Mysql 当查询有GROUP BY时，如何获取总数的百分比？

假设我有一个非标准化的表，其中有电影演员的名字和他们看过的电影。例如

CREATE TABLE movies_actors (
  movies_actors_id INT,
  movie VARCHAR(255),
  actor VARCHAR(255),
  PRIMARY KEY (movies_actors_id)
);

我做了一个演员选择，从电影演员组中按演员数一，以了解演员已经拍了多少部电影。但我也想知道这位演员在电影中所占的比例

我想我可以做到：

SELECT
  actor,
  COUNT(1) AS total,
  COUNT(1) / (SELECT COUNT(1) FROM movies_actors) * 100 AS avg
FROM movies_actors
GROUP BY actor;

但这似乎。。。idk。。。恶心

---

有什么想法吗？

对于大型集合，联接可能比子查询执行得更好

SELECT ma.actor
     , COUNT(1) AS total
     , COUNT(1) / t.cnt * 100 AS `percentage`
  FROM movies_actors ma
 CROSS
  JOIN (SELECT COUNT(1) AS cnt FROM movies_actors) t
 GROUP
    BY ma.actor
     , t.cnt  

对于大型集合，并且当返回大量行时，联接操作的性能通常优于子查询。在您的例子中，它不是一个相关的子查询，所以MySQL不需要执行多次，所以它可能没有任何区别

---

给COUNT1的非粉丝们的提示。。。我们可以用COUNT*或IFNULLSUM1,0替换COUNT1的任何和所有实例，以获得相同的结果。

我不确定这是否更好，但您可以进行求和，然后在其他地方进行计算：

SELECT actor,
    COUNT(1) AS total,
    SUM(oneMoviePercentPts) AS percentage
FROM movies_actors
CROSS JOIN 
(
    SELECT 100 / CAST(COUNT(1) AS DECIMAL(15,4)) AS oneMoviePercentPts 
    FROM movies_actors
) t
GROUP BY actor

---

我希望MySQL优化器足够聪明，不会多次执行子查询，但联接语法使其更为明确。

只要您想从同一个表中获取经过处理的数据，就进行自交叉联接

SELECT
m.actor,
COUNT(m.actor) AS total,
(COUNT(m.actor) / t.total_movies) * 100 AS avg
FROM movies_actors m
cross (select count(*) as total_movies from movies_actors) t
GROUP BY m.actor;

---

不使用联接和多重查询：-

select actor,counter,  100 * counter / @total as percentage
from(
select actor, 
        case when actor is null
            then @total := count(*)
            else count(*)
        end as counter
    from movies_actors 
    group by actor
    with rollup
) mytable

这对我很有用：

SELECT tmpTotal.yearmonth, tmpTotal.rec_count, 
      (tmpTotal.rec_count / @myCumul) * 100 AS myPercentage
FROM
(
  SELECT tmpResult.*, @myCumul := @myCumul + tmpResult.rec_count AS myNewCumul
  FROM
  (
    SELECT date_format(d.created_at, '%Y/%m') as yearmonth, count(*) rec_count
    FROM cf4a_webapp.factTable d 
      join cf4a_webapp.dimTable c on (d.client_id = c.id)
    WHERE c.id = 25 
      AND d.created_at >= '2019-01-01 00:00:01' 
      AND d.created_at < '2020-01-01 00:00:01'
    GROUP BY yearmonth
  ) tmpResult
  JOIN (SELECT @myCumul := 0) tmpCumul
) tmpTotal;

---

需要注意的是，由于MySQL引擎的工作方式，count*的性能将与count1完全相同，而且在我看来，在语义上更为正确。SELECT list的表达式3不在GROUP BY子句中，并且包含未聚合的列“t.cnt”，它在功能上不依赖GROUP BY子句中的列；这与sql\u mode=only\u full\u组不兼容_by@brahimm：要使语句兼容，只需将t.cnt添加到GROUPBY子句。也就是说，由ma.actor，t.cnt组成的小组。内联视图派生的表t保证在成功运行时只返回一行，否则将抛出错误。