Advance Aggregation based on multiple conditions

Question

Hi members,
I have a situation where I need to aggregate data based on multiple conditions. Relevant details of the problem is as follows.
There is a table (let's call X). It has following columns:
Transaction_Time (date)
Transaction_direction (Possible values IN or OUT)
Column_1
Column_2
Based on the columns: Transaction_direction, Column_1, Column_2, the type of the transaction will be derived. For example, if transaction_direction='IN' then transaction type is IN, if 'OUT' then transaction types are Out. Similarly if Column_1=Column_2 then transaction type is Txn_3 otherwise 4.
Based on date and transaction types the aggregation will happen.The sample output would be:
Time, Transaction type (IN, OUT, Txn_3, Txn_4), Sum of transactions
10-June-2013 00:00  IN Transactions  2500
10-June-2013 00:00  Txn_3 Transactions 3590
and so.
IN and Out transactions are easy to be derived using decode() function. However avoiding multiple UNION ALL and write a single SQL for all four conditions is tricky one.
Hope I clarified.
Neeraj

Answer 1

What version of Oracle are you using?
If you're on 11.x you can use the UNPIVOT feature as follows:
with t (Transaction_Time, Transaction_direction, Column_1, Column_2) as (
select date '2013-06-10', 'IN', 1, 1 from dual union all
select date '2013-06-10', 'IN', 2, 2 from dual union all
select date '2013-06-10', 'IN', 1, 2 from dual union all
select date '2013-06-10', 'IN', 3, 4 from dual union all
select date '2013-06-10', 'OUT', 3, 3 from dual union all
select date '2013-06-10', 'OUT', 3, 4 from dual
select * from (
select
  transaction_time
, sum(case when transaction_direction = 'IN' then 1 end)  as IN_count
, sum(case when transaction_direction = 'OUT' then 1 end) as OUT_count
, sum(case when Column_1 = Column_2 then 1 end)           as Txn_3_count
, sum(case when Column_1 != Column_2 then 1 end)          as Txn_4_count
from t
group by transaction_time
unpivot (
  txn_count for transaction_type in (
     IN_count as 'IN'
  ,  OUT_count as 'OUT'
  ,  Txn_3_count as 'Txn_3'
  ,  Txn_4_count as 'Txn_4'
order by transaction_time, transaction_type
TRANSACTION_TIME TRANSACTION_TYPE TXN_COUNT
2013-06-10       IN               4      
2013-06-10       OUT              2     
2013-06-10       Txn_3            3      
2013-06-10       Txn_4            3      
If you're okay with getting one row per date with the 4 counts you can just use the inner select above, i.e.
select
  transaction_time
, sum(case when transaction_direction = 'IN' then 1 end)  as IN_count
, sum(case when transaction_direction = 'OUT' then 1 end) as OUT_count
, sum(case when Column_1 = Column_2 then 1 end)           as Txn_3_count
, sum(case when Column_1 != Column_2 then 1 end)          as Txn_4_count
from t
group by transaction_time
order by transaction_time
TRANSACTION_TIME IN_COUNT OUT_COUNT  TXN_3_COUNT  TXN_4_COUNT
2013-06-10       4        2          3            3      
If you want to sum transaction amounts then use the same logic, except in the case statements replace 1 with the column you want to sum.
Regards,
Bob