How to find value in a range of following rows - SQL Teradata - sql

Question

I have a table with the following columns:
account, validity_date,validity_month,amount.
For each row i want to check if the value in field "amount' exist over the rows range of the next month. if yes, indicator=1, else 0.
account validity_date validity_month amount **required_column**
------- ------------- --------------- ------- ----------------
123 15oct2019 201910 400 0
123 20oct2019 201910 500 1
123 15nov2019 201911 1000 0
123 20nov2019 201911 500 0
123 20nov2019 201911 2000 1
123 15dec2019 201912 400
123 15dec2019 201912 2000
Can anyone help?
Thanks

Answer 1

validity_month/100*12 + validity_month MOD 100 calculates a month number (for comparing across years, Jan to previous Dec) and the inner ROW_NUMBER reduces multiple rows with the same amount per month to a single row (kind of DISTINCT):
SELECT dt.*
,CASE -- next row is from next month
WHEN Lead(nextMonth IGNORE NULLS)
Over (PARTITION BY account, amount
ORDER BY validity_date)
= (validity_month/100*12 + validity_month MOD 100) +1
THEN 1
ELSE 0
END
FROM
(
SELECT t.*
,CASE -- one row per account/month/amount
WHEN Row_Number()
Over (PARTITION BY account, amount, validity_month
ORDER BY validity_date ) = 1
THEN validity_month/100*12 + validity_month MOD 100
END AS nextMonth
FROM tab AS t
) AS dt
Edit:
The previous is for exact matching amounts, for a range match the query is probably very hard to write with OLAP-functions, but easy with a Correlated Subquery:
SELECT t.*
,CASE
WHEN
( -- check if there's a row in the next month matching the current amount +/- 10 percent
SELECT Count(*)
FROM tab AS t2
WHERE t2.account_ = t.account_
AND (t2.validity_month/100*12 + t2.validity_month MOD 100)
= ( t.validity_month/100*12 + t.validity_month MOD 100) +1
AND t2.amount BETWEEN t.amount * 0.9 AND t.amount * 1.1
) > 0
THEN 1
ELSE 0
END
FROM tab AS t
But then performance might be really bad...

Answer 2

Assuming the values are unique within a month and you have a value for each month for each account, you can simplify this to:
select t.*,
(case when lead(seqnum) over (partition by account, amount order by validity_month) = seqnum + 1
then 1 else 0
end)
from (select t.*,
dense_rank() over (partition by account order by validity_month) as seqnum
from t
) t;
Note: This puts 0 for the last month rather than NULL, but that can easily be adjusted.
You can do this without the subquery by using month arithmetic. It is not clear what the data type of validity_month is. If I assume a number:
select t.*,
(case when lead(floor(validity_month / 100) * 12 + (validity_month mod 100)
) over (partition by account, amount order by validity_month) =
(validity_month / 100) * 12 + (validity_month mod 100) - 1
then 1 else 0
end)
from t;

Answer 3

Just to add another way to do this using Standard SQL. This query will return 1 when the condition is met, 0 when it is not, and null when there isn't a next month to evaluate (as implied in your result column).
It is assumed that we're partitioning on the account field. Also includes a 10% range match on the amount field based on the comment made. Note that if you have an id field, you should include it (if two rows have the same account, validity_date, validity_month, amount there will only be one resulting row, due to DISTINCT).
Performance-wise, should be similar to the answer from #dnoeth.
SELECT DISTINCT
t1.account,
t1.validity_date,
t1.validity_month,
t1.amount,
CASE
WHEN t2.amount IS NOT NULL THEN 1
WHEN MAX(t1.validity_month) OVER (PARTITION BY t1.account) > t1.validity_month THEN 0
ELSE NULL
END AS flag
FROM `project.dataset.table` t1
LEFT JOIN `project.dataset.table` t2
ON
t2.account = t1.account AND
DATE_DIFF(
PARSE_DATE("%Y%m", CAST(t2.validity_month AS STRING)),
PARSE_DATE("%Y%m", CAST(t1.validity_month AS STRING)),
MONTH
) = 1 AND
t2.amount BETWEEN t1.amount * 0.9 AND t1.amount * 1.1;