Please explain the query?

Question

hello all,
please explain below query used in solution below, thanks in advance!
ELECT MAX(P1.ET) AS ST, P2.ST AS ET
FROM XYZ AS P1
INNER JOIN XYZ AS P2 ON (P1.ST < P2.ST)
GROUP BY P2.ST
HAVING MAX(P1.ET) < P2.ST
IF OBJECT_ID('XYZ') IS NOT NULL
DROP TABLE XYZ
GO
CREATE TABLE XYZ
id int identity(1,1),
ST smalldatetime NOT NULL,
ET smalldatetime NOT NULL
GO
INSERT INTO XYZ (ST, ET)
VALUES ('2010-01-01 9:00AM', '2010-01-01 10:00AM')
INSERT INTO XYZ (ST, ET)
VALUES ('2010-01-01 9:00AM', '2010-01-01 12:00PM')
INSERT INTO XYZ (ST, ET)
VALUES ('2010-01-01 1:00PM', '2010-01-01 2:00PM')
INSERT INTO XYZ (ST, ET)
VALUES ('2010-01-01 3:00PM', '2010-01-01 5:00PM')
INSERT INTO XYZ (ST, ET)
VALUES ('2010-01-01 11:00AM', '2010-01-01 12:00PM')
GO
WITH Gaps(Gap) AS
SELECT COALESCE(SUM(DATEDIFF(MINUTE,ST,ET)), 0)
FROM (
SELECT MAX(P1.ET) AS ST, P2.ST AS ET
FROM XYZ AS P1
INNER JOIN XYZ AS P2 ON (P1.ST < P2.ST)
GROUP BY P2.ST
HAVING MAX(P1.ET) < P2.ST
) gaps
SELECT (
COALESCE(DATEDIFF(MINUTE, MIN(ST), MAX(ET)), 0)
- (SELECT Gap FROM Gaps)
) / 60.0 TotalHrs
FROM XYZ

Answer 1

SELECT MAX(P1.ET) AS ST, P2.ST AS ET
FROM XYZ AS P1
INNER JOIN XYZ AS P2 ON (P1.ST < P2.ST)
GROUP BY P2.ST
HAVING MAX(P1.ET) < P2.ST
Finds all of the gaps (that is time that is not in any interval in your original data.  To see how it works, let's look at an example.  I'm going to use data that is a little different than your original data because your data has no gaps. 
The query still works if you have no gaps, but it is easier to see what it is doing if the data has some gaps.  Also, I'm going to explicitly set the id column instead of making it an identity.  This will make it a little easier to identify each
row in the following explanation. So the data I'm going to work with is
CREATE TABLE XYZ
id int,
ST smalldatetime NOT NULL,
ET smalldatetime NOT NULL
GO
INSERT INTO XYZ (id, ST, ET)
VALUES (1, '2010-01-01 9:00AM', '2010-01-01 10:00AM')
INSERT INTO XYZ (id, ST, ET)
VALUES (2, '2010-01-01 9:00AM', '2010-01-01 12:00PM')
INSERT INTO XYZ (id, ST, ET)
VALUES (3, '2010-01-02 1:00PM', '2010-01-02 2:00PM')
INSERT INTO XYZ (id, ST, ET)
VALUES (4, '2010-01-03 3:00PM', '2010-01-03 5:00PM')
INSERT INTO XYZ (id, ST, ET)
VALUES (5, '2010-01-03 7:00PM', '2010-01-03 9:00PM')
Notice that the gaps here are from row 2 to row 3 (12PM on the 1st to 1PM on the 2nd) and row 3 to row 4 (2PM on the 2nd to 3PM on the 3rd) and row 4 to row 5 (5PM on the 3rd to 7PM on the 3rd).  So that's what the above subquery should be finding for
us.
To see what a query you don't understand is doing, simplify it to the smallest part you can and see what it returns and then build it up to the final query.  So the simplest thing we can do is just the from clause.  That gives us
SELECT *
FROM XYZ AS P1
INNER JOIN XYZ AS P2 ON (P1.ST < P2.ST)
/* That gives us the result
1 2010-01-01 09:00:00 2010-01-01 10:00:00 3 2010-01-02 13:00:00 2010-01-02 14:00:00
2 2010-01-01 09:00:00 2010-01-01 12:00:00 3 2010-01-02 13:00:00 2010-01-02 14:00:00
1 2010-01-01 09:00:00 2010-01-01 10:00:00 4 2010-01-03 15:00:00 2010-01-03 17:00:00
2 2010-01-01 09:00:00 2010-01-01 12:00:00 4 2010-01-03 15:00:00 2010-01-03 17:00:00
3 2010-01-02 13:00:00 2010-01-02 14:00:00 4 2010-01-03 15:00:00 2010-01-03 17:00:00
1 2010-01-01 09:00:00 2010-01-01 10:00:00 5 2010-01-03 19:00:00 2010-01-03 21:00:00
2 2010-01-01 09:00:00 2010-01-01 12:00:00 5 2010-01-03 19:00:00 2010-01-03 21:00:00
3 2010-01-02 13:00:00 2010-01-02 14:00:00 5 2010-01-03 19:00:00 2010-01-03 21:00:00
4 2010-01-03 15:00:00 2010-01-03 17:00:00 5 2010-01-03 19:00:00 2010-01-03 21:00:00
Now we want to Group by P2.ST and get the MAX(P1.ET) and P2.ST, so that gives us
SELECT MAX(P1.ET) AS ST, P2.ST AS ET
FROM XYZ AS P1
INNER JOIN XYZ AS P2 ON (P1.ST < P2.ST)
GROUP BY P2.ST
/* Result is
2010-01-01 12:00:00 2010-01-02 13:00:00
2010-01-02 14:00:00 2010-01-03 15:00:00
2010-01-03 17:00:00 2010-01-03 19:00:00
Now with this sample data there are no rows in the output with MAX(P1.ET) > P2.ST.  But if there was one, you would not want that row because it is not a real gap (obviously, a gap can't start today and end yesterday).  (If you want to see how
you could get a case like that, add a row 6 to the sample data with a ST of 2010-01-01 7:00PM and an ET of 2010-01-03 9:00PM.)
So we add a HAVING MAX(P1.ET) < P2.ST to remove those cases.
That leaves us with all of the gaps.  So then with
SELECT COALESCE(SUM(DATEDIFF(MINUTE,ST,ET)), 0)
FROM (
SELECT MAX(P1.ET) AS ST, P2.ST AS ET
FROM XYZ AS P1
INNER JOIN XYZ AS P2 ON (P1.ST < P2.ST)
GROUP BY P2.ST
HAVING MAX(P1.ET) < P2.ST
) gaps
we get the total amount of time in all gaps.  Then the final result is just the time from the earliest ST to the latest ET minus the total time from the gap.
Tom