sql-server-2008 - SQL 性能 - 按多个字段分组

Question

我想知道哪个更好和/或更快。在下面，customer_id 是主键。按主键和其他 5 个字段分组，或者仅按主键分组，然后进行连接以获取其他字段。或者这是一个“视情况而定”的问题？

select customer_id, customer_name, customer_address, 5 other description fields.., 
  sum(amount) as SalesAmount
from customer c
inner join orders o
 on o.customer_id = c.customer_id
group by customer_id, customer_name, customer_address, 5 other description fields..

对比

select customer_id, customer_name, customer_address, 5 other description fields..,
     SalesAmount
from 
(
   select customer_id,
   sum(amount) as SalesAmount
   from customer c
   inner join orders o
     on o.customer_id = c.customer_id
   group by customer_id
) t
inner join customer c
   on c.customer_id = t.customer_id

Answer 1

这是一个依赖问题。后一种变体将再执行一次连接，这可能会更慢。但是如果你从这个 join 得到的额外数据很大，它可以更快，因为分组操作不需要对所有数据进行 shuffle。

所以你看，两种变体都有比另一个更快的场景。你需要测量。

我的猜测是，对于真实数据和查询，第一个变体通常要快得多。