问题
This is a follow up on the other question I asked.
Quarter Segment Customer *Counter*
Q1 2018 A A1 1
Q2 2018 A A1 2
Q3 2018 A A1 3
Q4 2018 B A1 1
Q1 2019 B A1 2
Q2 2019 A A1 1
Q1 2020 A A1 *1* I want 1 not 2 here because it's not consecutive (we don't have Q3 & Q4 2019)
Q2 2020 A A1 *2* I want 2 not 3 here because it reset in Q1 2020.
Below query works if the dates are consecutive. How would I adjust the query to get what I'm looking for? I tried adding a new column that is 1 row lag, and tried to check if the quarter value in the new column is essentially the true previous 1 quarter (if yes, use the counter, if not, use "1" to restart). But it doesn't reset the following records. Then I am not sure what to do.
select quarter, customer, segment,
row_number() over (partition by customer, segment, seqnum - seqnum_cs order by right(quarter, 4), left(quarter, 2)) as counter
from (select t.*,
row_number() over (partition by customer order by right(quarter, 4), left(quarter, 2)) as seqnum,
row_number() over (partition by customer, segment order by right(quarter, 4), left(quarter, 2)) as seqnum_cs
from t
) t
order by customer, seqnum;
回答1:
You need to fix your data model! In any case, this is solved by enumerating the quarters.
select quarter, customer, segment,
row_number() over (partition by customer, segment, q_seqnum - seqnum_cs order by q_seqnum) as counter
from (select t.*,
row_number() over (partition by customer, segment order by q_seqnum) as seqnum_cs
from (select t.*,
cast(right(quarter, 4) as int) * 4 + cast(substring(quarter, 2, 1) as int) as q_seqnum
from t
) t
) t
order by customer, q_seqnum;
来源:https://stackoverflow.com/questions/59431244/sql-gaps-and-island-problem-date-not-consecutive-causing-rank-inaccurate