Ranking over consecutive dates

Question

Hi guys,
I have been struggling with this for some days already and I have tried to implement different solutions I was able to get from searching the internet, and applying some approeaches learned from this forum on similar questions, but maybe if I put que question here someone knows exactly the best way to handle it:
So I have a table like this:
with payments as
select '1' as ID, '20130331' as DateR, 'Not_paid' as Status from dual
union
select '1' as ID, '20130430' as DateR, 'Paid' as Status from dual
union
select '1' as ID, '20130531' as DateR, 'Not_paid' as Status from dual
union
select '2' as ID, '20130331' as DateR, 'Not_paid' as Status from dual
union
select '2' as ID, '20130430' as DateR, 'Not_paid' as Status from dual
union
select '3' as ID, '20130331' as DateR, 'Paid' as Status from dual
union
select '3' as ID, '20130430' as DateR, 'Paid' as Status from dual
union
select '3' as ID, '20130531' as DateR, 'Paid' as Status from dual
And I am trying to get the amount of the current consecutive non-payments from a user. Now, it starts to count everytime the previous payments was Paid or it´s the first payment.
I have tried to get through dense rank, giving me this as output:
select ID, dater, status, dense_rank() over (partition by ID, status order by dater asc) rnk from payments
But I need to get something like this:
ID DATER STATUS RNK
1   20130331  Not_paid  1
1   20130430  Paid  1
1   20130531  Not_Paid  1
2   20130331  Not_paid  1
2   20130430  Not_paid  2
3   20130331  Paid  1
3   20130430  Paid  2
3   20130531  Paid  3
Such that if I want to get the max(rank) to check how many non-payments a user currently has I get that ID has 1 unpayment, ID 2 two consecutive unpayments, and ID 3 has 0 unpayments. This is because on the fourth consecutive non-payment I have to consider the user as churned.

Answer 1

Hi,
Here's one way:
WITH got_grp_num AS
    SELECT  p.*
    ,      ROW_NUMBER () OVER ( PARTITION BY  id
                                ORDER BY      dater
         - ROW_NUMBER () OVER ( PARTITION BY  id, status
                                ORDER BY      dater
                              )   AS grp_num
    FROM    payments  p
SELECT    id, dater, status
,         ROW_NUMBER () OVER ( PARTITION BY  id, status, grp_num
                               ORDER BY      dater
                             )   AS rnk
FROM      got_grp_num
ORDER BY  id, dater
For an explanation of the Fixed Difference technique used here, see
https://forums.oracle.com/forums/thread.jspa?threadID=2302781 and/or
https://forums.oracle.com/forums/thread.jspa?threadID=2303591
By the way, storing date information in a sting column is simply asking for trouble.  Use a DATE column instead.