Question

我有这个查询，它基本上得到了客户在过去一年和3个月的平均支出：

SELECT SQL_CALC_FOUND_ROWS 
       customer_id,
       customer_name,
       AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 1 YEAR, "%Y-%m-01"),
            spend_amount,
            NULL
       )) AS 1_year_average_spend,
       AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 3 MONTH, "%Y-%m-01"),
            spend_amount,
            NULL
       )) AS 3_month_average_spend

FROM   customer_spends
GROUP  BY customer__id

但我还需要获得支出平均值的百分比差异：

E.g。（伪代码）

if (1_year_average_spend = 0)
    change = N/A
else 
    change = 3_month_average_spend / 1_year_average_spend - 1

如何，或者你建议我做什么来实现这个？

我能想到的唯一方法是可怕的：

IF(
    AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 1 YEAR, "%Y-%m-01"),
            `spend_amount`,
            NULL
    )) > 0,
    AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 3 MONTH, "%Y-%m-01"),
            spend_amount,
            NULL
    )) / AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 1 YEAR, "%Y-%m-01"),
            `spend_amount`,
            NULL
    )) - 1,
    "N/A"
) AS 3_month_performance

Answer 1

使用内部选择（它就像一个临时视图）并从中进行选择。这应该有效：

SELECT
  customer_id,
  customer_name,
  1_year_average_spend,
  3_month_average_spend,
  if (1_year_average_spend = 0, "N/A", (3_month_average_spend / 1_year_average_spend) - 1) AS 3_month_performance
FROM (SELECT
  customer_id,
  customer_name,
  AVG(IF(DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
    DATE_FORMAT(NOW() - INTERVAL 1 YEAR, "%Y-%m-01"), spend_amount, NULL)) AS 1_year_average_spend,
  AVG(IF(DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
    DATE_FORMAT(NOW() - INTERVAL 3 MONTH, "%Y-%m-01"), spend_amount, NULL)) AS 3_month_average_spend)
FROM customer_spends 
GROUP BY customer_id, customer_name ) x

Answer 2

NULL会N/A值吗？如果是这样，您可以将NULLIF()应用于这样的分母：

a / NULLIF(b, 0) - 1

如果第一个参数等于第二个参数，则

NULLIF返回NULL。如果操作数为NULL，则整个表达式的计算结果为NULL。

与@Bohemian一样，我也建议使用subselect。这是完整的查询：

SELECT SQL_CALC_FOUND_ROWS
  customer_id,
  customer_name,
  1_year_average_spend,
  3_month_average_spend,
  3_month_average_spend / NULLIF(1_year_average_spend, 0) - 1 AS change
FROM (
  SELECT
    customer_id,
    customer_name,
    AVG(IF(
         DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
             DATE_FORMAT(NOW() - INTERVAL 1 YEAR, "%Y-%m-01"),
         spend_amount,
         NULL
    )) AS 1_year_average_spend,
    AVG(IF(
         DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
             DATE_FORMAT(NOW() - INTERVAL 3 MONTH, "%Y-%m-01"),
         spend_amount,
         NULL
    )) AS 3_month_average_spend
  FROM customer_spends
  GROUP BY customer__id
) s

Answer 3

如果您乐意使用MySQL特定代码，可以使用User-Defined Variables这样的（简化版）：

SELECT @avg1 := ROUND((1 + 2 + 3) / 3, 2) AS avg1,
       @avg2 := ROUND((4 + 5 + 6) / 3, 2) AS avg2,
   IF( @avg1, ROUND(@avg2 / @avg1 - 1, 2), NULL ) AS result;

+------+------+--------+
| avg1 | avg2 | result |
+------+------+--------+
| 2.00 | 5.00 |   1.50 |
+------+------+--------+

那将成为：

SELECT SQL_CALC_FOUND_ROWS 
       customer_id,
       customer_name,
       @1_year_average_spend := AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 1 YEAR, "%Y-%m-01"),
            spend_amount,
            NULL
       )) AS 1_year_average_spend,
       @3_month_average_spend := AVG(IF(
            DATE(CONCAT(year_of_spend, "-", month_of_spend, "-01")) >=
                DATE_FORMAT(NOW() - INTERVAL 3 MONTH, "%Y-%m-01"),
            spend_amount,
            NULL
       )) AS 3_month_average_spend,
       IF( @1_year_average_spend,
           @3_month_average_spend / @1_year_average_spend - 1,
           NULL
       ) AS diff

FROM   customer_spends
GROUP  BY customer__id

注1：我已使用diff作为差异的列名，因为change是保留字，因此最终可能会导致问题。

注2：您需要了解文档中的以下警告，因为它们可能会影响您的结果：

十进制值和实数值的分配不会保留精度或者价值的规模。

和

作为一般规则，您不应该为用户变量赋值并在同一语句中读取值。你可能会得到你期望的结果，但这不能保证。的顺序涉及用户变量的表达式的评估是未定义的可能会根据给定语句中包含的元素进行更改。在SELECT @ a，@ a：= @ a + 1，...中，您可能会认为MySQL会进行评估首先是@a，然后是第二个任务。但是，改变了声明（例如，通过添加GROUP BY，HAVING或ORDER BY 子句）可能导致MySQL选择具有不同的执行计划评估顺序。

因此请谨慎使用，并进行适当的测试！

Answer 4

摆脱IF()，DATE()和CONCAT()函数调用。您的查询现在必须扫描整个customer_spends表并检查所有行的复杂条件，即使它们是10年的数据。

这也会使用(year_of_spend, month_of_spend)或(customer_id, year_of_spend, month_of_spend)的索引来加速查询：

SELECT c.customer_id
     , c.customer_name
     , 1_year_average_spend
     , 3_month_average_spend
     , CASE WHEN 1_year_average_spend = 0
              THEN 'N/A'
              ELSE (3_month_average_spend / 1_year_average_spend) - 1
       END AS percent_difference
FROM
    customer AS c
  JOIN
    ( SELECT customer_id
           , AVG(spend_amount) AS 1_year_average_spend 
      FROM customer_spends 
      WHERE (year_of_spend, month_of_spend) >=
               ( YEAR(CUR_DATE() - INTERVAL 1 YEAR)
               , MONTH(CUR_DATE() - INTERVAL 1 YEAR)
               )
      GROUP BY customer_id
    ) AS grp1year
    ON grp1year.customer_id = c.customer_id
  LEFT JOIN
    ( SELECT customer_id
           , AVG(spend_amount) AS 3_month_average_spend
      FROM customer_spends 
      WHERE (year_of_spend, month_of_spend) >=
               ( YEAR(CUR_DATE() - INTERVAL 3 MONTH)
               , MONTH(CUR_DATE() - INTERVAL 3 MONTH)
               )
      GROUP BY customer_id
    ) AS grp3month
    ON grp3month.customer_id = c.customer_id

计算两个计算字段的差异

4 个答案: