Bigquery，查找组的滚动最新值并找到组的最小值

Question

假设我们在bigquery中有产品、卖家、价格和库存变化数据，如下所示。该数据来自产品列表表的更改（CDC）。

一种产品可能有多个卖家，并且不同的卖家对该产品有不同的价格。有时卖家会缺货，我们会在最低计算中忽略该卖家。因此当时该产品的最低价格取决于卖家的库存。

WITH a AS
(
  SELECT 1 AS dt, 'p' as product_id, 'A' AS seller, 10 as stock, 100 as price
  UNION ALL
  SELECT 2 AS dt, 'p' as product_id, 'B' AS seller, 10 as stock, 120 as price
  UNION ALL
  SELECT 3 AS dt, 'p' as product_id, 'C' AS seller, 10 as stock, 150 as price
  UNION ALL
  SELECT 4 AS dt, 'p' as product_id, 'D' AS seller, 10 as stock, 300 as price
  UNION ALL
  SELECT 5 AS dt, 'p' as product_id, 'E' AS seller, 10 as stock, 400 as price
  UNION ALL
  SELECT 6 AS dt, 'p' as product_id, 'F' AS seller, 10 as stock, 500 as price
  UNION ALL
  SELECT 7 AS dt, 'p' as product_id, 'G' AS seller, 10 as stock, 600 as price
  UNION ALL
  SELECT 8 AS dt, 'p' as product_id, 'A' AS seller, 0 as stock, 100 as price
  UNION ALL
  SELECT 9 AS dt, 'p' as product_id, 'B' AS seller, 10 as stock, 110 as price
  UNION ALL
  SELECT 10 AS dt, 'p' as product_id, 'B' AS seller, 10 as stock, 190 as price
  UNION ALL
  SELECT 11 AS dt, 'p' as product_id, 'G' AS seller, 10 as stock, 800 as price
  UNION ALL
  SELECT 12 AS dt, 'p' as product_id, 'G' AS seller, 10 as stock, 100 as price
)

SELECT *
FROM a

我想每次计算该产品的最低价格和卖家的最低价格：

所需输出：

dt	产品_id	最低价格	卖家最低价格
1	p	100	A
2	p	100	A
3	p	100	A
4	p	100	A
5	p	100	A
6	p	100	A
7	p	100	A
8	p	120	B
9	p	110	B
10	p	150	C
11	p	150	C
12	p	100	G

在时间 1，只有一个卖家销售该产品。所以最低价格是 100，最低价格的卖家是 A。在时间 2，出现了第二个卖家，但卖家 A 的最低价格仍然是 100。 直到时间7，状态都是一样的。

在时间8，卖家A缺货，所以该产品对卖家B的最低价格是120。在时间9，卖家B降价，所以对卖家B的最低价格是110。在时间10，B去缺货。在时间 11，G 提高了价格，因此没有效果。在时间 12，卖家 G 降低了价格。所以时间 12 的最低价格（对于有库存的卖家）是 100，卖家是 G。

简而言之，我想为有库存的卖家找到一个产品在不同时间的最低价格。

为了计算这个逻辑，我找到了一个包含交叉连接的解决方案，但它需要太长的时间和太多的资源。我想知道是否有更好的解决方案。我搜索了 stackoverflow + google，但找不到好的解决方案。

Answer 1

这样的事情怎么样？ （做得很快，所以可能也不是最有效的。）

创建所有不同时期和产品的列表（“distinct_periods”），然后加入拥有这些时期信息的卖家（“all_sellers”），并使用聚合仅筛选出每个卖家的最新信息。

然后最后一次聚合以找到您的分钟数（留下这些是 array_aggs，因为您可能需要整个历史记录 - 只需删除限制）。

WITH a AS
(
  SELECT 1 AS dt, 'p' as product_id, 'A' AS seller, 10 as stock, 100 as price
  UNION ALL
  SELECT 2 AS dt, 'p' as product_id, 'B' AS seller, 10 as stock, 120 as price
  UNION ALL
  SELECT 3 AS dt, 'p' as product_id, 'C' AS seller, 10 as stock, 150 as price
  UNION ALL
  SELECT 4 AS dt, 'p' as product_id, 'D' AS seller, 10 as stock, 300 as price
  UNION ALL
  SELECT 5 AS dt, 'p' as product_id, 'E' AS seller, 10 as stock, 400 as price
  UNION ALL
  SELECT 6 AS dt, 'p' as product_id, 'F' AS seller, 10 as stock, 500 as price
  UNION ALL
  SELECT 7 AS dt, 'p' as product_id, 'G' AS seller, 10 as stock, 600 as price
  UNION ALL
  SELECT 8 AS dt, 'p' as product_id, 'A' AS seller, 0 as stock, 100 as price
  UNION ALL
  SELECT 9 AS dt, 'p' as product_id, 'B' AS seller, 10 as stock, 110 as price
  UNION ALL
  SELECT 10 AS dt, 'p' as product_id, 'B' AS seller, 10 as stock, 190 as price
  UNION ALL
  SELECT 11 AS dt, 'p' as product_id, 'G' AS seller, 10 as stock, 800 as price
  UNION ALL
  SELECT 12 AS dt, 'p' as product_id, 'G' AS seller, 10 as stock, 100 as price
), distinct_periods as (
  select distinct dt, product_id
  from a
), all_sellers as (
  select distinct distinct_periods.dt, distinct_periods.product_id, 
    any_value(case when stock = 0 then null else price end) over(partition by distinct_periods.dt, distinct_periods.product_id, seller order by distinct_periods.dt desc) as instock_price,
    any_value(seller) over(partition by a.dt, distinct_periods.product_id, seller order by a.dt desc) as seller 
  from distinct_periods
  left join a
  on distinct_periods.product_id = a.product_id
  and distinct_periods.dt >= a.dt
)
select dt, product_id, array_agg(instock_price order by instock_price asc limit 1) as minimum_price, array_agg(seller order by instock_price asc limit 1) as seller_with_minimum_price
from all_sellers
where instock_price is not null
group by 1, 2