1

我有一个例行公事。但是太慢了。如何改进查询？

我的记录：http ://www.sqlfiddle.com/#!9/14cceb/1/0

我的查询：

CREATE DEFINER = 'root'@'localhost'
PROCEDURE example.ssa()
BEGIN
drop table if exists gps_table;
drop table if exists exam_datas;
CREATE TEMPORARY TABLE gps_table(ID int PRIMARY KEY AUTO_INCREMENT,timei 
int, 
trun_date_time datetime, tadd_meter int, tin_here int null);
insert into gps_table(timei,trun_date_time,tadd_meter,tin_here) select 
imei, run_date_time, add_meter, in_here from example_table;
CREATE TEMPORARY TABLE exam_datas(ID int PRIMARY KEY AUTO_INCREMENT,vimei 
int, vbas_run_date_time datetime, vbit_run_date_time datetime, vdifff int);

select tin_here from gps_table limit 1 into @onceki_durum;
select count(id) from gps_table into @kayit_sayisi;
set @i = 1;
set @min_mes = 0;
set @max_mes = 0;
set @frst_id = 0;
set @imei = 0;
set @run_date_time = '0000-00-00 00:00:00';
set @run_date_time2 = '0000-00-00 00:00:00';

   myloop: WHILE (@i <= @kayit_sayisi) DO 
  select tin_here from gps_table where id = @i into @in_here_true;
  if (@in_here_true = 1) then
    select id,trun_date_time, tadd_meter from gps_table where id = @i into @frst_id,@run_date_time2, @min_mes;    
    select id from gps_table where id > @frst_id and tin_here =0 order by id asc limit 1 INTO @id; 
    SET @id = @id-1;                                                                                 
    select id, timei, trun_date_time, tadd_meter from gps_table                                          
    where id = @id and tin_here =1 limit 1 into @i, @imei, @run_date_time, @max_mes;                    

      if(@i-@frst_id>3) then
        set @i:=@i+1;
        insert into exam_datas(vimei,vbas_run_date_time,vbit_run_date_time,vdifff) Values (@imei, @run_date_time2, @run_date_time, @max_mes-@min_mes);
        SELECT * FROM exam_datas;
        SET @asd =1;
      elseif 1=1 then
        set @i:=@i+1;
        End if;
  ELSEIF 1=1
    THEN SET @i:=@i+1;
  End if;
  IF (@i = @kayit_sayisi) 
    THEN set @tamam =1; LEAVE myloop;
  END IF;
END WHILE myloop;  

select DISTINCT * from exam_datas;


drop table if exists exam_datas;
drop table if exists gps_table;
END

我需要： id= 6 first true 和 id= 11 last_true

firs_true - last_true = 304-290= 14

  1. id=14 第一个 true 和 id=18 last_true

firs_true - last_true = 332-324= 8

这个套路太慢了。

MySql 版本为 5.7，表中有 2 百万条记录。

更新：

查询在这里。这里

谢谢@LukStorms

4

1 回答 1

0

有可能在 1 个查询中得到这样的结果。
从而避免对记录进行 WHILE 循环。

此示例在不使用窗口函数的情况下工作。只需使用查询中的变量来计算排名。然后用于获取组的最小值和最大值。

select
 imei, 
 min(run_date_time) as start_dt,
 max(run_date_time) as stop_dt, 
 max(add_meter) - min(add_meter) as diff
from
( 
  select imei, id, run_date_time, add_meter, in_here, 
  case
  when @prev_imei = imei and @prev_ih = in_here then @rnk
  when @rnk := @rnk + 1 then @rnk
  end as rnk,
  @prev_imei := imei as prev_imei,
  @prev_ih := in_here as prev_ih
  from example_table t
  cross join (select @rnk := 0, @prev_ih := null, @prev_imei := null) vars
  order by imei, id, run_date_time
) q
where in_here = 1 
group by imei, rnk 
having count(*) > 4
order by imei, min(id);

在该过程中，此类查询可用于填充该最终临时表。

对db<>fiddle的测试在这里

于 2019-01-23T21:11:05.227 回答