Oracle Efficiently joining tables with subquery in FROM - sql

Question

Table 1:
| account_no | **other columns**...
+------------+-----------------------
| 1 |
| 2 |
| 3 |
| 4 |
Table 2:
| account_no | TX_No | Balance | History |
+------------+-------+---------+------------+
| 1 | 123 | 123 | 12.01.2011 |
| 1 | 234 | 2312 | 01.03.2011 |
| 3 | 232 | 212 | 19.02.2011 |
| 4 | 117 | 234 | 24.01.2011 |
I have multiple join query, one of the tables(Table 2) inside a query is problematic as it is a view which computes many other things, that is why each query to that table is costly. From Table 2, for each account_no in Table 1 I need the whole row with the greatest TX_NO, this is how I do it:
SELECT * FROM TABLE1 A LEFT JOIN
( SELECT
X.ACCOUNT_NO,
HISTORY,
X.BALANCE
FROM TABLE2 X INNER JOIN
(SELECT
ACCOUNT_NO,
MAX(TX_NO) AS TX_NO
FROM TABLE2
GROUP BY ACCOUNT_NO) Y ON X.ACCOUNT_NO = Y.ACCOUNT_NO) B
ON B.ACCOUNT_NO = A.ACCOUNT_NO
As I understand at first it will make the inner join for all the rows in Table2 and after that left join needed account_no's with Table1 which is what I would like to avoid.
My question: Is there a way to find the max(TX_NO) for only those accounts that are in Table1 instead of going through all? I think it will help to increase the speed of the query.

Answer 1

I think you are on the right track, but I don't think that you need to, and would not myself, nest the subqueries the way you have done. Instead, if you want to get each record from table 1 and the matching max record from table 2, you can try the following:
SELECT * FROM TABLE1 t1
LEFT JOIN
(
SELECT t.*,
ROW_NUMBER() OVER (PARTITION BY account_no ORDER BY TX_No DESC) rn
FROM TABLE2 t
) t2
ON t1.account_no = t2.account_no AND
t2.rn = 1
If you want to continue with your original approach, this is how I would do it:
SELECT *
FROM TABLE1 t1
LEFT JOIN TABLE2 t2
ON t1.account_no = t2.account_no
INNER JOIN
(
SELECT account_no, MAX(TX_No) AS max_tx_no
FROM TABLE2
GROUP BY account_no
) t3
ON t2.account_no = t3.account_no AND
t2.TX_No = t3.max_tx_no
Instead of using a window function to find the greatest record per account in TABLE2, we use a second join to a subquery instead. I would expect the window function approach to perform better than this double join approach, and once you get used to it can even easier to read.

Answer 2

If table1 is comparatiely less expensive then you could think of doing a left outer join first which would considerable decrease the resultset and from that pick the latest transaction id records alone
select <required columns> from
(
select f.<required_columns),row_number() over (partition by account_no order by tx_id desc ) as rn
from
(
a.*,b.tx_id,b.balance,b.History
from table1 a left outer join table2 b
on a.account_no=b.account_no
)f
)g where g.rn=1