SQL - How to SELECT the best two months which are next to each other - sql

Question

How to select in PostgreSQL the best two months from Table.
Table:
ID Month Value
1 2019-06 100
2 2019-07 120
3 2019-08 70
4 2019-09 200
5 2019-10 100
6 2019-11 50
I would like to select ID where sum(Value) of two months which are next to each other is the highest.
In the following case the result will be:
4 2019-09
5 2019-10
where the sum of values is equal to 300.

Answer 1

You can put the data on one row using join:
select t1.*, t2.*
from t t1 join
t t2
on t2.month = t1.month + interval '1 month'
order by t1.value + t.value desc
limit 1;
Getting separate rows is trickier. You can easily get the first row using lead():
select t.*
from (select t.*, lead(value, 1, 0) over (order by month) as next_value
from t
) t
order by (value + next_value) desc
limit 1;
Getting the second month is much trickier. I am thinking that the simplest method is to unpivot the first results:
select t.*
from (select t1, t2
from t t1 join
t t2
on t2.month = t1.month + interval '1 month'
order by t1.value + t.value desc
limit 1
) cross join lateral
unnest(array[t1, t2]) t
order by t.month;

Answer 2

Here is a solution that uses solely window functions and that does not assume that month is of a date-like datatype.
This works as follows:
first rank records per increasing month with row_number() (aliased as rn) and compute the sum of the current and previous value (aliased as vals)
rank records by vals (aliased rnk)
exhibit the row number the record that has the highest vals (aliased rn_max)
finally pull out the this record and the preceeding one (ie the one that has the previous row number)
Query:
select id, month, value
from (
select t.*, first_value(rn) over(order by rnk) rn_max
from (
select t.*, rank() over(order by vals desc) rnk
from (
select
t.*,
value + lag(value, 1, 0) over (order by month) vals,
row_number() over(order by month) rn
from mytable t
) t
) t
) t
where rn in (rn_max, rn_max - 1)
order by month
Step-by-step demo on DB Fiddle:
id | month | value
-: | :------ | ----:
4 | 2019-09 | 200
5 | 2019-10 | 100