Oracle 10g - Using max on date column that is both partition and index

Question

I have a table that is both partitioned by and indexed on a DATE column d_dly_label .

I'm trying to understand the impact of this on max(d_dly_label) . Why does it produce different values for the below two queries:

Query 1:

select 
  max(d_dly_label) 
from 
  ECP.TSPT105
where 
  d_dly_label <= to_date('12/06/2015', 'dd/mm/yyyy');
--result: 13/JUL/14

Query 2:

select 
  max(d_dly_label) 
from 
  ECP.TSPT105
where 
  d_dly_label <= to_date('12/06/2015', 'dd/mm/yyyy')
  and d_dly_label >= to_date('12/06/2015', 'dd/mm/yyyy') - 1;
--result: 11/JUN/15

How does adding more conditions to the where clause result in a higher max(d_dly_label) ?

---

UPDATE: I've grabbed the DDL from SQL Developer and cut it down into a short example that replicates the issue.

drop table tmp_105_copy;

CREATE TABLE tmp_105_copy
(
  D_DLY_LABEL DATE NOT NULL 
)
PARTITION BY LIST (D_DLY_LABEL) 
(
  PARTITION JUL2014_13 VALUES (TO_DATE(' 2014-07-13 00:00:00', 'SYYYY-MM-DD HH24:MI:SS', 'NLS_CALENDAR=GREGORIAN')) 
  , PARTITION JUN2015_11 VALUES (TO_DATE(' 2015-06-11 00:00:00', 'SYYYY-MM-DD HH24:MI:SS', 'NLS_CALENDAR=GREGORIAN')) 
);

CREATE INDEX tmp_I1_TSPT105 ON tmp_105_copy(D_DLY_LABEL ASC) 
LOCAL 
(
  PARTITION JUL2014_13 
, PARTITION JUN2015_11 
);

insert into tmp_105_copy values (to_date('13jul2014','ddmonyyyy'));
insert into tmp_105_copy values (to_date('11jun2015','ddmonyyyy'));

commit;

select 
  max(d_dly_label) 
from 
  tmp_105_copy
where 
  d_dly_label <= to_date('12/06/2015', 'dd/mm/yyyy');
-- 13/JUL/14

analyze table tmp_105_copy estimate statistics;  

select 
  max(d_dly_label) 
from 
  tmp_105_copy
where 
  d_dly_label <= to_date('12/06/2015', 'dd/mm/yyyy');
-- still 13/JUL/14

select 
  max(d_dly_label) 
from 
  tmp_105_copy
where 
  d_dly_label <= to_date('12/06/2015', 'dd/mm/yyyy')
  and d_dly_label >= to_date('12/06/2015', 'dd/mm/yyyy') - 1;
--11/JUN/15

select 
  max(to_date(d_dly_label))
from 
  tmp_105_copy
where 
  d_dly_label <= to_date('12/06/2015', 'dd/mm/yyyy');
-- also 11/JUN/15

Answer 1

Maybe this is an indexed field in a table? If statistics are wrong, such incorrect results can appear. You can try to update the statistics of table and check your query again.