开始触发器和结束触发器之间的GROUP行

Question

如果你能帮我解决我的问题，那将会很酷。我有一张桌子，你可以从游戏中看到事件。每个玩家都有一定数量的黄金，你可以在Log in之后的Value列中看到。我需要找出在Log in之后有多少黄金玩家开始了他们的FIRST游戏。之前玩家开始游戏，他们可以Buy gold或Get bonus将由Value

**Player ID  Timestamp  Action  Value**  
    1111   09:11  Log in  500 
    1111   09:25  Buy gold 100 
    1111   09:28  Get bonus 50 
    1111   09:30  Start game  
    2222   11:14  Log in  800 
    3333   12:01  Log in  700 
    3333   12:04  Get bonus 50 
    3333   12:08  Start game 
    3333   12:15  Buy gold 100 
    3333   12:18  Start game 
    1111   14:40  Log in  300 
    1111   14:42  Buy gold 100 
    1111   14:50  Start game 
    2222   15:22  Log in  600 
    2222   15:25  Buy gold 100 
    2222   16:25  Log in  400 
    2222   16:30  Get bonus 50 
    2222   16:35  Start game  

特结果我要寻找提高他们的黄金量是类似的东西。它基本上是玩家的Start game和最近的Log in之间的值的总和。

**Match number Player ID  Value** 
    1    1111   650 
    2    3333   750 
    3    1111   400 
    4    2222   450 

谢谢！

Answer 1

一种方法是搜索每个Log in行。然后查找下一行，即Log in或Start game。如果该行是Start game，则表示登录后可以识别第一个游戏。然后，您可以找到值的总和为对之间的所有行：

select row_number() over (order by start.Timestamp) as MatchNumber 
,  start.[Player ID] 
,  vals.SumValue 
from YourTable login 
cross apply 
     (
     select top 1 * 
     from YourTable start 
     where login.[Player ID] = start.[Player ID] 
       and login.Timestamp < start.Timestamp 
       and start.Action in ('Start game', 'Log in') 
     order by 
       start.Timestamp 
     ) start 
cross apply 
     (
     select sum(cast(vals.Value as int)) as SumValue 
     from YourTable vals 
     where vals.[Player ID] = start.[Player ID] 
       and vals.Action in ('Log in', 'Get bonus', 'Buy gold') 
       and vals.Timestamp between login.Timestamp and start.Timestamp 
     ) vals 
where login.Action = 'Log in' 
     and start.Action = 'Start game' 

Example at SQL Fiddle.

Answer 2

你可以尝试一些相关子查询：

SELECT PlayerID, SUM(value) FROM table a 
WHERE 
TimeStamp >= (SELECT MAX(TimeStamp) FROM table WHERE a.PlayerID = PlayerID AND Action = 'Log in') 
GROUP BY PlayerID 

Answer 3

这是一个有点复杂。我的做法是为每次登录找到下一个时间戳，即“登录”或“开始游戏”，然后返回到原始表以查找所有金条目。

select tli.playerId, tli.TimeStamp, sum(value) 
from (select t.*, 
      (select top 1 timestamp 
       from t t2 
       where t2.playerId = t.playerId and 
        t2.timestamp > t.timestamp and 
        t2.action in ('log in', 'start game') 
       order by timestamp 
      ) as nextTimestamp 
     from t 
     where t.action in ('log in') 
    ) tli join 
    t 
    on t.playerId = tli.playerId and 
     t.timestamp >= tli.timestamp and 
     t.timestamp < tli.nextTimeStamp 
group by tli.playerId, tli.timeStamp