Mysql SQL聚合分数

我有一张有City和ComplaintType的桌子

我正在尝试创建具有以下计算的规范化列：

（伪）选择特定城市中特定类型的编号）/（特定城市中所有投诉的编号

我目前拥有以下SQL：

SELECT City AS city_name, ComplaintType AS complaint_type, 
    count(*) / (SELECT count(City) FROM data GROUP BY City) AS complaint_frac,
    count(*) AS count_freq,
    (SELECT count(City) FROM data GROUP BY City) AS count_city
FROM data
GROUP BY City, ComplaintType
ORDER BY complaint_frac DESC

这给了我下表：

城市中的投诉总数（

count\u city

）不正确。但是，当我自己运行count\u city查询时，计数是正确的，并给出以下输出：

如何正确获取与城市投诉数量相关的

city\u计数

，以便计算正确的分数

冷硬数字示例：

布朗克斯和热水=79690

布朗克斯（投诉总数）=579363

投诉分数=79690/579363=0.13754761695

---

关联主表中的

子查询

SELECT City AS city_name, ComplaintType AS complaint_type, 
    count(*) / (SELECT count(City) FROM data GROUP BY City) AS complaint_frac,
    count(*) AS count_freq,
    (SELECT count(d1.City) FROM data d1 WHERE d1.City = d2.City GROUP BY d1.City) AS count_city
FROM data d2
GROUP BY City, ComplaintType
ORDER BY complaint_frac DESC

---

关联主表中的

子查询

SELECT City AS city_name, ComplaintType AS complaint_type, 
    count(*) / (SELECT count(City) FROM data GROUP BY City) AS complaint_frac,
    count(*) AS count_freq,
    (SELECT count(d1.City) FROM data d1 WHERE d1.City = d2.City GROUP BY d1.City) AS count_city
FROM data d2
GROUP BY City, ComplaintType
ORDER BY complaint_frac DESC

---

这不需要子查询，至少在MySQL 8+中是这样；窗口函数可以完成以下工作：

SELECT City AS city_name, ComplaintType AS complaint_type,
       count(*) / sum(count(*)) over (partition by city) as complaint_frac,
       count(*) as count_freq,
       sum(count(*)) over (partition by city) as count_city 
FROM data
GROUP BY City, ComplaintType
ORDER BY complaint_frac DESC

---

这不需要子查询，至少在MySQL 8+中是这样；窗口函数可以完成以下工作：

SELECT City AS city_name, ComplaintType AS complaint_type,
       count(*) / sum(count(*)) over (partition by city) as complaint_frac,
       count(*) as count_freq,
       sum(count(*)) over (partition by city) as count_city 
FROM data
GROUP BY City, ComplaintType
ORDER BY complaint_frac DESC