Calculating Timestamp Overlap Duration - sql

Question

I have a business that rents out one particular product. I would like to know the duration in minutes of when a particular location is out of stock.
The first data set has transaction history with the location ID, rental starting time stamp, and rental ending time stamp.
The second data set has the location ID, date, and number of units available for rent that day. The number of units can change day to day as units are added/removed frequently.
I need to calculate how many minutes per day per location that all available units were out on rent.
Ex: Location A has 3 units on 2/1/2016. How many minutes (if any) on 2/1/2016 were there 3 units out on rent at the same time?
SQL is the language I need to use.
See sample data set Y below:
LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS
A 30Jan2016 19:54:37.000 01Feb2016 10:00:24.053
A 31Jan2016 16:30:23.000 01Feb2016 9:07:06.588
A 01Feb2016 9:22:22.000 02Feb2016 10:00:23.716
A 01Feb2016 9:36:11.000 01Feb2016 11:05:34.249
A 01Feb2016 10:27:34.000 01Feb2016 12:59:29.883
A 01Feb2016 10:40:38.000 01Feb2016 15:36:27.119
A 01Feb2016 12:43:10.000 01Feb2016 14:23:15.914
A 01Feb2016 13:28:20.000 01Feb2016 14:40:15.573
A 01Feb2016 17:03:13.000 01Feb2016 19:02:57.413
A 01Feb2016 17:17:12.000 01Feb2016 18:54:14.708
Sample data set Z below:
LOC_ID, Date, Unit_Count
A 01Feb2016 3
A 02Feb2016 4
A 03Feb2016 3
B 01Feb2016 2
B 02Feb2016 2
B 03Feb2016 2
Since Location A has 3 total units on Feb 1st then the desired output would be 25 minutes which is the total amount of time that 3 units were out on rent at the same time on Feb 1st at Location A. Between 10:40am and 11:05am 3 units were out on rent at the same time.

Answer 1

Here is some T-SQL code (you didn't indicate which DBMS you were using)
-- DECLARING SAMPLE DATA
DECLARE #Z table (LOC_ID char, Date date, Unit_Count int)
INSERT #Z ( LOC_ID, Date, Unit_Count ) VALUES ( 'A', '2016-02-01', 3)
INSERT #Z ( LOC_ID, Date, Unit_Count ) VALUES ( 'A', '2016-02-02', 4)
INSERT #Z ( LOC_ID, Date, Unit_Count ) VALUES ( 'A', '2016-02-03', 3)
INSERT #Z ( LOC_ID, Date, Unit_Count ) VALUES ( 'B', '2016-02-01', 2)
INSERT #Z ( LOC_ID, Date, Unit_Count ) VALUES ( 'B', '2016-02-02', 2)
INSERT #Z ( LOC_ID, Date, Unit_Count ) VALUES ( 'B', '2016-02-03', 2)
DECLARE #Y table (LOC_ID char, ACT_RNTL_BGN_TS datetime, ACT_RNTL_END_TS datetime)
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-01-30 19:54:37.000', '2016-02-01 10:00:24.053')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-01-31 16:30:23.000', '2016-02-01 09:07:06.588')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 09:22:22.000', '2016-02-02 10:00:23.716')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 09:36:11.000', '2016-02-01 11:05:34.249')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 10:27:34.000', '2016-02-01 12:59:29.883')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 10:40:38.000', '2016-02-01 15:36:27.119')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 12:43:10.000', '2016-02-01 14:23:15.914')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 13:28:20.000', '2016-02-01 14:40:15.573')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 17:03:13.000', '2016-02-01 19:02:57.413')
INSERT #Y ( LOC_ID, ACT_RNTL_BGN_TS, ACT_RNTL_END_TS ) VALUES ('A', '2016-02-01 17:17:12.000', '2016-02-01 18:54:14.708')
;
-- START OF QUERY
WITH SplittedRentedIntervals AS ( -- Intervals that span more than one day are splited.
SELECT Z.LOC_ID
, Z.Date
, BGN = CASE WHEN Y.ACT_RNTL_BGN_TS > CAST(Z.Date AS datetime) THEN Y.ACT_RNTL_BGN_TS ELSE CAST(Z.Date AS datetime) END
, [END] = CASE WHEN Y.ACT_RNTL_END_TS < CAST(DATEADD(DAY, 1, Z.Date) AS datetime) THEN Y.ACT_RNTL_END_TS ELSE CAST(DATEADD(DAY, 1, Z.Date) AS datetime) END
FROM #Z Z
JOIN #Y Y
ON Y.LOC_ID = Z.LOC_ID
AND (Z.Date = CAST(Y.ACT_RNTL_BGN_TS AS date)
OR Z.Date = CAST(Y.ACT_RNTL_END_TS AS date)
OR (Z.Date BETWEEN Y.ACT_RNTL_BGN_TS AND Y.ACT_RNTL_END_TS))
), Times AS ( -- All times starting and ending any interval, including start and end of day.
SELECT DISTINCT LOC_ID, Date, TIME = BGN FROM SplittedRentedIntervals
UNION SELECT DISTINCT LOC_ID, Date, TIME = [END] FROM SplittedRentedIntervals
UNION SELECT LOC_ID, Date, TIME = CAST(Date AS datetime) FROM #Z
UNION SELECT LOC_ID, Date, TIME = CAST(DATEADD(DAY, 1, Date) AS datetime) FROM #Z
), OrderedTimes AS (
SELECT LOC_ID
, Date
, TIME
, NUM = ROW_NUMBER() OVER (PARTITION BY LOC_ID, Date ORDER BY TIME ASC)
FROM Times
), Intervals AS ( -- Intervals are conformed by two consecutives times.
SELECT OT1.LOC_ID
, OT1.Date
, BGN = OT1.TIME
, [END] = OT2.TIME
FROM OrderedTimes OT1
JOIN OrderedTimes OT2
ON OT2.LOC_ID = OT1.LOC_ID
AND OT2.Date = OT1.Date
AND OT2.NUM = OT1.NUM + 1
), StockByInterval AS ( -- Intersect all time intervals with rented intervals to calculate rented units.
SELECT I.LOC_ID
, I.Date
, I.BGN
, I.[END]
, STOCK = Z.Unit_Count
- (SELECT COUNT(*)
FROM SplittedRentedIntervals SRI
WHERE I.Date = SRI.Date
AND I.LOC_ID = SRI.LOC_ID
AND SRI.BGN < I.[END]
AND SRI.[END] > I.BGN)
FROM Intervals I
JOIN #Z Z
ON Z.Date = I.Date
AND Z.LOC_ID = I.LOC_ID
), WithuotStock AS ( -- Sum the minutes of intervals where there is no stock.
SELECT LOC_ID
, Date
, MinutesWithoutStock = SUM(DATEDIFF(MINUTE, BGN, [END]))
FROM StockByInterval
WHERE STOCK <= 0 -- Sample data has some intervals where there are more items rented than are available.
GROUP BY LOC_ID, Date
)
SELECT Z.LOC_ID
, Z.Date
, MinutesWithoutStock = ISNULL(WS.MinutesWithoutStock, 0)
FROM #Z Z
LEFT JOIN WithuotStock WS
ON WS.Date = Z.Date
AND WS.LOC_ID = Z.LOC_ID
ORDER BY Z.LOC_ID, Z.Date
Sample output
LOC_ID Date MinutesWithoutStock
------ ---------- -------------------
A 2016-02-01 374
A 2016-02-02 0
A 2016-02-03 0
B 2016-02-01 0
B 2016-02-02 0
B 2016-02-03 0