postgres：SQL和條紋？

試圖找出填充此表最後一列的最快方法，streak列存儲運行結果值（win或loss）。postgres：SQL和條紋？

game_date | team | result | streak 
------------+---------+--------+-------- 
2017-10-05 | ABC  | win | 1 
2017-10-07 | ABC  | win | 2 
2017-10-11 | ABC  | loss | 1 
2017-10-14 | ABC  | win | 1 
2017-10-15 | ABC  | win | 2 
2017-10-18 | ABC  | win | 3 
2017-10-21 | ABC  | win | 4 
2017-10-23 | ABC  | loss | 1 
2017-10-24 | ABC  | win | 1 
2017-10-26 | ABC  | win | 2 
2017-10-28 | ABC  | win | 3 
2017-10-30 | ABC  | loss | 1 
2017-11-02 | ABC  | win | 1 
2017-11-04 | ABC  | loss | 1 
2017-11-07 | ABC  | win | 1 
2017-11-09 | ABC  | loss | 1 
2017-11-12 | ABC  | loss | 1 
2017-11-14 | ABC  | loss | 2 
2017-11-16 | ABC  | loss | 3 
2017-11-18 | ABC  | win | 1 
2017-11-19 | ABC  | loss | 1 
2017-11-22 | ABC  | loss | 2 
2017-11-24 | ABC  | loss | 3 
2017-11-25 | ABC  | win | 1 
2017-11-28 | ABC  | win | 2 
2017-11-30 | ABC  | win | 3 
2017-12-01 | ABC  | win | 4 
(27 rows)

正如你看到的，streak應該重置爲1時，先結果爲空，或者不同的，否則一個遞增。

來源

2017-12-02 Wells

您可以通過使用row_numbers()的區別來識別結果值的「孤島」，從而識別獲勝和丟失的序列。其餘僅僅是最終的應用row_number()：

select g.*, 
     row_number() over (partition by seqnum - seqnum_r, result order by game_date) as streak 
from (select g.*, 
      row_number() over (order by game_date) as seqnum, 
      row_number() over (partition by result order by game_date) as seqnum_r 
     from games g 
    ) g

來源

2017-12-03 00:12:46

postgres：SQL和條紋？

回答

相關問題