Using an index to access the last n values of a certain combination

Question

I have a large table of the following form:
Name Typ
ID NUMBER
M_DATE DATE
M_LOC NUMBER
M_ARTICLE VARCHAR2(30)
M_TYPE VARCHAR2(30)
M_MEASURE VARCHAR2(50)
VALUE NUMBER
In this table there is only a small number of different locations, articles, types and maesures but a large number of m_date values.
I want to access the last 10 values for a certain combination of locations, articles, types and measures.
I have an index
create index i_res_2 on results (m_loc, m_article, m_type, m_measure, m_date);
In my opinion it should be possible to access the last 10 values directly by this index, but
when I do my select in the following way it seems to me that first all datasets with the same
locations, articles, types and maesures are selected, then sorted by date and then the last 10
dates are selected. As those datasets are spread over the whole table it needs a higher number
of physical reads. Is there a way to have an index on m_date and an other form of this query
that selects the requested values with less phyical reads? (Database version is 10.2)
SQL> select
2 id,
3 m_date,
4 value
5 from (
6 select
7 id,
8 m_date,
9 value,
10 rank() over (order by m_date desc) r
11 from results
12 where m_loc = '14001'
13 and m_article = '11211-00-00'
14 and m_type = '0002'
15 and m_measure = '032')
16 where r <= 10
17 order by m_date asc;
ID M_DATE Value
3931958 05.03.10 69.00
3931960 05.03.10 69.00
3905712 10.03.10 68.60
3999535 10.03.10 69.70
3985125 11.03.10 69.40
3957851 11.03.10 69.60
3949799 21.03.10 68.70
4017369 21.03.10 69.00
3951981 22.03.10 68.80
3983554 22.03.10 68.80
Abgelaufen: 00:00:03.00
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
| 0 | SELECT STATEMENT | | 1 | 48 | 6 (34)| 00:00:01 |
| 1 | SORT ORDER BY | | 1 | 48 | 6 (34)| 00:00:01 |
|* 2 | VIEW | | 1 | 48 | 5 (20)| 00:00:01 |
|* 3 | WINDOW SORT PUSHED RANK | | 1 | 39 | 5 (20)| 00:00:01 |
| 4 | TABLE ACCESS BY INDEX ROWID| RESULTS | 1 | 39 | 4 (0)| 00:00:01 |
|* 5 | INDEX RANGE SCAN | I_RES_2 | 1 | | 3 (0)| 00:00:01 |
Predicate Information (identified by operation id):
2 - filter("R"<=10)
3 - filter(RANK() OVER ( ORDER BY INTERNAL_FUNCTION("M_DATE") DESC )<=10)
5 - access("M_LOC"=14001 AND "M_ARTICLE"='11211-00-00' AND "M_TYPE"='0002' AND
"M_MEASURE"='032')
Statistiken
1 recursive calls
0 db block gets
547 consistent gets
474 physical reads
0 redo size
766 bytes sent via SQL*Net to client
380 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
10 rows processed

Answer 1

I found a solution, which is much better in terms of physical reads and consistent gets even if the optimizer calculates higher costs. But I still think, that there should be an easier way to do this:
SQL> select
2 id, m_date, value from results
3 where
4 (m_loc, m_article, m_type, m_measure, m_date)
5 in
6 (select m_loc, m_article, m_type, m_measure, m_date
7 from (
8 select /*+ first_rows(10) */ m_loc, m_article, m_type, m_measure, m_date
9 from results
10 where m_loc = '14001'
11 and m_article = '11370-00-00'
12 and m_type = '0002'
13 and m_measure = '032'
14 order by m_loc,m_article,m_type,m_measure,m_date desc
15 ) where rownum < 10);
ID M_DATE VALUE
2495995 01.02.09 70,4
2457657 19.01.09 66,9
2556262 13.02.09 67,4
2496232 01.02.09 67,6
2465051 20.01.09 67
2454994 19.01.09 67,4
2545951 13.02.09 67,4
2502216 01.02.09 67,5
2497533 01.02.09 67,2
9 Zeilen ausgewählt.
Ausführungsplan
Plan hash value: 3924867000
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
| 0 | SELECT STATEMENT | | 1 | 122 | 8 (25)| 00:00:01 |
| 1 | TABLE ACCESS BY INDEX ROWID| RESULTS | 1 | 39 | 3 (0)| 00:00:01 |
| 2 | NESTED LOOPS | | 1 | 122 | 8 (25)| 00:00:01 |
| 3 | VIEW | VW_NSO_1 | 1 | 83 | 4 (25)| 00:00:01 |
| 4 | HASH UNIQUE | | 1 | 31 | | |
|* 5 | COUNT STOPKEY | | | | | |
| 6 | VIEW | | 1 | 31 | 4 (25)| 00:00:01 |
|* 7 | SORT ORDER BY STOPKEY| | 1 | 31 | 4 (25)| 00:00:01 |
|* 8 | INDEX RANGE SCAN | I_RES_2 | 1 | 31 | 3 (0)| 00:00:01 |
|* 9 | INDEX RANGE SCAN | I_RES_2 | 1 | | 2 (0)| 00:00:01 |
Predicate Information (identified by operation id):
5 - filter(ROWNUM<10)
7 - filter(ROWNUM<10)
8 - access("M_LOC"=14001 AND "M_ARTICLE"='11370-00-00' AND "M_TYPE"='0002'
AND "M_MEASURE"='032')
9 - access("M_LOC"="$nso_col_1" AND "M_ARTICLE"="$nso_col_2" AND
"M_TYPE"="$nso_col_3" AND "M_MEASURE"="$nso_col_4" AND "M_DATE"="$nso_col_5")
Statistiken
1 recursive calls
0 db block gets
36 consistent gets
14 physical reads
0 redo size
748 bytes sent via SQL*Net to client
381 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
9 rows processed