运行排除具有重复列值的行的总和

Question

一个例子table:

video	encoding	video time spent	encoding bytes	encoding bytes running sum	video time spent running sum (expected)	video time spent running sum (actual)
A	1	1	500	500	1	1
A	2	1	400	900	1	2
B	3	2	300	1200	3	5
B	4	2	200	1400	3	8
B	5	2	100	1500	3	11
B	6	2	100	1600	3	14

video time spent 列有一个视频被观看了多少时间；观看了哪种编码并不重要。
花费的视频时间运行总和是我想要得到的。它应该只计算在视频级别花费的时间，忽略编码。

我想 select 尽可能多地编码字节，同时保持在花费的视频时间总和 < X 以内。

我目前的查询：

SELECT *
FORM (
   SELECT 
      ...,
      SUM(encoding_bytes) OVER(ORDER BY encoding_bytes desc) AS encoding_bytes_running_sum, 
      SUM(video_time_spent) OVER (ORDER BY encoding_bytes desc) AS video_time_spent_running_sum
      ...
) 
WHERE video_time_spent_running_sum < X

但 video_time_spent_running_sum 不够聪明，无法跳过同一视频中的其他编码。执行此操作的最佳方法是什么？

每个视频的编码数量不是恒定的。

创建脚本 table:

SELECT
    *,
    SUM(encoding_bytes) OVER(
        ORDER BY
            encoding_bytes DESC
    ) AS encoding_bytes_running_sum,
    SUM(video_time_spent) OVER (
        ORDER BY
            encoding_bytes DESC ROWS UNBOUNDED PRECEDING
    ) AS video_time_spent_running_sum
FROM (
    VALUES
        ('a', 1, 1, 500),
        ('a', 2, 1, 400),
        ('b', 3, 2, 300),
        ('b', 4, 2, 200),
        ('b', 5, 2, 100),
        ('b', 6, 2, 100)
) AS t (video, encoding, video_time_spent, encoding_bytes)

Answer 1

一种方法如下（我相信它可以简化）；您使用 ROW_NUMBER 函数只计算每个视频的第一行。

WITH cte AS (
    SELECT
        *
        , SUM(encoding_bytes) OVER (ORDER BY encoding_bytes DESC) AS encoding_bytes_running_sum
        --, SUM(video_time_spent) OVER (ORDER BY encoding_bytes DESC ROWS UNBOUNDED PRECEDING) AS video_time_spent_running_sum
        , ROW_NUMBER() OVER (PARTITION BY video ORDER BY video, [encoding]) rn
    FROM (
        VALUES
            ('a', 1, 1, 500),
            ('a', 2, 1, 400),
            ('b', 3, 2, 300),
            ('b', 4, 2, 200),
            ('b', 5, 2, 100),
            ('b', 6, 2, 100)
    ) AS t (video, [encoding], video_time_spent, encoding_bytes)
)
SELECT video, [encoding], video_time_spent, encoding_bytes, encoding_bytes_running_sum
    , SUM(CASE WHEN rn = 1 THEN video_time_spent ELSE 0 END) OVER (ORDER BY video ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) video_time_spent_running_sum
FROM cte;

这个returns:

video encoding	video_time_spent	encoding_bytes	encoding_bytes_running_sum	video_time_spent_running_sum
a	1	1	500	500
a	2	1	400	900
b	3	2	300	1200
b	4	2	200	1400
b	5	2	100	1600
b	6	2	100	1600

运行排除具有重复列值的行的总和

Running sum excluding rows with duplicate column value

sql

tsql

sql-server

window-functions

运行 排除具有重复列值的行的总和

Running sum excluding rows with duplicate column value

sql

tsql

sql-server

window-functions

运行排除具有重复列值的行的总和