如何在 SQL 中的 window 中获取组中的第一个元素？

Question

我想获取组的第一个元素，但是必须为每个window计算组。我想做这样的事情：

架构

TABLE
id
target
groups
capture_date
event_date

SELECT
    AVG(
        FIRST(target) GROUP BY id ORDER BY capture_date DESC WHERE capture_date <= MAX(event_date)
    ) OVER (
        PARTITION BY groups
        ORDER BY event_date
        RANGE BETWEEN INTERVAL 7 DAYS PRECEDING AND CURRENT ROW
    )
FROM table

我想在 sql 或 pyspark 中执行此操作，无论更简单。有任何想法吗？谢谢！

Answer 1

这是一个完整的 SQL 版本，但如果需要，可以在 spark/pyspark 中 re-written。我使用了 groupby，但你也可以运行秒 window with row_number & where

with raw_averages as ( -- short cut for subquery
 SELECT
    AVG(
        target
    ) OVER (
        PARTITION BY groups
        ORDER BY event_date
        RANGE BETWEEN INTERVAL 7 DAYS PRECEDING AND CURRENT ROW
    ) as average, 
    ID, 
    capture_date,
    event_date
FROM table ),
grouped_result as -- more shortcut for subquery
(SELECT 
 id, 
 avg(average) as average, -- the average of the entire group is the same -->math trick
 reverse( -- sort descending
  array_sort( --sort ascending by first item ( event_date )
   arrays_zip( -- create one array of below arrays
    collect_list( event_date ), -- collect the grouped items *has to be first to get the ordering you want*
    collect_list( capture_date ) -- collect the grouped items
   )
  )
 )[0] as values --getting first will return max (reference first item in array)
from raw_averages 
GROUP BY groups, id 
-- HAVING values.`0` = values.`1` -- having might work here but I didn't explore it
)
select 
 groups, 
 id, 
 average,
 values.`0` as event_date -- awkward syntax because of arrays_zip
 values.`1` as capture_date
from 
 grouped_result
where values.`0` = values.`1`

如何在 SQL 中的 window 中获取组中的第一个元素？

How to get the first element in groups over a window in SQL?

sql

pyspark

apache-spark-sql