假设我有两个实体：事件和活动
事件是在（看似）随机时间发生的事情，如日出，日落，风暴，雾等。
我有一张表：
create table Event (
eventKey int,
eventDesc varchar(100),
started datetime
)
EventKey | EventDesc | Started
1 "Sunset" 2009-07-03 6:51pm
2 "Sunrise" 2009-07-04 5:33am
3 "Fog" 2009-07-04 5:52pm
4 "Sunset" 2009-07-04 6:49pm
5 "Full Moon" 2009-07-04 10:12pm
6 "Sunrise" 2009-07-05 5:34am
然后我有一张人们参与的活动表，以及他们与之相关的活动（即一个行动可以长时间运行并跨越多个事件：“周末露营”）：
create table EventTask (
activityKey int,
activityDesc varchar(100),
startEventKey int,
endEventKey int
)
ActivityKey | ActivityDesc | StartEventKey | EndEventKey
123 "Camp-out" 1 5
234 "Drive home" 6 6
我想输出由发生的事件标记的操作的时间轴：
ActivityKey | ActivityDesc | EventKey | EventDesc
123 "Camp-out" 1 "Sunset"
123 "Camp-out" 2 "Sunrise"
123 "Camp-out" 3 "Fog"
123 "Camp-out" 4 "Sunset"
123 "Camp-out" 5 "Full Moon"
234 "Drive Home" 6 "Sunrise"
是否可以编写一个以线性时间similar to this question执行此操作的查询？还请推荐索引或您能想到的任何其他优化。当前的解决方案是用C＃编写的，但我会喜欢快速的SQL解决方案。
执行此操作的最佳查询是什么？
答案 0 :(得分：2)
/*
create table Event (
eventKey int,
eventDesc varchar(100),
started timestamp
);
insert into event values( 1, 'Sunset' , '2009-07-03 6:51pm');
insert into event values(2, 'Sunrise', '2009-07-04 5:33am');
insert into event values(3, 'Fog' , '2009-07-04 5:52pm');
insert into event values(4, 'Sunset' , '2009-07-04 6:49pm');
insert into event values(5, 'Full Moon', '2009-07-04 10:12pm');
insert into event values(6, 'Sunrise' , '2009-07-05 5:34am');
select * from event;
create table EventTask (
activityKey int,
activityDesc varchar(100),
startEventKey int,
endEventKey int
)
insert into eventtask values(123 , 'Camp-out', 1 , 5);
insert into eventtask values(234, 'Drive home', 6, 6);
select * from eventtask;
*/
select a.activitykey, a.activitydesc, b.eventkey, b.eventdesc
from
eventtask a
join event b on b.eventkey between a.starteventkey and a.endeventkey
order by
a.activitykey, b.eventkey;
activitykey activitydesc eventkey eventdesc
-------------- --------------- ----------- ------------
123 Camp-out 1 Sunset
123 Camp-out 2 Sunrise
123 Camp-out 3 Fog
123 Camp-out 4 Sunset
123 Camp-out 5 Full Moon
234 Drive home 6 Sunrise
6 record(s) selected [Fetch MetaData: 3/ms] [Fetch Data: 1/ms]
[Executed: 7/7/09 4:24:34 PM EDT ] [Execution: 15/ms]
如果你的表很大，你肯定需要在event.eventkey，eventtask.starteventkey和eventtask.endeventkey上建立索引。
请注意，索引可提高查询速度，但会降低插入和更新速度。
这是不需要event.eventkey列具有重要性的版本（更正确）：
select a.activitykey, a.activitydesc, d.eventkey, d.eventdesc
from
eventtask a
join event b on b.eventkey = a.starteventkey
join event c on c.eventkey = a.endeventkey
join event d on d.started between b.started and c.started
order by
a.activitykey, d.started;
activitykey activitydesc eventkey eventdesc
-------------- --------------- ----------- ------------
123 Camp-out 1 Sunset
123 Camp-out 2 Sunrise
123 Camp-out 3 Fog
123 Camp-out 4 Sunset
123 Camp-out 5 Full Moon
234 Drive home 6 Sunrise
6 record(s) selected [Fetch MetaData: 2/ms] [Fetch Data: 0/ms]
[Executed: 7/8/09 10:01:25 AM EDT ] [Execution: 4/ms]
答案 1 :(得分：1)
我会重新定义活动表，因此有一个startTime和一个EndTime，而不是基于随机事件。然后，如果我真的想看看那段时间发生的'事件'，我会加入时间范围。从OO /灵活性的角度来看，这更有意义，尽管您会看到更高的性能成本。
declare @Event table(
id int,
name varchar(100),
[time] datetime
);
insert into @Event values(1, 'Sunset', '2009-07-03 6:51pm');
insert into @Event values(2, 'Sunrise', '2009-07-04 5:33am');
insert into @Event values(3, 'Fog', '2009-07-04 5:52pm');
insert into @Event values(4, 'Sunset', '2009-07-04 6:49pm');
insert into @Event values(5, 'Full Moon', '2009-07-04 10:12pm');
insert into @Event values(6, 'Sunrise', '2009-07-05 5:34am');
select * from @Event;
declare @Activity table (
id int,
name varchar(100),
startTime datetime,
endTime datetime
)
insert into @Activity values(123, 'Camp-out', '2009-07-03 6:00pm', '2009-07-05 5:00am');
insert into @Activity values(234, 'Drive home', '2009-07-05 5:00am', '2009-07-05 6:00am');
select *
from @Activity A
join @Event E on E.[time] > A.startTime and E.[time] < A.endTime
order by A.startTime
答案 2 :(得分：1)
我最近写了两种方法来优化这些查询（加入BETWEEN条件）：Using CROSS APPLY to optimize joins on BETWEEN conditions
可能的查询（无法在没有样本INSERT的情况下进行测试）：
SELECT et.activityKey,
et.activityDesc,
e.*
FROM Event AS e CROSS APPLY(SELECT TOP 1 * FROM EventTask AS et
WHERE et.startEventKey <= e.started
AND e.started < endEventKey
ORDER BY et.startEventKey
) AS et