我已经在一个表上运行了以下测试，该表具有100万个与1万个类别随机相关的产品（MariaDB 10.0.19）：

I've run the following test on a table with 1M products randomly related to 10K categories (MariaDB 10.0.19):

select p.categoryId, count(*) as total
from products p
group by p.categoryId
having count(*) = 100

执行时间：156毫秒

select p.categoryId, count(*) as total
from products p
group by p.categoryId
having total = 100

执行时间：156毫秒

因此，性能似乎没有任何差异。

So there doesn't seem to be any difference in performance.

请注意，使用 ORDER BY 引擎会将结果复制到临时表中（请参见说明：使用临时表；使用文件排序）。因此，即使您使用 ORDER BY COUNT（*），也无法重新计算该值。

Note that with ORDER BY the engine will copy the result into a temporary table (See EXPLAIN: Using temporary; Using filesort). So the value can't be recalculated, even when you use ORDER BY COUNT(*).

但是-当我使用 ORDER BY COUNT（DISTINGT ...）时（我无法解释）：

However - There is a difference (which I can not explain) when I use ORDER BY COUNT(DISTINGT ...):

select p.categoryId, count(distinct p.productData) as total
from products p
group by p.categoryId
order by total

配置文件：863毫秒，用于复制到tmp表

Profile: 863 msec for Copying to tmp table

select p.categoryId, count(distinct p.productData) as total
from products p
group by p.categoryId
order by count(distinct p.productData)

配置文件：963毫秒复制到tmp表

Profile: 963 msec for Copying to tmp table