方法1 :通过行号来进行加减
1.新建表1添加自增行号列(考虑到自增id 有丢失数据现象)
2.复制上表1为表2
3.根据表1、表2行号进行where或者left join on的处理
* where 进行处理的条件为表1表2完全的匹配
*left join on 处理的条件为 左表完全的匹配
#设置内存变量
SET @arownum=0;
SET @brownum=0;
#查询sql
SELECT a.id ,a.terminal_id,a.TIMESTAMP,a.seq-b.seq FROM
( SELECT @arownum:=@arownum+1 AS id ,terminal_id,TIMESTAMP,seq FROM `msg_basic` ORDER BY terminal_id, TIMESTAMP )AS a
LEFT JOIN
( SELECT @brownum:=@brownum+1 AS id ,terminal_id,TIMESTAMP,seq FROM `msg_basic` ORDER BY terminal_id, TIMESTAMP )AS b
ON a.id = b.id+1
#也可新建表
CREATE TABLE tab2 AS SELECT @brownum:=@brownum+1 AS id ,terminal_id,TIMESTAMP,seq FROM `msg_basic` ORDER BY terminal_id, TIMESTAMP
方法1优化:
优化原因1:方法1中的a.id = b.id+1进行查询速度很慢(1min以上),但a.id = b.id进行查询的速度很快;于是开始想到update id (UPDATE tabname SET id =id-1 测试可行 ;)
优化原因2:消息序列在筛选时容易出现terminal_id 不一致现象,所以增加条件
优化原因3,查看计算结果的过程明细
优化后语句:
SET @arownum=0;
SET @brownum=1;
#查询sql
SELECT a.id ,b.id,a.terminal_id,a.TIMESTAMP,b.TIMESTAMP,a.seq,b.seq ,a.seq-b.seq AS dseq FROM
( SELECT @arownum:=@arownum+1 AS id ,terminal_id,TIMESTAMP,seq FROM `msg_basic` ORDER BY terminal_id, TIMESTAMP )AS a,
( SELECT @brownum:=@brownum+1 AS id ,terminal_id,TIMESTAMP,seq FROM `msg_basic` ORDER BY terminal_id, TIMESTAMP )AS b
WHERE a.id = b.id AND a.terminal_id=b.terminal_id #筛选设备相同、seq差距为1
AND a.seq-b.seq>1 #根据消息序号进行筛选,丢失消息数量需要大于1,如果连续则为1
AND LEFT(b.TIMESTAMP,6)=LEFT(a.TIMESTAMP,6)#筛选日期相同的,进行消息包减法
AND CAST(a.seq AS SIGNED INTEGER) > 12 #排除熄火后重新点火后seq变为1
AND CAST(a.timestamp AS SIGNED INTEGER) > 180201095138 #排除 18-02-01之前的数据
AND a.terminal_id <> '01020408101F' # 排除设备编号为01020408101F
方法2:通过limit不能实现原因
前提 t1=[2-6]5个数字,t2:=[1:6] 6个数字 ,t1为t2删去一行
因为mysql的相减 必须通过 where 条件找到共同的条件,类似于t1.id= t2.id;直接让其相减则能够得到, 5*6=30个结果;