Identify repeating raws (loans) in sql - sql

Question

I have a table like this:
Client Branch Amount Date
1 2 1500 1.1.14
1 2 1400 3.1.14
1 3 1500 1.1.14
1 4 300 7.1.14
1 5 1500 1.1.14
------------------------------
2 2 300 1.1.14
2 2 300 1.1.14
2 5 300 1.1.14
2 3 400 4.1.14
------------------------------
3 2 300 1.1.14
3 2 300 1.1.14
3 5 300 1.1.14
3 5 300 1.1.14
3 3 400 4.1.14
4 2 300 1.1.14
4 2 300 1.1.14
4 5 300 1.1.14
4 5 300 1.1.14
4 5 300 1.1.14
And my desired output should be like this:
Client Branch Amount Date Ind Loan_Distinct_Num
1 2 1500 1.1.14 0 1
1 2 1400 3.1.14 0 2
1 3 1500 1.1.14 1 1
1 4 300 7.1.14 0 3
1 5 1500 1.1.14 1 1
-------------------------------------------------
2 2 300 1.1.14 0 1
2 2 300 1.1.14 0 2
2 5 300 1.1.14 1 2
2 3 400 4.1.14 0 3
--------------------------------------------------
3 2 300 1.1.14 0 1
3 2 300 1.1.14 0 2
3 5 300 1.1.14 1 1
3 5 300 1.1.14 1 2
3 3 400 4.1.14 0 3
------------------------------------------------
4 2 300 1.1.14 0 1
4 2 300 1.1.14 0 2
4 5 300 1.1.14 1 1
4 5 300 1.1.14 1 2
4 5 300 1.1.14 0 3
So what am I trying to do? (Comment: Those records are just a sample data)
Well, those are the rules:
A client has moved from one branch to another in the same bank. The problem is that the branches were writing the data for him several times. I want to identify loans that are repeating. Two steps needed:
Step1 :
Assumption : Same_Amount + Same_Date + Diffrent Date ---> Ind = 1 on the records that are after the first raw.
How Ind field works?
For example:
In the partition for client = 1 the amount 1500 is reapting 3 times for same date and diffrent branches but only the two last records for this details will get a "1" value for Ind, the first will get Ind = 0 because it is not a repeting loan, it is the first time that a record with this amount and date appears in the data.
In case like client = 2 there were two records for branch = 2 and only one for branch = 5 so in this case I will assum that the last record for the branch = 2 was being repeted.
In case like client = 3 there were two records in branch = 2 and two records for branch = 5 so in this case i will assume that both of the loans from branch 2 was repeted.
In Case client = 4, well it will be the same like client 3 but there was another record but I will consider it a new one because I have no extra past loan to communicate with her.
Step 2: I want to create my own distinct loan number for each client
Any help on how to approach for solving this nor simple problem?
comment : sql-server 2008.

Answer 1

First - set up your data into a table. I've added an identity column ID, so we have something to order by - you specified in a comment that your data was in a specific order.
declare #data table (ID int identity(1,1), Client int, Branch int, Amount int, [Date] date);
insert into #data values
(1,2, 1500,'2014-01-01'),
(1,2, 1400,'2014-03-01'),
(1,3, 1500,'2014-01-01'),
(1,4, 300,'2014-07-01'),
(1,5, 1500,'2014-01-01'),
(2,2, 300,'2014-01-01'),
(2,2, 300,'2014-01-01'),
(2,5, 300,'2014-01-01'),
(2,3, 400,'2014-04-01'),
(3,2, 300,'2014-01-01'),
(3,2, 300,'2014-01-01'),
(3,5, 300,'2014-01-01'),
(3,5, 300,'2014-01-01'),
(3,3, 400,'2014-04-01'),
(4,2, 300,'2014-01-01'),
(4,2, 300,'2014-01-01'),
(4,5, 300,'2014-01-01'),
(4,5, 300,'2014-01-01'),
(4,5, 300,'2014-01-01');
Here is where we do the query:
--In the first cte, we take all the data, and partition it up into individual loans (partition by Client, Amount, Date).
with cte1 as (
select *, ROW_NUMBER() over (partition by Client, Amount, Date order by ID) as rowno from #data
), cte2 as (
--in this cte, we get a list of distinct loans. We will use another rownumber in a bit to find our Loan_Distinct_Num
select distinct Client, Amount, [Date] from #data
)
select cte1.Client, cte1.Branch, cte1.Amount, cte1.[Date]
-- If rowno = 1, it's the first instance of that combination
, case when rowno = 1 then 0 else 1 end as ind
, b.Loan_Distinct_Num
from cte1
left join (select cte2.*, ROW_NUMBER() over (partition by Client order by [Date]) as Loan_Distinct_Num
-- This is where our distinct loan number comes from
from cte2
) as b
on b.Client = cte1.Client and b.Amount = cte1.Amount and b.[Date] = cte1.[Date]
order by ID

Answer 2

Here's an answer if ind should only be 1 if exists a previous record with a different branch # (see row 7). Also, uses dense_rank to group loans by amount/date in loan_distinct_num. The logic seems more complex for that column - if this is a one-time fix, I might use a cursor to loop through the table and apply some more complex logic to populate that one column, rather than try to calculate it in a query.
-- sample data
declare #data table (ID int identity(1,1), Client int, Branch int, Amount int, [Date] date);
insert into #data values
(1,2, 1500,'2014-01-01'),
(1,2, 1400,'2014-03-01'),
(1,3, 1500,'2014-01-01'),
(1,4, 300,'2014-07-01'),
(1,5, 1500,'2014-01-01'),
(2,2, 300,'2014-01-01'),
(2,2, 300,'2014-01-01'),
(2,5, 300,'2014-01-01'),
(2,3, 400,'2014-04-01'),
(3,2, 300,'2014-01-01'),
(3,2, 300,'2014-01-01'),
(3,5, 300,'2014-01-01'),
(3,5, 300,'2014-01-01'),
(3,3, 400,'2014-04-01'),
(4,2, 300,'2014-01-01'),
(4,2, 300,'2014-01-01'),
(4,5, 300,'2014-01-01'),
(4,5, 300,'2014-01-01'),
(4,5, 300,'2014-01-01');
-- query
select client, branch, amount, date,
case when exists (select * from #data t2 where client = tbl.client and branch <> tbl.branch and amount = tbl.amount and date = tbl.date and id < tbl.id) then 1 else 0 end as ind,
DENSE_RANK() over (partition by client order by date, amount asc) as loan_disinct_num
from #data tbl
order by id;