Wrong cardinality with date literal

Question

Hi guys,
There is table T with column D (DATE). The table consists of data from 1999 year.
NUM_ROWS = 33.7M
The column D does NOT have histogram.
There is index on the column D (and columns C1, C2).
create index ... on T (D, C1, C2)There is a SQL and plan
SQL> explain plan for
  2  select *
  3    from t
  4   where t.D > to_date('01.01.2010', 'dd.mm.rrrr')
  5  ;
Explained
SQL> @plan
Session altered
Plan hash value: 2411994532
| Id  | Operation                   | Name                  | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |                       |     1 |   278 |     5   (0)| 00:00:01 |
|   1 |  TABLE ACCESS BY INDEX ROWID| T                     |     1 |   278 |     5   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | D                     |     1 |       |     4   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - access("T"."D">TO_DATE(' 2010-01-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))
14 rows selectedNotice, cardinality = 1 (!!!) Why?
When min data is 1999 year, and max is 2010 year in the SQL above I expect get cardinality about 1/10 of NUM_ROWS.
But cardinality = 1.
This is the question.
As I know (68992.1), in this case selectivity is calculate as 1 - (High - Value / High - Low). But in my 10053 trace I see very small value 1.6733e-08.
SINGLE TABLE ACCESS PATH
  BEGIN Single Table Cardinality Estimation
  Column (#25): D(DATE)
    AvgLen: 8.00 NDV: 7765172 Nulls: 0 Density: 1.2878e-07 Min: 2451269 Max: 2453250
  Using prorated density: 1.6733e-08 of col #25 as selectivity of out-of-range value pred
  Table: T  Alias: T    
    Card: Original: 29880736  Rounded: 1  Computed: 0.50  Non Adjusted: 0.50
  END   Single Table Cardinality Estimation
  Access Path: TableScan
    Cost:  785120.33  Resp: 785120.33  Degree: 0
      Cost_io: 572319.00  Cost_cpu: 57593829444
      Resp_io: 572319.00  Resp_cpu: 57593829444
  Using prorated density: 1.6733e-08 of col #25 as selectivity of out-of-range value pred
  Using prorated density: 1.6733e-08 of col #25 as selectivity of out-of-range value pred
  Access Path: index (RangeScan)
    Index: TIDX
    resc_io: 5.00  resc_cpu: 31164
    ix_sel: 1.6733e-08  ix_sel_with_filters: 1.6733e-08
    Cost: 5.12  Resp: 5.12  Degree: 1And the second one - how can I translate HIGH_VALUE and LOW_VALUE from dba_tab_col_statistics from RAW to DATE?

Answer 1

793769 wrote:
SINGLE TABLE ACCESS PATH
BEGIN Single Table Cardinality Estimation
Column (#25): D(DATE)
AvgLen: 8.00 NDV: 7765172 Nulls: 0 Density: 1.2878e-07 Min: 2451269 Max: 2453250
Using prorated density: 1.6733e-08 of col #25 as selectivity of out-of-range value predAnd the second one - how can I translate HIGH_VALUE and LOW_VALUE from dba_tab_col_statistics from RAW to DATE?I know you've said that the question is answered - but
a) The "out of range value" is a good clue that the low/high values are not what you expect
b) For dates, the values are the Julian format of the date, and in your case we see:
SQL> select to_char(to_date('01.01.2010','dd.mm.rrrr'),'J') from dual;
TO_CHAR
2455198
1 row selected.Your high value is well below 1st Jan 2010.
The value you have can be derived from :
SQL> select to_char(to_date('2453250','J'),'dd-mon-yyyy') from dual;
TO_CHAR(TO_
01-sep-2004
1 row selected.Regards
Jonathan Lewis