Split one record into multiple based on another table - sql

Question

I have a table tracking location stays. There's an ID, startdatetime, enddatetime, and other fields.
I have another table with events that occur within each of those stays, with similar start and end times, and linked on the ID field.
What I need to do is merge the two and split the location table up into its individual events. The trick here is a location may start on 2017-08-02 but the first event might not start for a few days. Thus i'd need a record for that gap at the start.
sample data
CREATE TABLE #Stays (
EpID INT, StayId INT, StayStartDate DateTime, StayEndDate DateTime);
CREATE TABLE #Events (
EpID INT, EventId INT, EventStartDate DateTime, EventEndDate DateTime, EventNumber INT);
INSERT INTO #Events SELECT 1, 7897, '2016-11-24 00:00:00.000','2016-11-26 00:00:00.000', 1
INSERT INTO #Events SELECT 1, 7898, '2016-11-26 00:00:00.000','2016-11-28 00:00:00.000', 2
INSERT INTO #Stays SELECT 1, 10, '2016-11-22 08:15:00.000','2016-11-24 10:54:00.000'
INSERT INTO #Stays SELECT 1, 11, '2016-11-24 10:54:00.000','2016-11-24 11:17:00.000'
INSERT INTO #Stays SELECT 1, 12, '2016-11-24 11:17:00.000','2016-11-25 08:16:00.000'
INSERT INTO #Stays SELECT 1, 13, '2016-11-25 08:16:00.000','2016-11-28 23:15:00.000'
expected output would be
EpId StartDate EndDate EventNumber
1 2016-11-22 08:15:00.000 2016-11-23 23:59:59.000 NULL
1 2016-11-24 00:00:00.000 2016-11-25 23:59:59.000 7897
1 2016-11-26 00:00:00.000 2016-11-27 23:59:59.000 7898
1 2016-11-28 00:00:00.000 2016-11-28 23:15:00.000 NULL
here is what i'm trying. It currently doesn't work properly, and i'm sure the method i'm working on is probably not the best. It's currently not melding the two datasets together.
My guess is theres a much easier way to do it with outer or cross apply, but my knowledge of how they work is rather limited.
Any help?
;with e as (
SELECT [EpID]
,EventId
,[EventNumber]
,case when [EventStartDate] > DayStart then [EventStartDate] else DayStart end as [EventStart]
,case when [EventEndDate] < DayEnd then [EventEndDate] else DayEnd end as [EventEnd]
FROM [Events] e
inner join DimStaySegmentDayReference d on d.DayEnd >= e.[EventStartDate] and d.DayStart <= e.[EventEndDate]
),
s as (
select
[EpID]
,StayId
,case when StayStartDate > DayStart then StayStartDate else DayStart end as [StayStart]
,case when StayEndDate < DayEnd then StayEndDate else DayEnd end as [StayEnd]
from Stays s
inner join DimStaySegmentDayReference d on d.DayEnd >= StayStartDate and d.DayStart <= StayEndDate
),
u as (select 'stay' as source, [EpID], StayStart, StayEnd, '' as event from s
union all
select 'event' as source, [EpID], [EventStart], [EventEnd], eventnumber as event from e)
select Source,
[EpID],
Staystart,
stayend,
case when lag(stayend) over (partition by EpId ORDER BY STAYSTART) < StayEnd-0.0001 AND source='event' then lag(stayend) over (partition by EpId ORDER BY STAYSTART) else staystart end as staystartnew,
case when lead(staystart) over (partition by EpID ORDER BY StayStart) < stayend then lead(staystart) over (partition by EpID ORDER BY StayStart) else stayend end as stayendnew,
event
from u
where StayStart <> stayend
order by StayStart
The DayReference table is simply every day with a start and end time so i can split the record into day segments.
I'm using SQL Server 2012
Edit for some context
I've updated my sample data to make it a bit clearer.
The stay table tracks location stays. In this provided case i'm ignoring multiple locations to make finding a solution easier.
Locations and Events are agnostic to each other, other than occurring for the same EpID within the same time frame.
As an example consider tracking time at work, you start at 9am and finish at 5pm. For this work day you'll have say 5 location stays making up the full shift. 9-11 desk, 11-12 meeting, 12-1 lunch, 1-3 meeting, 3-5 desk.
You then have a series of events, lets call it drinking coffee. You drink coffee between 9:30 and 10, and 2-4.
What I need to do is mesh together these two sets of data creating a single timeline.
9-930 desk, 930-10 coffee, 10-11 desk, 11-12 meeting, 12-1 lunch, 1-2 meeting, 2-4 coffee, 4-5 desk.
Hope this helps

Answer 1

Probably some things can be simplified, but will be easy to read what I am validating for each case, also, I think that one row is missing in your output example, I got a last one from 2018-09-14 16:00 To 2018-09-15 12:00 and I did not find a reason on the logic or the question to discard it
Extra validations and a left join to the Stays with no registered events would be needed, but here is my approach
;WITH CTE AS (
SELECT D.*, s.StayId,
EventNumber,
LAG(D.DStart) OVER (ORDER BY EventNumber) As LagStart,
LAG(StayID) OVER (ORDER BY EventNumber) As LagStay,
LAG(Event) OVER (ORDER BY EventNumber) As LagEvent,
LEAD(D.DEnd) OVER (ORDER BY EventNumber) As LeadEnd,
LEAD(StayID) OVER (ORDER BY EventNumber) As LeadStay,
LEAD(Event) OVER (ORDER BY EventNumber) As LeadEvent
FROM #Events E
CROSS APPLY
(
SELECT TOP 1 * FROM #Stays S WHERE E.EventStartDate BETWEEN S.StayStartDate AND S.StayEndDate
UNION
SELECT TOP 1 * FROM #Stays S WHERE E.EventEndDate BETWEEN S.StayStartDate AND S.StayEndDate
) S
CROSS APPLY (
SELECT StayStartDate AS DStart, EventStartDate DEnd, Null AS Event, 1 as c WHERE StayStartDate < EventStartDate
UNION
SELECT EventStartDate, EventEndDate, EventNumber, 2 WHERE EventStartDate >= StayStartDate AND EventEndDate <= StayEndDate
UNION
SELECT StayStartDate, EventEndDate, EventNumber, 3 WHERE StayStartDate > EventStartDate AND EventEndDate < StayEndDate
UNION
SELECT EventStartDate, StayEndDate, EventNumber, 4 WHERE StayStartDate < EventStartDate AND EventEndDate > StayEndDate
UNION
SELECT EventEndDate, StayEndDate, Null, 5 WHERE EventEndDate < StayEndDate
) D
)
SELECT DISTINCT
CASE WHEN LagStay = StayId AND Event IS NULL AND LagEvent IS NULL THEN LagStart
ELSE DStart END AS StartDate,
CASE WHEN LeadStay = StayId AND Event IS NULL AND LeadEvent IS NULL THEN LeadEnd
ELSE DEnd END AS EndDate,
Event, StayID
FROM CTE
ORDER BY StartDate