如何在多个 CTE 中引用日期数组？

Question

我有一个工作查询，它根据截至该月底的客户交易计算摘要指标（在计算中使用该月的最后一天）。

我现在正在尝试运行针对文件中的所有月份进行这些计算。

我有两个引用报告月份的 CTE，但我无法弄清楚如何以及在何处将日期数组加入查询。

如果有任何建议，我将不胜感激！

下面的工作查询将月末日期硬编码到查询中。

With input_data AS (

SELECT '#1238' as order_id, DATE('2021-12-15') as order_date, 'c12345' as cust_id, 18 as order_value
UNION ALL SELECT '#1201' as order_id, DATE('2021-10-10') as order_date, 'c12345' as cust_id, 18 as order_value
UNION ALL SELECT '#1198' as order_id, DATE('2021-07-05') as order_date, 'c12345' as cust_id, 20 as order_value
UNION ALL SELECT '#1134' as order_id, DATE('2020-10-15') as order_date, 'c12345' as cust_id, 10 as order_value
UNION ALL SELECT '#1112' as order_id, DATE('2019-08-10') as order_date, 'c12345' as cust_id, 5 as order_value
UNION ALL SELECT '#1234' as order_id, DATE('2021-07-05') as order_date, 'c11111' as cust_id, 118 as order_value
UNION ALL SELECT '#1294' as order_id, DATE('2021-01-05') as order_date, 'c11111' as cust_id, 68 as order_value
UNION ALL SELECT '#1290' as order_id, DATE('2021-01-01') as order_date, 'c11111' as cust_id, 82 as order_value
UNION ALL SELECT '#1284' as order_id, DATE('2020-01-15') as order_date, 'c22222' as cust_id, 98 as order_value),

-- date array that I'm trying to use to group by in final step but not used yet
dates AS (
  SELECT
   LAST_DAY(array_date, month) AS mth_end_date
  FROM 
    UNNEST(GENERATE_DATE_ARRAY( 
      (SELECT MIN(order_date) FROM input_data), 
      (SELECT MAX(order_date) FROM input_data), INTERVAL 1 MONTH)) as array_date),

-- for each order record add the date of that customers prior order
-- exclude all orders in the future (relative to the report date)
enhanced_orders AS (
SELECT 
*
,lag(order_date) OVER (PARTITION by cust_id ORDER BY order_date) AS cust_prev_order_date
FROM input_data
WHERE order_date < Date('2021-12-31')),

-- for each customer determine the most recent, and the next most recent, order dates  
cust_calc as (
SELECT
cust_id
, max(order_date) as max_order_date
, max(cust_prev_order_date) as max_prev_order_date
FROM enhanced_orders
GROUP by 1),

-- for each customer calculate how many days ago their most recent order was from the report date, 
-- and calculate the days between their last two most recent orders
-- exclude customers if their most recent order was >12m ago, or they don't have an order prior to their most recent one
cust_calc2 AS (
SELECT
cust_id
, date_diff(Date('2021-12-31'),  max_order_date, day) as recency
, date_diff(max_order_date, max_prev_order_date, day) as latency
FROM cust_calc
WHERE max_prev_order_date IS NOT NULL 
AND max_order_date > date_sub(Date('2021-12-31'), INTERVAL 1 YEAR))

-- average the recency and latency of all customers at the report date
SELECT
--- add in format_date('%Y-%m', eom_date)
avg(recency) AS monthly_recency
,avg(latency) AS monthly_latency
FROM cust_calc2 
-- Group by 1

Answer 1

以下是基于 PostgreSQL14 的测试。但是建议的更改应该可以直接应用。

数据中仅月份的结果：

The PG 14 fiddle

计算所有 input_data 的先前订单日期，没有限制。我们稍后会添加该限制：

   , enhanced_orders AS (
       -- for each order record add the date of that customers prior order
       SELECT *
            , lag(order_date) OVER (PARTITION by cust_id ORDER BY order_date) AS cust_prev_order_date
         FROM input_data
     )

现在引入报表日期列表（隐含交叉连接），并根据每个对应的mth_end_date限制行，在GROUP BY项中添加mth_end_date：

   , cust_calc as (
       -- for each customer determine the most recent, and the next most recent, order dates
       -- exclude all orders in the future (relative to the report date)
       SELECT cust_id
            , mth_end_date
            , max(order_date)           as max_order_date
            , max(cust_prev_order_date) as max_prev_order_date
         FROM enhanced_orders, dates
        WHERE order_date < mth_end_date
        GROUP BY 1, 2
     )

根据需要在下一个术语中包含 mth_end_date。为 eom_date 添加别名。

注意：对 PG 进行了一些细微的更改。只关注 mth_end_date 调整：

   , cust_calc2 AS (
       -- for each customer calculate how many days ago their most recent order was from the report date, 
       -- and calculate the days between their last two most recent orders
       -- exclude customers if their most recent order was >12m ago, or they don't have an order prior to their most recent one
       SELECT cust_id
            , mth_end_date AS eom_date
            , age(mth_end_date,   max_order_date)      as recency
            , age(max_order_date, max_prev_order_date) as latency
         FROM cust_calc
        WHERE max_prev_order_date IS NOT NULL 
          AND max_order_date > (mth_end_date - INTERVAL '1' YEAR)
     )

最后，最后一个查询表达式按eom_date分组：

-- average the recency and latency of all customers at the report date
-- add in format_date('%Y-%m', eom_date)
SELECT eom_date
     , avg(recency) AS monthly_recency
     , avg(latency) AS monthly_latency
  FROM cust_calc2 
 GROUP BY 1
 ORDER BY eom_date
;

最终PGSQL:

CREATE OR REPLACE FUNCTION last_day(date)
RETURNS date AS
$$
  SELECT (date_trunc('MONTH', ) + INTERVAL '1 MONTH - 1 day')::date;
$$ LANGUAGE 'sql' IMMUTABLE STRICT;

With RECURSIVE input_data AS (
       SELECT '#1238' as order_id, DATE('2021-12-15') as order_date, 'c12345' as cust_id, 18 as order_value UNION ALL
       SELECT '#1201' as order_id, DATE('2021-10-10') as order_date, 'c12345' as cust_id, 18 as order_value UNION ALL
       SELECT '#1198' as order_id, DATE('2021-07-05') as order_date, 'c12345' as cust_id, 20 as order_value UNION ALL
       SELECT '#1134' as order_id, DATE('2020-10-15') as order_date, 'c12345' as cust_id, 10 as order_value UNION ALL
       SELECT '#1112' as order_id, DATE('2019-08-10') as order_date, 'c12345' as cust_id, 5 as order_value UNION ALL
       SELECT '#1234' as order_id, DATE('2021-07-05') as order_date, 'c11111' as cust_id, 118 as order_value UNION ALL
       SELECT '#1294' as order_id, DATE('2021-01-05') as order_date, 'c11111' as cust_id, 68 as order_value UNION ALL
       SELECT '#1290' as order_id, DATE('2021-01-01') as order_date, 'c11111' as cust_id, 82 as order_value UNION ALL
       SELECT '#1284' as order_id, DATE('2020-01-15') as order_date, 'c22222' as cust_id, 98 as order_value
     )
   , dates AS (
       -- date array to use to group by in final step
       SELECT DISTINCT LAST_DAY(order_date::DATE) AS mth_end_date
         FROM input_data
     )
   , enhanced_orders AS (
       -- for each order record add the date of that customers prior order
       SELECT *
            , lag(order_date) OVER (PARTITION by cust_id ORDER BY order_date) AS cust_prev_order_date
         FROM input_data
     )
   , cust_calc as (
       -- for each customer determine the most recent, and the next most recent, order dates
       -- exclude all orders in the future (relative to the report date)
       SELECT cust_id
            , mth_end_date
            , max(order_date)           as max_order_date
            , max(cust_prev_order_date) as max_prev_order_date
         FROM enhanced_orders, dates
        WHERE order_date < mth_end_date
        GROUP BY 1, 2
     )
   , cust_calc2 AS (
       -- for each customer calculate how many days ago their most recent order was from the report date, 
       -- and calculate the days between their last two most recent orders
       -- exclude customers if their most recent order was >12m ago, or they don't have an order prior to their most recent one
       SELECT cust_id
            , mth_end_date AS eom_date
            , age(mth_end_date,   max_order_date)      as recency
            , age(max_order_date, max_prev_order_date) as latency
         FROM cust_calc
        WHERE max_prev_order_date IS NOT NULL 
          AND max_order_date > (mth_end_date - INTERVAL '1' YEAR)
     )
-- average the recency and latency of all customers at the report date
-- add in format_date('%Y-%m', eom_date)
SELECT eom_date
     , avg(recency) AS monthly_recency
     , avg(latency) AS monthly_latency
  FROM cust_calc2 
 GROUP BY 1
 ORDER BY eom_date
;

如何在多个 CTE 中引用日期数组？

How do I reference a date array in multiple CTE's?

sql

google-bigquery