SQL, using Group by until specific trend (increment, decrement, same) - sql-server

Question

I would like to know how can i modify my code for considering all the same values of suppose 10 as UP till the time it is incrementing and then down for decrement and SAME if there is no change till the time there is no variation in the value (increment, decrement, same).
Here is my code :
;with etape1 as
(
select ROW_NUMBER() OVER(ORDER BY mnth) AS id,* from [InsideTSQL2008].[alioune].[Sales]
)
,
etape2 as
(
select
a.id, b.mnth AS START , a.mnth AS FINISH ,
a.qty - b.qty AS TREND
FROM
etape1 a
LEFT JOIN etape1 b
on a.id = b.id+1
)
select * from etape2;
My Result is :
id START FINISH TREND
1 NULL 2007-12-01 NULL
2 2007-12-01 2008-01-01 10
3 2008-01-01 2008-02-01 10
4 2008-02-01 2008-03-01 10
5 2008-03-01 2008-04-01 10
6 2008-04-01 2008-05-01 0
7 2008-05-01 2008-06-01 -10
8 2008-06-01 2008-07-01 -10
9 2008-07-01 2008-08-01 -10
10 2008-08-01 2008-09-01 -10
11 2008-09-01 2008-10-01 10
12 2008-10-01 2008-11-01 -10
13 2008-11-01 2008-12-01 20
14 2008-12-01 2009-01-01 10
15 2009-01-01 2009-02-01 10
16 2009-02-01 2009-03-01 -40
My final result as required should be like :
Start End Trend
200712 200712 unknown
200801 200804 UP
200805 200805 SAME
200806 200809 DOWN
200810 200810 UP
200811 200811 DOWN
200812 200812 UP
200903 200903 DOWN
200904 200905 SAME
200906 200907 UP
Any help would be really helpful; Thanks

Answer 1

Took me a few goes (and a few hours), but I think I have what you want:
DECLARE #Sales AS TABLE (mnth datetime, qty int)
INSERT INTO #Sales
SELECT '2016-01-01', 10 UNION ALL
SELECT '2016-02-01', 20 UNION ALL
SELECT '2016-03-01', 30 UNION ALL
SELECT '2016-04-01', 40 UNION ALL
SELECT '2016-05-01', 40 UNION ALL
SELECT '2016-06-01', 30 UNION ALL
SELECT '2016-07-01', 20 UNION ALL
SELECT '2016-08-01', 30 UNION ALL
SELECT '2016-09-01', 40 UNION ALL
SELECT '2016-10-01', 45 UNION ALL
SELECT '2016-11-01', 50
;WITH etape1 AS (
SELECT ROW_NUMBER() OVER(ORDER BY mnth) AS id, * FROM #Sales
)
, etape2 AS (
SELECT id, lag(mnth) OVER (ORDER BY id) AS START, mnth AS FINISH, CASE WHEN qty - LAG(qty) OVER (ORDER BY id) < 0 THEN -1 WHEN qty - LAG(qty) OVER (ORDER BY id) > 0 THEN 1 ELSE 0 END AS TREND
FROM etape1
)
, etape3 AS (
SELECT id, START, FINISH, TREND, lag(TREND) OVER (ORDER BY id) AS PrevTrend
FROM etape2
)
, etape4 AS (
SELECT id, START, FINISH, TREND, SUM(CASE WHEN TREND = PREVTREND THEN 0 ELSE 1 END) OVER (ORDER BY id ROWS UNBOUNDED PRECEDING) AS Change
FROM etape3
)
SELECT MIN(START) AS START, MAX(FINISH) AS FINISH, CASE WHEN MIN(TREND) IS NULL THEN 'Unknown' WHEN MIN(TREND) < 0 THEN 'Down' WHEN MIN(TREND) > 0 THEN 'Up' WHEN MIN(Start) is NULL THEN 'Unknown' ELSE 'Same' END AS TREND
FROM etape4
GROUP BY Change
ORDER BY START
Results are:
START FINISH TREND
NULL 2016-01-01 Unknown
2016-01-01 2016-04-01 Up
2016-04-01 2016-05-01 Same
2016-05-01 2016-07-01 Down
2016-07-01 2016-11-01 Up