Question

我有一个表，其中包含客户的销售数据，如下所示；

|CustomerCode|SalesDate|Profit
|123455666   |2018-06  |120
|123455666   |2018-06  |100
|123455666   |2018-05  |10
|123455666   |2018-04  |60
|666452342   |2018-06  |900
|666452342   |2018-05  |1000
|666452342   |2018-05  |900
|666452342   |2018-06  |800

我想要一个表，显示公司三个月的利润。预期表；

|CustomerCode|P_This_Month|P_1_Month_Ago|P_2_Month_Ago
|123455666   |220         |10           |60
|666452342   |900         |1900         |800

做到这一点的最佳方法是什么？目前，我正在使用JOIN运算符。但是，即使它适用于第一次联接，也不适用于第二次联接（获得了意外的高值）。

我该怎么做？

SELECT 
  This_Month.*, 
  SUM(_1_Month_Ago.UCSALES) sales_1_month_ago,
  SUM(_2_Months_Ago.UCSALES) sales_2_months_ago
FROM 
  SalesTable This_Month
LEFT JOIN
  SalesTable _1_Month_Ago ON This_Month.CustomerCode = _1_Month_Ago AND DATE_SUB(This_Month.SalesDate, INTERVAL 1 MONTH) = _1_Month_Ago.SalesDate
LEFT JOIN
  SalesTable _2_Months_Ago ON This_Month.CustomerCode = _2_Months_Ago AND DATE_SUB(This_Month.SalesDate, INTERVAL 1 MONTH) = _2_Months_Ago.SalesDate
GROUP BY 
  This_Month.CustomerCode

Answer 1

您可以通过条件聚合来做到这一点。

select
CustomerCode,
sum(case when sales_date=current_date then profit else 0 end) P_This_Month,
sum(case when sales_date=date_sub(current_date, INTERVAL 1 month) then profit else 0 end) P_1_Month,
sum(case when sales_date=date_sub(current_date, INTERVAL 2 month) then profit else 0 end) P_2_Month,
from <tableName>
group by CustomerCode

Answer 2

您可以在从日期中提取月份的情况下使用案例

    with cte as
    (
     select EXTRACT(month FROM SalesDate) as monthofday,CustomerCode,sum(Profit) as total
     from tbale

    ) select CustomerCode,
 sum(case when monthofday=4 then total else 0 end) as Forththmonthtotal,
 sum(case when monthofday=5 then total else 0 end) as fifththmonthtotal,
 --you can apply for 12 month this logic 
 from cte 
group by CustomerCode

Answer 3

以下是用于BigQuery标准SQL

此解决方案避免进行自我联接，并且通用性足以覆盖多个月的数据

#standardSQL
WITH months AS (
  SELECT FORMAT_DATE('%Y-%m', month) month
  FROM (
    SELECT 
      MIN(PARSE_DATE('%Y-%m', SalesDate)) min_month, 
      MAX(PARSE_DATE('%Y-%m', SalesDate)) max_month
    FROM `project.dataset.table`
  ), UNNEST(GENERATE_DATE_ARRAY(min_month, max_month, INTERVAL 1 MONTH)) month  
)
SELECT CustomerCode, month, 
  SUM(IF(diff = 0, P_This_Month, 0)) P_This_Month,
  SUM(IF(diff = 1, P_This_Month, 0)) P_1_Month_Ago,
  SUM(IF(diff = 2, P_This_Month, 0)) P_2_Month_Ago
FROM (
  SELECT CustomerCode, P_This_Month, PARSE_DATE('%Y-%m', month) month, 
    DATE_DIFF(PARSE_DATE('%Y-%m', month), PARSE_DATE('%Y-%m', SalesDate), MONTH) diff
  FROM months m CROSS JOIN (
    SELECT CustomerCode, SalesDate, SUM(Profit) P_This_Month
    FROM `project.dataset.table`
    GROUP BY CustomerCode, SalesDate
  ) t
)
WHERE diff BETWEEN 0 AND 2
GROUP BY CustomerCode, month

您可以使用问题中的示例数据来测试，玩游戏，如下例所示

#standardSQL
WITH `project.dataset.table` AS (
  SELECT 123455666 CustomerCode, '2018-06' SalesDate, 120 Profit UNION ALL
  SELECT 123455666, '2018-06', 100 UNION ALL
  SELECT 123455666, '2018-05', 10 UNION ALL
  SELECT 123455666, '2018-04', 60 UNION ALL
  SELECT 666452342, '2018-06', 900 UNION ALL
  SELECT 666452342, '2018-05', 1000 UNION ALL
  SELECT 666452342, '2018-05', 900 UNION ALL
  SELECT 666452342, '2018-04', 800 
), months AS (
  SELECT FORMAT_DATE('%Y-%m', month) month
  FROM (
    SELECT 
      MIN(PARSE_DATE('%Y-%m', SalesDate)) min_month, 
      MAX(PARSE_DATE('%Y-%m', SalesDate)) max_month
    FROM `project.dataset.table`
  ), UNNEST(GENERATE_DATE_ARRAY(min_month, max_month, INTERVAL 1 MONTH)) month  
)
SELECT CustomerCode, month, 
  SUM(IF(diff = 0, P_This_Month, 0)) P_This_Month,
  SUM(IF(diff = 1, P_This_Month, 0)) P_1_Month_Ago,
  SUM(IF(diff = 2, P_This_Month, 0)) P_2_Month_Ago
FROM (
  SELECT CustomerCode, P_This_Month, PARSE_DATE('%Y-%m', month) month, 
    DATE_DIFF(PARSE_DATE('%Y-%m', month), PARSE_DATE('%Y-%m', SalesDate), MONTH) diff
  FROM months m CROSS JOIN (
    SELECT CustomerCode, SalesDate, SUM(Profit) P_This_Month
    FROM `project.dataset.table`
    GROUP BY CustomerCode, SalesDate
  ) t
)
WHERE diff BETWEEN 0 AND 2
GROUP BY CustomerCode, month
-- ORDER BY CustomerCode, month

有结果

Row CustomerCode    month       P_This_Month    P_1_Month_Ago   P_2_Month_Ago    
1   123455666       2018-04-01  60              0               0    
2   123455666       2018-05-01  10              60              0    
3   123455666       2018-06-01  220             10              60   
4   666452342       2018-04-01  800             0               0    
5   666452342       2018-05-01  1900            800             0    
6   666452342       2018-06-01  900             1900            800

同一表上的迭代查询

3 个答案: