Lag analytical function and controlling the offset

Question

Can you help me on that? Small challenge. At least I gave up in half a day.
Data
ACCOUNT_NUMBER     BATCH_ID     TRANSACTION_DATE     TRANSACTION_TYPE     TRANSACTION_NUMBER     PARENT_TRANSACTION_NUMBER
124680     ZY000489     1/11/2011     62     377     NULL
124680     ZY000489     1/11/2011     1     378     NULL
124680     ZY000489     1/11/2011     1     379     NULL
124680     ZY000489     1/11/2011     1     380     NULL
124680     ZY000489     1/11/2011     62     381     NULL
124680     ZY000489     1/11/2011     1     381     NULL
124681     ZY000490     1/11/2011     350     4000     NULL
124681     ZY000490     1/11/2011     1     4001     NULL
124681     ZY000490     1/11/2011     1     4002     NULL
I want to identify parent Transaction Number for each row in above data.
The way to identify it is My parent transaction Id is
-     All child transaction have type as 1
-     One main transaction can have multiple line items.
-     Any transaction (type) can have an related child transaction (Transaction Type as 1)
-     Each logical group of transactions have same account number, batch id, transaction date and consecutive Transaction Number (like 377, 378, 379, 380 in above example)
The data should look like below once I identified parent transaction columns:
ACCOUNT_NUMBER     BATCH_ID     TRANSACTION_DATE     TRANSACTION_TYPE     TRANSACTION_NUMBER     PARENT_TRANSACTION_NUMBER
124680     ZY000489     1/11/2011     62     377     377
124680     ZY000489     1/11/2011     1     378     377
124680     ZY000489     1/11/2011     1     379     377
124680     ZY000489     1/11/2011     1     380     377
124680     ZY000489     1/11/2011     62     381     381
124680     ZY000489     1/11/2011     1     382     381
124681     ZY000490     1/11/2011     350     4000     4000
124681     ZY000490     1/11/2011     1     4001     4000
124681     ZY000490     1/11/2011     1     4002     4000
I tried using LAG Analytical function trying to lag dynamically with offset but had difficulties dynamically expanding the offset. Its an Control Break kind of functionality that i want to achieve in single SQL.
i Know we can do it using pl/sql construct but the challenge is to do it using single sql. Please help
Please let me know if you are able to do it in single SQL.
Thanks

Answer 1

rohitgoswami wrote:
Can you help me on that? Small challenge. At least I gave up in half a day.
Data
ACCOUNT_NUMBER     BATCH_ID     TRANSACTION_DATE     TRANSACTION_TYPE     TRANSACTION_NUMBER     PARENT_TRANSACTION_NUMBER
124680     ZY000489     1/11/2011     62     377     NULL
124680     ZY000489     1/11/2011     1     378     NULL
124680     ZY000489     1/11/2011     1     379     NULL
124680     ZY000489     1/11/2011     1     380     NULL
124680     ZY000489     1/11/2011     62     381     NULL
124680     ZY000489     1/11/2011     1     381     NULL
124681     ZY000490     1/11/2011     350     4000     NULL
124681     ZY000490     1/11/2011     1     4001     NULL
124681     ZY000490     1/11/2011     1     4002     NULL
I want to identify parent Transaction Number for each row in above data.
The way to identify it is My parent transaction Id is
-     All child transaction have type as 1
-     One main transaction can have multiple line items.
-     Any transaction (type) can have an related child transaction (Transaction Type as 1)
-     Each logical group of transactions have same account number, batch id, transaction date and consecutive Transaction Number (like 377, 378, 379, 380 in above example)
The data should look like below once I identified parent transaction columns:
ACCOUNT_NUMBER     BATCH_ID     TRANSACTION_DATE     TRANSACTION_TYPE     TRANSACTION_NUMBER     PARENT_TRANSACTION_NUMBER
124680     ZY000489     1/11/2011     62     377     377
124680     ZY000489     1/11/2011     1     378     377
124680     ZY000489     1/11/2011     1     379     377
124680     ZY000489     1/11/2011     1     380     377
124680     ZY000489     1/11/2011     62     381     381
124680     ZY000489     1/11/2011     1     382     381
124681     ZY000490     1/11/2011     350     4000     4000
124681     ZY000490     1/11/2011     1     4001     4000
124681     ZY000490     1/11/2011     1     4002     4000
I tried using LAG Analytical function trying to lag dynamically with offset but had difficulties dynamically expanding the offset. Its an Control Break kind of functionality that i want to achieve in single SQL.
i Know we can do it using pl/sql construct but the challenge is to do it using single sql. Please help
Please let me know if you are able to do it in single SQL.
ThanksCan probably pretty this up ... i just went for functional code for the moment.
TUBBY_TUBBZ?with
  2     data (acc_no, batch_id, trans_date, trans_type, trans_no) as
  3  (
  4    select 124680, 'ZY000489', to_date('1/11/2011', 'mm/dd/yyyy'), 62   , 377   from dual union all
  5    select 124680, 'ZY000489', to_date('1/11/2011', 'mm/dd/yyyy'), 1    , 378   from dual union all
  6    select 124680, 'ZY000489', to_date('1/11/2011', 'mm/dd/yyyy'), 1    , 379   from dual union all
  7    select 124680, 'ZY000489', to_date('1/11/2011', 'mm/dd/yyyy'), 1    , 380   from dual union all
  8    select 124680, 'ZY000489', to_date('1/11/2011', 'mm/dd/yyyy'), 62   , 381   from dual union all
  9    select 124680, 'ZY000489', to_date('1/11/2011', 'mm/dd/yyyy'), 1    , 382   from dual union all
10    select 124681, 'ZY000490', to_date('1/11/2011', 'mm/dd/yyyy'), 350  , 4000  from dual union all
11    select 124681, 'ZY000490', to_date('1/11/2011', 'mm/dd/yyyy'), 1    , 4001  from dual union all
12    select 124681, 'ZY000490', to_date('1/11/2011', 'mm/dd/yyyy'), 1    , 4002  from dual
13  )
14  select
15    acc_no,
16    batch_id,
17    trans_date,
18    trans_type,
19    trans_no,
20    case when trans_type != 1
21    then
22      trans_no
23    else
24      lag
25      (
26        case when trans_type = 1
27        then
28           null
29        else
30           trans_no
31        end
32        ignore nulls
33      ) over (partition by acc_no, batch_id, trans_date order by trans_no asc)
34    end as parent_trans_no
35  from data;
            ACC_NO BATCH_ID                 TRANS_DATE                         TRANS_TYPE           TRANS_NO    PARENT_TRANS_NO
            124680 ZY000489                 11-JAN-2011 12 00:00                       62                377                377
            124680 ZY000489                 11-JAN-2011 12 00:00                        1                378                377
            124680 ZY000489                 11-JAN-2011 12 00:00                        1                379                377
            124680 ZY000489                 11-JAN-2011 12 00:00                        1                380                377
            124680 ZY000489                 11-JAN-2011 12 00:00                       62                381                381
            124680 ZY000489                 11-JAN-2011 12 00:00                        1                382                381
            124681 ZY000490                 11-JAN-2011 12 00:00                      350               4000               4000
            124681 ZY000490                 11-JAN-2011 12 00:00                        1               4001               4000
            124681 ZY000490                 11-JAN-2011 12 00:00                        1               4002               4000
9 rows selected.
Elapsed: 00:00:00.01
TUBBY_TUBBZ?