如何用前一行的值填充空列？

Question

我想在这里做点什么。我想让所有的列都充满价值。但是当我有空列时，我想让它填充上一个非空列的值。

with cte as (
select '2019-11-12 16:01:55' as timestamp, null as owner_id, null as owner_assigneddate, null as lastmodifieddate union all
select '2019-11-12 19:03:18' as timestamp, 39530934 as owner_id, '2019-11-12 19:03:18' as owner_assigneddate, '2019-11-12 19:03:18' as lastmodifieddate union all
select '2019-11-12 19:03:19' as timestamp, null as owner_id, null as owner_assigneddate, '2019-11-12 19:03:19' as lastmodifieddate union all
select '2019-11-12 19:03:20' as timestamp, null as owner_id, null as owner_assigneddate, '2019-11-12 19:03:20' as lastmodifieddate union all
select '2019-11-12 19:03:31' as timestamp, 40320368 as owner_id, '2019-11-12 19:03:31' as owner_assigneddate, '2019-11-12 19:03:31' as lastmodifieddate union all
select '2019-11-12 19:03:33' as timestamp, null as owner_id, null as owner_assigneddate, '2019-11-12 19:03:33' as lastmodifieddate union all
select '2019-11-12 19:03:56' as timestamp, null as owner_id, null as owner_assigneddate, '2019-11-12 19:03:356' as lastmodifieddate)

select timestamp,
       owner_id,
       owner_assigneddate,
       lastmodifieddate,
       COALESCE(owner_id, LEAD(owner_id) OVER(ORDER BY timestamp DESC)) AS test_column
from cte order by timestamp asc

通过前面的查询，我已经设法将值仅放在下一行中。

我想要做的是用基于上一行的值填充所有列。第 4 行的值应为 39530934，第 7 行的值应为 40320368。我想我在这里遗漏了一些东西，但我不知道是什么。

Answer 1

这应该适用于您的 cte 定义：

...
select timestamp,
       owner_id,
       owner_assigneddate,
       lastmodifieddate,
       LAST_VALUE(owner_id IGNORE NULLS) 
            OVER(ORDER BY timestamp ASC ROWS BETWEEN 
                 UNBOUNDED PRECEDING AND CURRENT ROW) AS test_column
from cte order by timestamp asc

Answer 2

就此而言，Big Query 不支持 window 函数中的 ignore null。这是一个依赖于 window max 的解决方案来定位保存最后一个非空 owner_id 的记录（这假定时间戳的唯一性）。有了这些信息，您就可以通过连接引入相应的 owner_id。

select 
    c.timestamp,
    coalesce(c.owner_id, c_lag.owner_id) owner_id,
    c.owner_assigneddate,
    c.lastmodifieddate
from 
    (
        select
            cte.*,
            max(case when owner_id is not null then timestamp end) 
                over(order by timestamp rows unbounded preceding) target_timestamp
        from cte
    ) c
    left join cte c_lag 
        on c.owner_id is null 
        and c_lag.timestamp = c.target_timestamp

Demo on DB Fiddle:

timestamp           | owner_id | owner_assigneddate  | lastmodifieddate    
:------------------ | -------: | :------------------ | :-------------------
2019-11-12 16:01:55 |     null | null                | null                
2019-11-12 19:03:18 | 39530934 | 2019-11-12 19:03:18 | 2019-11-12 19:03:18 
2019-11-12 19:03:19 | 39530934 | null                | 2019-11-12 19:03:19 
2019-11-12 19:03:20 | 39530934 | null                | 2019-11-12 19:03:20 
2019-11-12 19:03:31 | 40320368 | 2019-11-12 19:03:31 | 2019-11-12 19:03:31 
2019-11-12 19:03:33 | 40320368 | null                | 2019-11-12 19:03:33 
2019-11-12 19:03:56 | 40320368 | null                | 2019-11-12 19:03:356

注意：如果需要，为了更好地理解逻辑，您可以运行独立地查看内部查询 returns（请参阅数据库 fiddle）。

编辑

重读此文，我发现 window max 提供的信息在您的原始数据中已经可用，在列 owner_assigneddate 中...所以这要简单得多：

select 
    c.timestamp,
    coalesce(c.owner_id, c_lag.owner_id) owner_id,
    c.owner_assigneddate,
    c.lastmodifieddate
from
    cte c
    left join cte c_lag 
        on c.owner_id is null 
        and c_lag.timestamp = c.owner_assigneddate

Answer 3

在 BigQuery 中使用 LAST_VALUE() 和 IGNORE NULLS 选项和 COALESCE():

select timestamp,
       COALESCE(owner_id, last_value(owner_id ignore nulls) over (order by timestamp)) as owner_id,
       COALESCE(owner_assigneddate, LAST_VALUE(owner_assigneddate IGNORE NULLS) OVER (ORDER BY TIMESTAMP)) as owner_assigneddate,
       COALESCE(lastmodifieddate, LAST_VALUE(lastmodifieddate IGNORE NULLS) OVER (ORDER BY TIMESTAMP)) as lastmodifieddate
from cte order by timestamp asc

如何用前一行的值填充空列？

How do I fill null columns with values from the previous row?

sql

data-warehouse

left-join

gaps-and-islands

google-bigquery