为什么我不能在下一个 SELECT 表达式中使用列别名？

Question

我可以修改下一个以在表达式ROUND(avg_time * cnt, 2) 中使用列别名avg_time 和cnt 吗？

SELECT 
    COALESCE(ROUND(stddev_samp(time), 2), 0) as stddev_time, 
    MAX(time) as max_time, 
    ROUND(AVG(time), 2) as avg_time, 
    MIN(time) as min_time, 
    COUNT(path) as cnt, 
    ROUND(avg_time * cnt, 2) as slowdown, path
FROM 
    loadtime
GROUP BY
    path
ORDER BY
    avg_time DESC
LIMIT 10;

引发下一个错误：

ERROR:  column "avg_time" does not exist
LINE 7:  ROUND(avg_time * cnt, 2) as slowdown, path

不过，下一个可以正常工作（使用主表达式而不是列别名：

SELECT 
    COALESCE(ROUND(stddev_samp(time), 2), 0) as stddev_time, 
    MAX(time) as max_time, 
    ROUND(AVG(time), 2) as avg_time, 
    MIN(time) as min_time, 
    COUNT(path) as cnt, 
    ROUND(AVG(time) * COUNT(path), 2) as slowdown, path
FROM 
    loadtime
GROUP BY
    path
ORDER BY
    avg_time DESC
LIMIT 10;

Answer 1

您可以在GROUP BY 或HAVING 语句中使用先前创建的别名，但不能在SELECT 或WHERE 语句中使用。这是因为程序同时处理了所有的SELECT 语句，还不知道别名的值。

解决方案是将查询封装在子查询中，然后别名在外面可用。

SELECT stddev_time, max_time, avg_time, min_time, cnt, 
       ROUND(avg_time * cnt, 2) as slowdown
FROM (
        SELECT 
            COALESCE(ROUND(stddev_samp(time), 2), 0) as stddev_time, 
            MAX(time) as max_time, 
            ROUND(AVG(time), 2) as avg_time, 
            MIN(time) as min_time, 
            COUNT(path) as cnt, 
            path
        FROM 
            loadtime
        GROUP BY
            path
        ORDER BY
            avg_time DESC
        LIMIT 10
   ) X;

Answer 2

查询的执行顺序（以及表达式和别名的求值）与它的编写方式不同。 “一般”位置是子句按以下顺序评估：

FROM
WHERE
GROUP BY
HAVING
SELECT
ORDER BY

因此，列别名对于大多数查询来说是未知的直到选择子句完成（这就是您可以使用别名的原因在 ORDER BY 子句中）。但是，在 from 子句中建立的 表别名 在 where to order by 子句中可以理解。

最常见的解决方法是将查询封装到“派生表”中

推荐阅读：Order Of Execution of the SQL query

注意：不同的 SQL dbms 对于别名的使用有不同的具体规则

编辑 提醒读者注意逻辑子句序列的目的是通常（但不总是）别名仅在声明别名的子句之后才变得可引用。其中最常见的是SELECT 子句中声明的别名可以被ORDER BY 子句使用。特别是，在 SELECT 子句中声明的别名不能在同一 SELECT 子句中引用。

但请注意，由于产品的差异，并非每个 dbms 都会以这种方式运行

Answer 3

在实际创建虚拟关系之前，别名不可用，如果您想使用别名本身执行其他表达式，则必须使用 as 子查询创建虚拟关系，而不是在其之上运行附加查询。因此，我会将您的查询修改为以下内容：

SELECT stddev_time, max_time, avg_time, min_time, ROUND(avg_time * cnt, 2) as slowdown, path FROM
(
SELECT 
    COALESCE(ROUND(stddev_samp(time), 2), 0) as stddev_time, 
    MAX(time) as max_time, 
    ROUND(AVG(time), 2) as avg_time, 
    MIN(time) as min_time, 
    COUNT(path) as cnt, 
    ROUND(AVG(time) * COUNT(path), 2) as slowdown, path
FROM 
    loadtime
GROUP BY
    path
ORDER BY
    avg_time DESC
LIMIT 10;
)

我想在此处添加您的第二个查询有效的原因是因为查询规划器将这些列识别为直接在您查询它们的表中定义的。

Answer 4

要么重复表达式：

ROUND(ROUND(AVG(time), 2) * COUNT(path), 2) as slowdown

或使用子查询：

SELECT *, ROUND(avg_time * cnt, 2) as slowdown FROM (
  SELECT 
    COALESCE(ROUND(stddev_samp(time), 2), 0) as stddev_time, 
    MAX(time) as max_time, 
    ROUND(AVG(time), 2) as avg_time, 
    MIN(time) as min_time, 
    COUNT(path) as cnt, 
    path
  FROM loadtime
  GROUP BY path) x
ORDER BY avg_time DESC
LIMIT 10;