Start and End Dates check

Question

Hi,
I have a requirement to find the list of persons who have overlapping dates.
Each person can have many start and End Dates, but only one row that has a NULL end date, the active record. If the start and end dates are overlapped, like in the example below, the second row should have a start date of 24-DEC-2006 and not 22-DEC-2006. Is it possible to list those persons who have overlapping dates/gap in the dates.
Empl_ID Start Date End Date
12345 01-JAN-2006 23-DEC-2006
12345 22-DEC-2006 13-JAN-2007
12345 14-JAN-2007 NULL
Any help is greatly appreciated.
Thanks!

Answer 1

Hi Patrick
I like where you are coming from. I hope you don't mind but I have extended your idea slightly to add in a PARTITION BY clause to make sure we look at records for the same person. Also, if we LAG to get a previous record shouldn't we LAG the END_DATE and compare it with the current START_DATE?
LAG(END_DATE,1) OVER (PARTITION BY EMPL_ID ORDER BY START_DATE)
Doing this, then the results would look like this:
Empl_ID Start Date End Date Lag Date
12345 01-JAN-2006 23-DEC-2006 NULL
12345 22-DEC-2006 13-JAN-2007 23-DEC-2006
12345 14-JAN-2007 NULL 13-JAN-2007
This would give the SIGN an issue because in the first record the LAG would be NULL and in the last record the END_DATE would be NULL. So it looks like some manipulation of NULLs has to take place as well which I will deal with later.
SIGN((LAG(END_DATE,1) OVER (PARTITION BY EMPL_ID ORDER BY START_DATE)) - START_DATE)
Anytime the difference between the Lag Date and the Start Date is positive this means this is an overlap which means we do need to look for any variance where the SIGN is 1.
Using the other function, we could use LEAD like this:
LEAD(START_DATE,1) OVER (PARTITION BY EMPL_ID ORDER BY START_DATE)
Doing this, then the results would look like this:
Empl_ID Start Date End Date Lead Date
12345 01-JAN-2006 23-DEC-2006 22-DEC-2006
12345 22-DEC-2006 13-JAN-2007 14-JAN-2007
12345 14-JAN-2007 NULL NULL
which when combined with SIGN becomes: SIGN((LEAD(START_DATE,1) OVER (PARTITION BY EMPL_ID ORDER BY START_DATE)) - END_DATE)
This time you would be looking for the SIGN to be -1
Personally, I think I would prefer the LEAD function because a) there would always be a START_DATE, and b) we could replace all of the possible NULLs with the SYSDATE to get a final calculation like this:
SIGN(NVL(LEAD(START_DATE,1) OVER (PARTITION BY EMPL_ID ORDER BY START_DATE),TRUNC(SYSDATE)) - NVL(END_DATE, TRUNC(SYSDATE)))
Based on the current date being 17-NOV-2009, you would now get this result:
Empl_ID Start Date End Date Lead Date
12345 01-JAN-2006 23-DEC-2006 22-DEC-2006
12345 22-DEC-2006 13-JAN-2007 14-JAN-2007
12345 14-JAN-2007 17-NOV-2009 17-NOV-2009
Best wishes
Michael