Sql query to get the given output

Question

Hi Friends
My table data is given below
user transaction_date transaction_type
1 01-Aug-2011 a
1 02-Aug-2011 c
1 03-Aug-2011 a
1 03-Aug-2011 b
1 04-Aug-2011 a
1 04-Aug-2011 b
2 03-Aug-2011 a
2 03-Aug-2011 b
2 04-Aug-2011 c
2 04-Aug-2011 b
2 05-Aug-2011 a
2 05-Aug-2011 b
2 07-Aug-2011 a
2 07-Aug-2011 b
I want the count for each user as, how many times he did a transaction type 'b' immediately after transaction type a ?
Like the output for above data should be
user count
1 2
2 3
Thanks in advance :)

Answer 1

Assuming there was information in the transaction date to indicate the sequence of the transactions then it would be something like...
SQL> ed
Wrote file afiedt.buf
  1  with t as (select 1 as usr, to_date('01-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS') as transaction_date, 'a' as transaction_type from dual union all
  2             select 1, to_date('02-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'c' from dual union all
  3             select 1, to_date('03-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'a' from dual union all
  4             select 1, to_date('03-Aug-2011 00:00:01','DD-MON-YYYY HH24:MI:SS'), 'b' from dual union all
  5             select 1, to_date('04-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'a' from dual union all
  6             select 1, to_date('04-Aug-2011 00:00:01','DD-MON-YYYY HH24:MI:SS'), 'b' from dual union all
  7             select 2, to_date('03-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'a' from dual union all
  8             select 2, to_date('03-Aug-2011 00:00:01','DD-MON-YYYY HH24:MI:SS'), 'b' from dual union all
  9             select 2, to_date('04-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'c' from dual union all
10             select 2, to_date('04-Aug-2011 00:00:01','DD-MON-YYYY HH24:MI:SS'), 'b' from dual union all
11             select 2, to_date('05-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'a' from dual union all
12             select 2, to_date('05-Aug-2011 00:00:01','DD-MON-YYYY HH24:MI:SS'), 'b' from dual union all
13             select 2, to_date('07-Aug-2011 00:00:00','DD-MON-YYYY HH24:MI:SS'), 'a' from dual union all
14             select 2, to_date('07-Aug-2011 00:00:01','DD-MON-YYYY HH24:MI:SS'), 'b' from dual
15            )
16  --
17  -- end of test data
18  --
19  select usr, sum(sum_ab) as sum_ab
20  from (
21        select usr
22              ,case when transaction_type = 'b'
23                     and lag(transaction_type) over (partition by usr order by transaction_date) = 'a'
24               then 1
25               else 0
26               end as sum_ab
27        from t
28       )
29* group by usr
SQL> /
       USR     SUM_AB
         1          2
         2          3which is essentially the same as Etbin posted, but without the assumption that the ordering is by transaction_type.