Find increasing and decreasing values in a column SNOWFLAKE ,SQL - snowflake-cloud-data-platform

Question

I have a table in snowflake like below, and wanted to add new column represents if that row is increasing or decreasing based on the next row!
id
value
1
70
2
70
3
70
4
70
5
70
6
71
7
72
8
73
9
73
10
73
11
74
12
74
13
74
14
74
15
73
16
72
17
73
18
72
19
72
20
72
21
72
22
71
23
71
24
72
25
72
Expected Output:
id
value
DESIRED OUTPUT
1
70
INCREASING_ID1
2
70
INCREASING_ID2
3
70
INCREASING_ID3
4
70
INCREASING_ID4
5
70
INCREASING_ID5
6
71
INCREASING_ID6
7
72
INCREASING_ID7
8
73
INCREASING_ID8
9
73
INCREASING_ID9
10
73
INCREASING_ID10
11
74
INCREASING_ID11
12
74
INCREASING_ID12
13
74
INCREASING_ID13
14
74
INCREASING_ID14
15
73
DECREASING_ID15
16
72
DECREASING_ID16
17
73
INCREASING_ID17
18
72
DECREASING_ID18
19
72
DECREASING_ID19
20
72
DECREASING_ID20
21
72
DECREASING_ID21
22
71
DECREASING_ID22
23
71
DECREASING_ID23
24
72
INCREASING_ID24
25
72
INCREASING_ID25

Answer 1

This is a two step process. First step, find the direction of the lag: increasing, decreasing, or null if neither direction. Second step, if the current direction for a row is null, refer back to the last non-null value (null previously defined as no change in direction):
with DIRECTION as
(
select ID
,VALUE
,case
when ID = 1 then 'INCREASING'
when VALUE > lag(VALUE) over (order by id) then 'INCREASING'
when VALUE < lag(VALUE) over (order by id) then 'DECREASING'
else NULL
end as OUTPUT
from T1
)
select ID
,VALUE
,case when OUTPUT is null then
lag(OUTPUT) ignore nulls over (order by ID) || '_ID' || ID
else OUTPUT || '_ID' || ID
end as OUTPUT
from DIRECTION
;

Answer 2

So this is the same answer is Greg's, I have just written it verbosely to show the workings:
firstly a CTE for data:
with data(id, value, DESIRED_OUTPUT) as (
select * from values
(1, 70, 'INCREASING_ID1'),
(2, 70, 'INCREASING_ID2'),
(3, 70, 'INCREASING_ID3'),
(4, 70, 'INCREASING_ID4'),
(5, 70, 'INCREASING_ID5'),
(6, 71, 'INCREASING_ID6'),
(7, 72, 'INCREASING_ID7'),
(8, 73, 'INCREASING_ID8'),
(9, 73, 'INCREASING_ID9'),
(10, 73, 'INCREASING_ID10'),
(11, 74, 'INCREASING_ID11'),
(12, 74, 'INCREASING_ID12'),
(13, 74, 'INCREASING_ID13'),
(14, 74, 'INCREASING_ID14'),
(15, 73, 'DECREASING_ID15'),
(16, 72, 'DECREASING_ID16'),
(17, 73, 'INCREASING_ID17'),
(18, 72, 'DECREASING_ID18'),
(19, 72, 'DECREASING_ID19'),
(20, 72, 'DECREASING_ID20'),
(21, 72, 'DECREASING_ID21'),
(22, 71, 'DECREASING_ID22'),
(23, 71, 'DECREASING_ID23'),
(24, 72, 'INCREASING_ID24'),
(25, 72, 'INCREASING_ID25')
)
then two selects, as the two LAG's cannot co-exist at the same level:
select *
,lag(change)ignore nulls over(order by id) as lag_change
,nvl(change, lag_change) || '_ID' || id as final_answer
,DESIRED_OUTPUT = final_answer as is_same
from (
select *
,lag(value)over(order by id) as lag_v
,case
when lag_v is null then 'INCREASING'
when value<lag_v then 'DECREASING'
when value>lag_v then 'INCREASING'
end as change
from data
)
order by 1;
ID
VALUE
DESIRED_OUTPUT
LAG_V
CHANGE
LAG_CHANGE
FINAL_ANSWER
IS_SAME
1
70
INCREASING_ID1
null
INCREASING
null
INCREASING_ID1
TRUE
2
70
INCREASING_ID2
70
null
INCREASING
INCREASING_ID2
TRUE
3
70
INCREASING_ID3
70
null
INCREASING
INCREASING_ID3
TRUE
4
70
INCREASING_ID4
70
null
INCREASING
INCREASING_ID4
TRUE
5
70
INCREASING_ID5
70
null
INCREASING
INCREASING_ID5
TRUE
6
71
INCREASING_ID6
70
INCREASING
INCREASING
INCREASING_ID6
TRUE
7
72
INCREASING_ID7
71
INCREASING
INCREASING
INCREASING_ID7
TRUE
8
73
INCREASING_ID8
72
INCREASING
INCREASING
INCREASING_ID8
TRUE
9
73
INCREASING_ID9
73
null
INCREASING
INCREASING_ID9
TRUE
10
73
INCREASING_ID10
73
null
INCREASING
INCREASING_ID10
TRUE
11
74
INCREASING_ID11
73
INCREASING
INCREASING
INCREASING_ID11
TRUE
12
74
INCREASING_ID12
74
null
INCREASING
INCREASING_ID12
TRUE
13
74
INCREASING_ID13
74
null
INCREASING
INCREASING_ID13
TRUE
14
74
INCREASING_ID14
74
null
INCREASING
INCREASING_ID14
TRUE
15
73
DECREASING_ID15
74
DECREASING
INCREASING
DECREASING_ID15
TRUE
16
72
DECREASING_ID16
73
DECREASING
DECREASING
DECREASING_ID16
TRUE
17
73
INCREASING_ID17
72
INCREASING
DECREASING
INCREASING_ID17
TRUE
18
72
DECREASING_ID18
73
DECREASING
INCREASING
DECREASING_ID18
TRUE
19
72
DECREASING_ID19
72
null
DECREASING
DECREASING_ID19
TRUE
20
72
DECREASING_ID20
72
null
DECREASING
DECREASING_ID20
TRUE
21
72
DECREASING_ID21
72
null
DECREASING
DECREASING_ID21
TRUE
22
71
DECREASING_ID22
72
DECREASING
DECREASING
DECREASING_ID22
TRUE
23
71
DECREASING_ID23
71
null
DECREASING
DECREASING_ID23
TRUE
24
72
INCREASING_ID24
71
INCREASING
DECREASING
INCREASING_ID24
TRUE
25
72
INCREASING_ID25
72
null
INCREASING
INCREASING_ID25
TRUE
and thus the final SQL can be:
select id, value
,nvl(change, lag(change)ignore nulls over(order by id)) || '_ID' || id as final_answer
from (
select id, value
,lag(value)over(order by id) as lag_v
,case
when lag_v is null then 'INCREASING'
when value<lag_v then 'DECREASING'
when value>lag_v then 'INCREASING'
end as change
from data
)
order by 1;
which is the same as Greg's other than I have used the implicit NULL result in the inner CASE, and I used a NVL instead of CASE in the outer SELECT, a COALESCE could also be used here.