这样做是为了保持解决方案易于维护,而无需一次性完成最终查询,这将使其大小几乎翻倍(在我看来).这是因为结果需要匹配并在具有匹配的In和Out事件的一行上表示.所以最后,我使用了一些工作表.它在存储过程中实现.

存储过程使用通过交叉连接引入的多个变量.将交叉连接视为初始化变量的机制.变量是安全维护的,所以我相信,这个document的精神经常在变量查询中引用.参考的重要部分是在线上安全处理变量,迫使它们在使用它们的其他列之前设置.这是通过maximum()和least()函数实现的,这些函数的优先级高于不使用这些函数的变量.另请注意,coalesce()通常用于相同的目的.如果它们的使用看起来很奇怪,例如将已知大于0或0的数字中的最大值,那么这是故意的.故意强制设置变量的优先顺序.

查询中名为dummy2等的列是未使用输出的列,但它们用于设置变量(例如,最大值()或其他变量).这是上面提到的.像7777这样的输出是第3个插槽中的占位符,因为使用的if()需要一些值.所以忽略这一切.

我已经包含了几个代码的屏幕截图,因为它逐层进行,以帮助您可视化输出.这些迭代的开发如何慢慢地进入下一阶段以扩展先前的阶段.

我相信我的同行可以在一个查询中改进这一点.我本可以这样完成它.但我相信这会导致一个令人困惑的混乱,如果被触及就会破裂.

架构：

create table attendance2(Id int, DateTime datetime, Door char(20), Active_door char(20));
INSERT INTO attendance2 VALUES
( 1, '2016-01-01 08:00:00', 'In', ''),
( 2, '2016-01-01 09:00:00', 'Out', ''),
( 3, '2016-01-01 09:15:00', 'In', ''),
( 4, '2016-01-01 09:30:00', 'In', ''),
( 5, '2016-01-01 09:35:00', '', 'On'),
( 6, '2016-01-01 10:00:00', 'Out', ''),
( 7, '2016-01-01 16:00:00', '', 'Off');
drop table if exists oneLinersDetail;
create table oneLinersDetail
( -- architect this depending on multi-user concurrency
id int not null,
dt datetime not null,
door int not null,
grpIn int not null,
grpInSeq int not null,
grpOut int not null,
grpOutSeq int not null
);
drop table if exists oneLinersSummary;
create table oneLinersSummary
( -- architect this depending on multi-user concurrency
id int not null,
grpInSeq int null,
grpOutSeq int null,
checkIn datetime null, -- we are hoping in the end it is not null
checkOut datetime null -- ditto
);
存储过程：
DROP PROCEDURE IF EXISTS fetchOneLiners;
DELIMITER $$
CREATE PROCEDURE fetchOneLiners()
BEGIN
truncate table oneLinersDetail; -- architect this depending on multi-user concurrency
insert oneLinersDetail(id,dt,door,grpIn,grpInSeq,grpOut,grpOutSeq)
select id,dt,door,grpIn,grpInSeq,grpOut,grpOutSeq
from
( select id,dt,door,
if(@lastEvt!=door and door=1,
greatest(@grpIn:=@grpIn+1,0),
7777) as dummy2, -- this output column we don't care about (we care about the variable being set)
if(@lastEvt!=door and door=2,
greatest(@grpOut:=@grpOut+1,0),
7777) as dummy3, -- this output column we don't care about (we care about the variable being set)
if (@lastEvt!=door,greatest(@flip:=1,0),least(@flip:=0,1)) as flip,
if (door=1 and @flip=1,least(@grpOutSeq:=0,1),7777) as dummy4,
if (door=1 and @flip=1,greatest(@grpInSeq:=1,0),7777) as dummy5,
if (door=1 and @flip!=1,greatest(@grpInSeq:=@grpInSeq+1,0),7777) as dummy6,
if (door=2 and @flip=1,least(@grpInSeq:=0,1),7777) as dummy7,
if (door=2 and @flip=1,greatest(@grpOutSeq:=1,0),7777) as dummy8,
if (door=2 and @flip!=1,greatest(@grpOutSeq:=@grpOutSeq+1,0),7777) as dummy9,
@grpIn as grpIn,
@grpInSeq as grpInSeq,
@grpOut as grpOut,
@grpOutSeq as grpOutSeq,
@lastEvt:=door as lastEvt
from
( select id,`datetime` as dt,
CASE
WHEN Door='in' or Active_door='on' THEN 1
ELSE 2
END as door
from attendance2
order by id
) xD1 -- derived table #1
cross join (select @grpIn:=0,@grpInSeq:=0,@grpOut:=0,@grpOutSeq:=0,@lastEvt:=-1,@flip:=0) xParams
order by id
) xD2 -- derived table #2
order by id;
-- select * from oneLinersDetail;
truncate table oneLinersSummary; -- architect this depending on multi-user concurrency
insert oneLinersSummary (id,grpInSeq,grpOutSeq,checkIn,checkOut)
select distinct grpIn,null,null,null,null
from oneLinersDetail
order by grpIn;
-- select * from oneLinersSummary;
update oneLinersSummary ols
join
( select grpIn,max(grpInSeq) m
from oneLinersDetail
where door=1
group by grpIn
) d1
on d1.grpIn=ols.id
set ols.grpInSeq=d1.m;
-- select * from oneLinersSummary;
update oneLinersSummary ols
join
( select grpOut,max(grpOutSeq) m
from oneLinersDetail
where door=2
group by grpOut
) d1
on d1.grpOut=ols.id
set ols.grpOutSeq=d1.m;
-- select * from oneLinersSummary;
update oneLinersSummary ols
join oneLinersDetail old
on old.door=1 and old.grpIn=ols.id and old.grpInSeq=ols.grpInSeq
set ols.checkIn=old.dt;
-- select * from oneLinersSummary;
update oneLinersSummary ols
join oneLinersDetail old
on old.door=2 and old.grpOut=ols.id and old.grpOutSeq=ols.grpOutSeq
set ols.checkOut=old.dt;
-- select * from oneLinersSummary;
-- dump out the results
select id,checkIn,checkOut
from oneLinersSummary
order by id;
-- rows are left in those two tables (oneLinersDetail,oneLinersSummary)
END$$
DELIMITER ;
测试：
call fetchOneLiners();
+----+---------------------+---------------------+
| id | checkIn | checkOut |
+----+---------------------+---------------------+
| 1 | 2016-01-01 08:00:00 | 2016-01-01 09:00:00 |
| 2 | 2016-01-01 09:35:00 | 2016-01-01 16:00:00 |
+----+---------------------+---------------------+

这是答案的结尾.以下是开发人员可视化完成存储过程的步骤.