在多个orderby参数postgres中获取用于排序的列

问题描述 投票：0回答：1

我使用下面的查询从表 Schedule 创建物化视图 Schedule_state 

CREATE MATERIALIZED VIEW schedule_state AS (
    WITH schedule_latest_events AS (
        SELECT
            *,
            row_number() over (
                PARTITION BY key_id
                ORDER BY
                    header_event_timestamp DESC,
                    ktimestamp DESC,
                    raw_load_timestamp DESC,
                    update_timestamp DESC
            ) AS row_number
        FROM
            schedule
    )
    SELECT
        *
    FROM
        schedule_latest_events
    WHERE
        row_number = 1
);

由于我使用多列进行排序，有什么方法可以找出哪一列用于对排名为 1 的列进行排序？

还可以获取四个按列排序的最大值并将结果分配给物化视图中的列吗？ 谢谢！

Table schedule

key_id,   header_event_timestamp,              ktimestamp,             raw_load_timestamp,         update_timestamp
k1,    2023-12-22 08:50:59.930000,  2023-12-22 08:50:59.930000, 2023-12-22 08:52:36.960000, 2023-12-22 08:50:58.100000
k1,    2023-12-22 08:50:37.530000,  2023-12-22 08:50:37.530000, 2023-12-22 08:52:36.960000, 2023-12-22 06:41:02.483000
k2,    2023-12-22 06:41:03.080000,  2023-12-22 06:41:03.080000, 2023-12-22 06:52:33.189000, 2023-12-22 06:41:02.483000
k2,    2023-12-22 06:41:03.080000,  2023-12-22 06:45:03.060000, 2023-12-22 06:52:33.189000, 2023-12-22 06:41:02.483000
k2,    2023-12-22 06:41:03.080000,  2023-12-20 15:09:36.370000, 2023-12-21 02:52:37.557000, 2023-12-20 14:04:37.323000
K3,    2023-12-20 14:39:00.909832,  2023-12-20 14:39:00.909000, 2023-12-21 02:52:37.557000, 2023-12-20 14:04:37.323000
K3,    2023-12-20 14:39:00.909832,  2023-12-20 14:39:00.909000, 2023-12-21 02:55:37.557000, 2023-12-20 14:04:37.323000
k4,    2023-12-22 06:41:03.080000,  2023-12-22 06:41:03.080000, 2023-12-22 06:52:33.189000, 2023-12-22 06:41:02.483000
k4,    2023-12-22 06:41:03.080000,  2023-12-22 06:41:03.080000, 2023-12-22 05:52:33.189000, 2023-12-22 06:41:02.483000
k4,    2023-12-22 06:41:03.080000,  2023-12-22 05:41:03.080000, 2023-12-22 05:52:33.189000, 2023-12-22 06:41:02.483000
k4,    2023-12-22 05:41:03.080000,  2023-12-22 05:41:03.080000, 2023-12-22 05:52:33.189000, 2023-12-22 06:41:02.483000

MATERIALIZED VIEW schedule_state

key_id,   header_event_timestamp,              ktimestamp,             raw_load_timestamp,         update_timestamp         row_number
k1,    2023-12-22 08:50:59.930000,  2023-12-22 08:50:59.930000, 2023-12-22 08:52:36.960000, 2023-12-22 08:50:58.100000          1
k2,    2023-12-22 06:41:03.080000,  2023-12-22 06:45:03.060000, 2023-12-22 06:52:33.189000, 2023-12-22 06:41:02.483000          1
K3,    2023-12-20 14:39:00.909832,  2023-12-20 14:39:00.909000, 2023-12-21 02:55:37.557000, 2023-12-20 14:04:37.323000          1
k4,    2023-12-22 06:41:03.080000,  2023-12-22 06:41:03.080000, 2023-12-22 06:52:33.189000, 2023-12-22 06:41:02.483000          1

在表中

  • 对于 key_id k1，第一个参数 header_event_timestamp 不同，因此最新 header_event_timestamp 的记录被分配排名 1。
  • 对于 key_id k2，第一个参数 header_event_timestamp 是相同的，因此根据第二个参数 ktimestamp 进行排序，因为 key k2 的第二个参数 ktimestamp 不同，并且具有最新 ktimestamp 的记录被分配排名 1。
  • 对于 key_id k3，第一个参数 header_event_timestamp 和第二个参数 ktimestamp 相同，因此根据第三个参数 raw_load_timestamp 进行排序，因为 key k3 的第三个参数 raw_load_timestamp 不同，并且具有最新 raw_load_timestamp 的记录被分配排名 1。
sql postgresql materialized-views
1个回答
0
投票

可以确定定义顺序的多个列中的第一列，以区分前 2 条记录。

我已经准备好了下面的完整示例，我立即评论：

  • 为了便于说明，我添加了一个 
    Schedule
    CTE，您需要将其删除才能获得真正的查询。
  • 有 1 条带有 
    key_id = ksingle
    的记录，以说明当一条记录单独存在时会发生什么情况。
  • 带有 
    key_id = kequal
    的两条记录是相同的，以说明完美平局时会发生什么。
  • 查询使用2个相似的窗口。我将它们的定义移到了查询的 
    WINDOW
    子句中，以避免重复 + 为了清楚起见，将它们放在一起。
  • 下面的查询实现了两种方法来告诉您区分列：
    • 4 列返回布尔值，用于比较订单中使用的每一列与同一列的下一个值（窗口函数 
      nth_value
      ，带参数 
      2
      ）。
    • 由与上面相同的布尔值构造的数组，应用了 
      array_position
      ，返回第一次出现的 
      false
      。
WITH Schedule(key_id, header_event_timestamp, ktimestamp, raw_load_timestamp, update_timestamp) AS ( VALUES
('ksingle', '2023-12-22 08:50:59.930000',  '2023-12-22 08:50:59.930000', '2023-12-22 08:52:36.960000', '2023-12-22 08:50:58.100000'),
('kequal',  '2023-12-22 08:50:59.930000',  '2023-12-22 08:50:59.930000', '2023-12-22 08:52:36.960000', '2023-12-22 08:50:58.100000'),
('kequal',  '2023-12-22 08:50:59.930000',  '2023-12-22 08:50:59.930000', '2023-12-22 08:52:36.960000', '2023-12-22 08:50:58.100000'),
('k1',      '2023-12-22 08:50:59.930000',  '2023-12-22 08:50:59.930000', '2023-12-22 08:52:36.960000', '2023-12-22 08:50:58.100000'),
('k1',      '2023-12-22 08:50:37.530000',  '2023-12-22 08:50:37.530000', '2023-12-22 08:52:36.960000', '2023-12-22 06:41:02.483000'),
('k2',      '2023-12-22 06:41:03.080000',  '2023-12-22 06:41:03.080000', '2023-12-22 06:52:33.189000', '2023-12-22 06:41:02.483000'),
('k2',      '2023-12-22 06:41:03.080000',  '2023-12-22 06:45:03.060000', '2023-12-22 06:52:33.189000', '2023-12-22 06:41:02.483000'),
('k2',      '2023-12-22 06:41:03.080000',  '2023-12-20 15:09:36.370000', '2023-12-21 02:52:37.557000', '2023-12-20 14:04:37.323000'),
('k3',      '2023-12-20 14:39:00.909832',  '2023-12-20 14:39:00.909000', '2023-12-21 02:52:37.557000', '2023-12-20 14:04:37.323000'),
('k3',      '2023-12-20 14:39:00.909832',  '2023-12-20 14:39:00.909000', '2023-12-21 02:55:37.557000', '2023-12-20 14:04:37.323000'),
('k4',      '2023-12-22 06:41:03.080000',  '2023-12-22 06:41:03.080000', '2023-12-22 06:52:33.189000', '2023-12-22 06:41:02.483000'),
('k4',      '2023-12-22 06:41:03.080000',  '2023-12-22 06:41:03.080000', '2023-12-22 05:52:33.189000', '2023-12-22 06:41:02.483000'),
('k4',      '2023-12-22 06:41:03.080000',  '2023-12-22 05:41:03.080000', '2023-12-22 05:52:33.189000', '2023-12-22 06:41:02.483000'),
('k4',      '2023-12-22 05:41:03.080000',  '2023-12-22 05:41:03.080000', '2023-12-22 05:52:33.189000', '2023-12-22 06:41:02.483000')
), schedule_latest_events AS (
    SELECT *,
           ROW_NUMBER()                         OVER w1 AS row_number,
           nth_value(header_event_timestamp, 2) OVER w2 AS next_header_event_timestamp,
           nth_value(ktimestamp, 2)             OVER w2 AS next_ktimestamp,
           nth_value(raw_load_timestamp, 2)     OVER w2 AS next_raw_load_timestamp,
           nth_value(update_timestamp, 2)       OVER w2 AS next_update_timestamp
    FROM schedule
    WINDOW
    w1 AS (PARTITION BY key_id ORDER BY header_event_timestamp DESC, ktimestamp DESC, raw_load_timestamp DESC, update_timestamp DESC),
    w2 AS (PARTITION BY key_id ORDER BY header_event_timestamp DESC, ktimestamp DESC, raw_load_timestamp DESC, update_timestamp DESC ROWS BETWEEN CURRENT ROW AND UNBOUNDED FOLLOWING)
)
SELECT *,
       header_event_timestamp = next_header_event_timestamp AS test_header_event_timestamp,
       ktimestamp             = next_ktimestamp             AS test_ktimestamp,
       raw_load_timestamp     = next_raw_load_timestamp     AS test_raw_load_timestamp,
       update_timestamp       = next_update_timestamp       AS test_update_timestamp,
       array_position(ARRAY[
           header_event_timestamp = next_header_event_timestamp,
           ktimestamp             = next_ktimestamp,
           raw_load_timestamp     = next_raw_load_timestamp,
           update_timestamp       = next_update_timestamp
       ], 'false') AS first_different_column
FROM schedule_latest_events
WHERE row_number = 1

由于我进行查询是为了说明表示您想要的结果的几种方法，因此上述查询将返回给您的内容存在一些冗余。我让你删除不想保留的列。

最新问题
© www.soinside.com 2019 - 2024. All rights reserved.