotedId 
       AND p.placeId = uvo.placeId 
     GROUP BY userVotedId, placeId) 
AS RESULT

which gives me this result:

Now, these are the rows I REALLY want:

What's missing in my query so I can get this?

• I want one result per place. So I should see only distinct placeIds, with the userVotedId who received the most votes.

• In the event of a tie, a random winner will do!

Answer 1

Seems like you need one more aggregate. Use the MAX() aggregate on your cnt value and GROUP BY placeId, userVotedId :

 
 
 
  
  SELECT placeId, userVotedId, max(cnt) FROM ( SELECT uvo.userVotedId, p.placeId, count(*) as cnt FROM users as u INNER JOIN users_votes as uvo ON u.userId = uvo.userVotedId INNER JOIN places as p ON p.placeId = uvo.placeId GROUP BY userVotedId, placeId ) AS RESULT GROUP BY placeId, userVotedId
 
  

Note: I changed your query to use JOIN syntax instead of the commas between the tables.

Edit, based on your comment the following should work:

select total.uservotedid,
  total.placeid,
  total.cnt
from
(
  SELECT uvo.userVotedId, p.placeId, count(*) as cnt 
  FROM users as u
  INNER JOIN users_votes as uvo
    ON u.userId = uvo.userVotedId 
  INNER JOIN places as p 
    ON p.placeId = uvo.placeId 
  GROUP BY userVotedId, placeId
) total
inner join
(
  select max(cnt) Mx, placeid
  from
  (
    SELECT uvo.userVotedId, p.placeId, count(*) as cnt 
    FROM users as u
    INNER JOIN users_votes as uvo
      ON u.userId = uvo.userVotedId 
    INNER JOIN places as p 
      ON p.placeId = uvo.placeId 
    GROUP BY userVotedId, placeId
  ) mx
  group by placeid
) src
  on total.placeid = src.placeid
  and total.cnt = src.mx

See SQL Fiddle with Demo

The result is:

| USERVOTEDID | PLACEID | CNT |
-------------------------------
|          65 |      11 |   1 |
|          67 |      13 |   1 |
|          67 |      25 |   1 |
|          67 |      51 |   2 |

Edit #2, if you want a random number returned if there is a tie, then you can use user variables:

select uservotedid,
  placeid, 
  cnt
from
(
  select total.uservotedid,
    total.placeid,
    total.cnt,
    @rownum := case when @prev = total.placeid then @rownum+1 else 1 end rownum,
    @prev := total.placeid pplaceid
  from
  (
    SELECT uvo.userVotedId, p.placeId, count(*) as cnt 
    FROM users as u
    INNER JOIN users_votes as uvo
      ON u.userId = uvo.userVotedId 
    INNER JOIN places as p 
      ON p.placeId = uvo.placeId 
    GROUP BY userVotedId, placeId
  ) total
  inner join
  (
    select max(cnt) Mx, placeid
    from
    (
      SELECT uvo.userVotedId, p.placeId, count(*) as cnt 
      FROM users as u
      INNER JOIN users_votes as uvo
        ON u.userId = uvo.userVotedId 
      INNER JOIN places as p 
        ON p.placeId = uvo.placeId 
      GROUP BY userVotedId, placeId
    ) mx
    group by placeid
  ) src
    on total.placeid = src.placeid
    and total.cnt = src.mx
  order by total.placeid, total.uservotedid
) src
where rownum = 1
order by placeid, uservotedid

See SQL Fiddle with Demo

Answer 2

SELECT placeId, userVotedId, MAX(cnt)
    FROM (SELECT uvo.userVotedId, p.placeId, count(*) AS cnt 
          FROM users as u, users_votes as uvo, places as p 
          WHERE u.userId = uvo.userVotedId AND p.placeId = uvo.placeId 
          GROUP BY userVotedId, placeId) AS RESULT
    GROUP BY  placeId

• Similar Question - SQL query max(), count()

• Demo

Answer 3

For simplicity, I called your query Test:

SELECT * 
FROM Test T JOIN (
SELECT t.placeId, Max(t.cnt) maxcnt
FROM Test t
GROUP BY t.placeId) T2 ON T.placeId = T2.placeId and T.cnt = T2.maxcnt

Here is the Fiddle .

BTW -- Test = :

SELECT uvo.userVotedId, p.placeId, count(*) as cnt 
  FROM users as u
  INNER JOIN users_votes as uvo
    ON u.userId = uvo.userVotedId 
  INNER JOIN places as p 
    ON p.placeId = uvo.placeId 
  GROUP BY userVotedId, placeId

Good luck.

--EDIT -- as requested, here is the final code:

SELECT * 
FROM (SELECT uvo.userVotedId, p.placeId, count(*) AS cnt 
          FROM users as u, users_votes as uvo, places as p 
          WHERE u.userId = uvo.userVotedId AND p.placeId = uvo.placeId 
          GROUP BY userVotedId, placeId) T JOIN (
SELECT t.placeId, Max(t.cnt) maxcnt
FROM (SELECT uvo.userVotedId, p.placeId, count(*) AS cnt 
          FROM users as u, users_votes as uvo, places as p 
          WHERE u.userId = uvo.userVotedId AND p.placeId = uvo.placeId 
          GROUP BY userVotedId, placeId) t
GROUP BY t.placeId) T2 ON T.placeId = T2.placeId and T.cnt = T2.maxcnt