ringa_lee2017-04-17 16:21:52
select t.userid,t.score from (select * from active_gamescore order by Score desc,created desc) as t group by t.userid limit 20;
巴扎黑2017-04-17 16:21:52
更新:
一個人只能回覆一次,很憂傷
感謝幾位抽出時間去幫我解題
測試了一下(5次平均)
在1w資料量的情況下,
@clcx_1315 :0.004s
@伊拉克 : 0.009s
@邢愛明 :0.006s
我自己的 :0.016s
在20w的資料量下:
@clcx_1315 :0.104s
@伊拉克 : 0.141s
@邢愛明 :0.165s
我自己的 :0.171s
所以@clcx_1315的方法最優,十分感謝,學到了一個想法。
從explain看,@clcx_1315的寫法只做了2次全表遍歷,其他都是3次,或許這就是原因了。
===========================之前的分隔線================== ==========
琢磨了一種寫法,但效率有待提高
select ta.user_id,ta.max_score,tb.min_time
from (
select a.user_id, max(a.score) max_score from active_gamescore a where a.active_id='58' group by a.user_id
) ta
join (
select a.user_id,a.score,min(a.created_at) min_time from active_gamescore a where a.active_id='58' group by a.user_id,a.score
) tb
on ta.user_id=tb.user_id and ta.max_score=tb.score
order by ta.max_score desc,tb.min_time asc
limit 20
高洛峰2017-04-17 16:21:52
假設同一使用者下的created_at欄位值不重複,可以試試看下面的語句:
SELECT t1.user_id, t1.score, t1.created_at
FROM (
SELECT
@row_num:=IF(@prev_col1=t.user_id, @row_num+1, 1) AS row_number,
t.*,
@prev_col1:=t.user_id
FROM (SELECT * FROM active_gamescore ORDER BY user_id, score DESC, created_at) t,
(SELECT @row_num:=1, @prev_col1:=NULL) var
) t1 WHERE row_number = 1
ORDER BY t1.score DESC, t1.created_at
LIMIT 20