当此标识符不再存在于数据库中时，特定标识符缺少日期而不添加额外日期 SQL

Question

用语言表达问题，我有一个庞大的 table，其中包括订阅者和每天的数据。如果订阅者不再存在，那么他们将没有更多记录，即 SUB123 从 28/10/2021 开始不再存在，那么该订阅者将每天都有记录，直到 27/10/2021。目前的问题是一些订阅者错过了日期，这可能是因为它是周末或其他问题。我想用空值填充这些记录，以便它们可以记录在案。

当前问题：

Subscriber	Date	Rev
sub123	25/10/2021	256
sub456	25/10/2021	282
sub123	26/10/2021	652
sub123	27/10/2021	396
sub456	28/10/2021	132
sub456	29/10/2021	484
sub456	01/11/2021	96
sub456	02/11/2021	45

所需的解决方案：

Subscriber	Date	Rev
sub123	25/10/2021	256
sub456	25/10/2021	282
sub123	26/10/2021	652
sub456	26/10/2021	NULL
sub123	27/10/2021	396
sub456	27/10/2021	NULL
sub456	28/10/2021	132
sub456	29/10/2021	484
sub456	30/10/2021	NULL
sub456	31/10/2021	NULL
sub456	01/11/2021	96
sub456	02/11/2021	45

我目前的尝试：

WITH all_dates as (
SELECT
     CAST(date_column AS DATE) date_column, b.subscriber, b.date
FROM
 (VALUES
     (SEQUENCE(
      min(b.date) OVER (PARTITION BY b.subscriber ORDER BY b.date),
      max(b.date) OVER (PARTITION BY b.subscriber ORDER BY b.date),
      INTERVAL '1' DAY)
     )
 ) AS t1(date_array)
CROSS JOIN
 UNNEST(date_array) AS t2(date_column) 
LEFT JOIN MAINTABLE b 
on t2.date_column = b.date
), 
customer_dates as (
SELECT distinct a.subscriber, a.date, b.date_column
from MAINTABLE a
left join all_dates b
on a.date = b.date_column
    )
    SELECT *
    from customer_dates a

此代码不起作用，但它是我尝试完成的尝试，如果我使用下面附带的代码，它将为所有订阅者生成从初始日期到结束日期的日期这不是我们想要的，因此上面的代码就是我们所尝试的。

WITH all_dates as (
SELECT
     CAST(date_column AS DATE) date_column, b.subscriber, b.date
FROM
 (VALUES
     (SEQUENCE(
      date('2021-10-25'),
      date('2022-04-30'),
      INTERVAL '1' DAY)
     )
 ) AS t1(date_array)
CROSS JOIN
 UNNEST(date_array) AS t2(date_column) 
LEFT JOIN MAINTABLE b 
on t2.date_column = b.date
), 
customer_dates as (
SELECT distinct a.subscriber, a.date, b.date_column
from MAINTABLE a
left join all_dates b
on a.date = b.date_column
    )
    SELECT *
    from customer_dates a

Answer 1

您可以使用 lag 函数生成缺失范围以使用 unnest 展平并另外处理 Rev：

-- sample data
WITH dataset (Subscriber, Date, Rev) AS (
    VALUES ('sub123',   date_parse('25-10-2021', '%d-%m-%Y'),   256),
    ('sub456',  date_parse('25-10-2021', '%d-%m-%Y'),   282),
    ('sub123',  date_parse('26-10-2021', '%d-%m-%Y'),   652),
    ('sub123',  date_parse('27-10-2021', '%d-%m-%Y'),   396),
    ('sub456',  date_parse('28-10-2021', '%d-%m-%Y'),   132),
    ('sub456',  date_parse('29-10-2021', '%d-%m-%Y'),   484),
    ('sub456',  date_parse('01-11-2021', '%d-%m-%Y'),   96),
    ('sub456',  date_parse('02-11-2021', '%d-%m-%Y'),   45)
) 

-- query
select subscriber, lifted_date as date, if(date = lifted_date, rev, NULL) rev
from
(
    select Subscriber, 
        Rev,
        cast(date as date) date, 
        lag(cast(date as date)) over(partition by Subscriber order by date) prev_date
    from dataset
)
cross join unnest(
        array_except(sequence(coalesce(prev_date, date), date, interval '1' day), array[prev_date])
    ) as t(lifted_date)
order by subscriber, date

输出：

subscriber	date	rev
sub123	2021-10-25 00:00:00.000	256
sub123	2021-10-26 00:00:00.000	652
sub123	2021-10-27 00:00:00.000	396
sub456	2021-10-25 00:00:00.000	282
sub456	2021-10-26 00:00:00.000
sub456	2021-10-27 00:00:00.000
sub456	2021-10-28 00:00:00.000	132
sub456	2021-10-29 00:00:00.000	484
sub456	2021-10-30 00:00:00.000
sub456	2021-10-31 00:00:00.000
sub456	2021-11-01 00:00:00.000	96
sub456	2021-11-02 00:00:00.000	45

当此标识符不再存在于数据库中时，特定标识符缺少日期而不添加额外日期 SQL

Missing dates for specific identifiers without adding extra dates when this identifier is no longer in the database SQL

sql

presto