mysql - 将 Count() 小于总数百分比的行分组在一起

Question

我正在为一家课后教育公司开发一个数据库，该数据库跟踪学生记录，包括类(class)注册情况和学生信息。

我想做的是编写一个查询，它可以返回每所学校的注册学生人数，但也可以将贡献低于总数一定百分比的学校分组在一起(我想在中显示信息一张图表，但我们有很多学校只有 1 名学生来自那所学校，我不希望图表有 50 个条形图或饼图等)

所以代替

+-------------+------------+
| School Name | # Students |
+-------------+------------+
| School A    |         52 |
| School B    |         27 |
| School C    |         15 |
| School D    |          2 |
| School E    |          1 |
| School F    |          1 |
+-------------+------------+

我愿意

+---------------+------------+
|  School Name  | # Students |
+---------------+------------+
| School A      |         52 |
| School B      |         27 |
| School C      |         15 |
| Other Schools |          4 |
+---------------+------------+

这是我现在拥有的查询的简化形式，它可以工作，但在使用多个 Select 查询相同信息时有点多余。有没有办法减少冗余？

SELECT @enrollmentSum := COUNT(StudentEnrollmentID) FROM StudentEnrollment;
SELECT SchoolName, COUNT(StudentEnrollmentID) ECount FROM Student
JOIN StudentEnrollment ON StudentEnrollment.StudentID = Student.StudentID
JOIN School ON Student.SchoolID = School.SchoolID
GROUP BY SchoolName
HAVING Ecount >= .025 * @enrollmentSum
UNION ALL
SELECT "Other Schools" as SchoolName, SUM(Ecount) as ECount FROM
(
    SELECT SchoolName, COUNT(StudentEnrollmentID) ECount FROM Student
    JOIN StudentEnrollment ON StudentEnrollment.StudentID = Student.StudentID
    JOIN School ON Student.SchoolID = School.SchoolID
    GROUP BY SchoolName
    HAVING Ecount < .025 * @enrollmentSum
) t2
ORDER BY Ecount DESC

如果需要，相关表的基本结构:

学生

+-----------+-------------+----------+
| StudentID | StudentName | SchoolID |
+-----------+-------------+----------+

学校

+----------+------------+
| SchoolID | SchoolName |
+----------+------------+

学生注册

+---------------------+-----------+---------+
| StudentEnrollmentID | StudentID | ClassID |
+---------------------+-----------+---------+

谢谢你的帮助

Answer 1

提示:

• count(x) 返回“x 不为空”的行数，因此 count(primary key) = count(*) 更易于阅读

• “JOIN School ON Student.SchoolID = School.SchoolID”可以重写为“JOIN School USING (SchoolID)”，它更具可读性，并且如果您在结果集中只提供一列“SchoolID”使用类似“select *”的东西

现在，查询...

SELECT SchoolName, sum(cnt) ECount FROM 
(SELECT IF(count(*)>=.025*@enrollmentSum, SchoolName, 'Others') AS SchoolName,
 COUNT(*) cnt FROM Student
 JOIN StudentEnrollment USING (StudentID)
 JOIN School USING (SchoolID)
 GROUP BY SchoolName) subq
GROUP BY SchoolName
ORDER BY Ecount DESC

对于低于阈值的所有学校，使用 IF() 会将学校名称替换为“其他”。请注意，这是在 GROUP BY 之后计算的，因此您实际上可以在选定的表达式中使用 count(*)。然后另一个 GROUP BY 将“其他”组合在一起。

编辑

这是一个很 hack，但它似乎做你想做的......

SET @total=0; 
SELECT IF(cnt/@total>=0.2, SchoolName, 'Others') SN, sum(cnt) FROM (
    SELECT SchoolName, cnt, @total:=@total+cnt FROM (
        SELECT SchoolName, count(*) cnt FROM st GROUP BY SchoolName
    ) AS foo -- ORDER BY cnt DESC
) AS bar
GROUP BY SN ORDER BY sum(cnt) DESC;

这是变态。 MySQL 似乎总是首先具体化子查询“foo”并将结果存储在缓冲区中，然后再处理子查询“bar”。我认为“ORDER BY cnt DESC”是必要的，但如果它被注释掉，它似乎也有效。

运行子查询“foo”的副作用是将@total 设置为我们想要的值!

因此当运行外部子查询时，总数是可用的。

这种方法的问题是它可能会在没有警告的情况下停止工作，因为它是一种 hack。

Postgresql version if you're curiosus...