今天在检查报表接口的时候发现了一个非常大的bug，在同事的帮助下 解决了，为了长记性，我就在这里对这个问题做一个详细的描述剖析，以便以后可以快速解决同一类的问题。
接口需求:
如下图所示，在选取时间范围之后，我需要点击分别按照 天，月，年 来查看销售额(总销售额，退货金额，减免金额，实际收入)

我的理解误区:
我刚开始写这个接口的时候 想通过一些sql来获取数据库所有的数据(完成的订单和退货的订单) 然后在java程序里面分别过来退货订单，然后通过计算 总销售额-减免-退货 = 实际营业额收入，后来发现 按天 月 年 来查询的时候，
退货的金额过滤不出来，这就导致了问题：

SELECT DATE_FORMAT(o.create_tm,'%Y-%m') as `period`,
     sum(o.price) as salesTotal,sum(o.price) as returnPrice,sum(o.prixx_redux) as      reduction,o.mode
    FROM
    `ordex` o
     WHERE (o.mode = 2 or o.mode= 1) 
     and o.pay = 1
      and o.shop_id = 10000xx
        and o.create_tm between '2015-12-12' and '2016-02-19'
        group by MONTH(o.create_tm)
    ORDER BY o.create_tm DESC
    

后来经过讨论，决定以下sql来解决这个问题:

   SELECT r.period ,salesTotal,reduction ,ifnull (returnPrice,0) as returnPrice , (salesTotal- (reduction+ ifnull (returnPrice,0))) as allin   from 
   (SELECT DATE_FORMAT(o.create_tm,'%Y-%m') as `period`,
     sum(o.price) as salesTotal,sum(o.price_reduce) as reduction
     FROM
    `order` o
     WHERE (o.mode = 2 or o.mode=1) 
     and o.pay = 1
       and o.shop_id = 10000xx
       and o.create_tm between '2015-01-12' and date_add('2016-02-19', INTERVAL 1 day)  
            group by MONTH(o.create_tm)
     ORDER BY o.create_tm DESC ) r
LEFT JOIN 
  (SELECT DATE_FORMAT(o.create_tm,'%Y-%m') as `period`,
    ifnull( sum(o.price),0) as returnPrice
    FROM
    `order` o
     WHERE o.mode = 2
     and o.pay = 1
      and o.shop_id = 10000xx
           and o.create_tm between '2015-01-12' and date_add('2016-02-19', INTERVAL 1 day)  
            group by MONTH(o.create_tm)
     ORDER BY o.create_tm DESC 
) r1
on r.period=r1.period

通过这样的连接 可以把mode=2的数据单独查出来 加入计算 已达计算的目的