As per my understanding Overlap of index is as follows:-
CREATE INDEX idx1 ON TabA ( Col1, Col2, Col3 );
CREATE INDEX idx2 ON TabA ( Col1, Col2 );
CREATE INDEX idx3 ON TabA ( Col1 DESC, Col2 DESC );
In the table structure above, the index idx1 is a superset (overlap) of the index idx2 and therefore redundant.So index1 is overlap of Index2.
Droping of such overlap indexes can improve performance.
But How can we find all the overlap index in entire schema of oracle 11gR2?
Can anybody please guide me?Thanks in advance.
You may query user_ind_columns ordering and concatenating columns by column_position and then do a self-join to do a comparison.
WITH indx AS ( SELECT table_name,index_name,LISTAGG(column_name,',') WITHIN GROUP(
ORDER BY column_position
) AS cols
FROM user_ind_columns a
GROUP BY table_name,index_name
)
SELECT a.table_name,a.index_name AS sup_index,b.index_name sub_indx,
a.cols as super_set,b.cols AS sub_set
FROM indx a
JOIN indx b ON a.table_name = b.table_name
AND a.index_name != b.index_name
where ','|| a.cols||',' like ',%'|| b.cols|| '%,';
Result
TABLE_NAME SUP_INDEX SUB_INDX SUPER_SET SUB_SET
TABA IDX1 IDX2 COL1,COL2,COL3 COL1,COL2
JOB_HISTORY JHIST_EMP_ID_ST_DATE_PK JHIST_EMPLOYEE_IX EMPLOYEE_ID,START_DATE EMPLOYEE_ID
2 rows selected.
Note: Oracle Database treats descending indexes as if they were function-based indexes and hence the column names can't be compared directly through this method.
Also note that this statement by you is disputable.
Droping of such overlap indexes (ie Index2) can improve performance.
Related
How can I achieve this to Select to one row only dynamically since
the objective is to get the uniqueness even on multiple columns
select distinct
coalesce(least(ColA, ColB),cola,colb) A1, greatest(ColA, ColB) B1
from T
The best solution is to use UNION
select colA from your_table
union
select colB from your_table;
Update:
If you want to find the duplicate then use the EXISTS as follows:
SELECT COLA, COLB FROM YOUR_TABLE T1
WHERE EXISTS (SELECT 1 FROM YOUR_tABLE T2
WHERE T2.COLA = T1.COLB OR T2.COLB = T1.COLA)
If I correctly understand words: objective is to get the uniqueness even on multiple columns, number of columns may vary, table can contain 2, 3 or more columns.
In this case you have several options, for example you can unpivot values, sort, pivot and take unique values. The exact code depends on Oracle version.
Second option is listagg(), but it has limited length and you should use separators not appearing in values.
Another option is to compare data as collections. Here I used dbms_debug_vc2coll which is simple table of varchars. Multiset except does main job:
with t as (select rownum rn, col1, col2, col3,
sys.dbms_debug_vc2coll(col1, col2, col3) as coll
from test )
select col1, col2, col3 from t a
where not exists (
select 1 from t b where b.rn < a.rn and a.coll multiset except b.coll is empty )
dbfiddle with 3-column table, nulls and different test cases
I have two tables. if the data in table1 is more than a predefined limit (say 2), i need to copy the remaining contents of table1 to table2 and delete those same contents from table1.
I used the below query to insert the excess data from table1 to table2.
insert into table2
SELECT * FROM table1 WHERE ROWNUM < ((select count(*) from table1)-2);
Now i need the delete query to delete the above contents from table1.
Thanks in advance.
A straightforward approach would be an interim storage in a temporary table. Its content can be used to determine the data to be deleted from table1 as well as the source to feed table 2.
Assume (slightly abusing notation) to be the PK column (or that of any candidate key) of table1 - usually there'll be some key that comprises only 1 column.
create global temporary table t_interim as
( SELECT <pk> pkc FROM table1 WHERE ROWNUM < ((select count(*) from table1)-2 )
;
insert into table2
select * from table1 where <pk> IN (
select pkc from t_interim
);
delete from table1 where <pk> IN (
select pkc from t_interim
);
Alternative
If any key of table1 spans more than 1 column, use an EXISTS clause instead as follows ( denoting the i-th component of a candidate key in table1):
create global temporary table t_interim as
( SELECT <ck_1> ck1, <ck_2> ck2, ..., <ck_n> ckn FROM table1 WHERE ROWNUM < ((select count(*) from table1)-2 )
;
insert into table2
select * from table1 t
where exists (
select 1
from t_interim t_i
where t.ck_1 = t_i.ck1
and t.ck_2 = t_i.ck2
...
and t.ck_n = t_i.ckn
)
;
delete from table1 t where
where exists (
select 1
from t_interim t_i
where t.ck_1 = t_i.ck1
and t.ck_2 = t_i.ck2
...
and t.ck_n = t_i.ckn
)
;
(Technically you could try to adjust the first scheme by synthesizing a key from the components of any CK, eg. by concatenating. You run the risk of introducing ambiguities ( (a bc, ab c) -> (abc, abc) ) or run into implementation limits ( max. varchar length ) using the first method)
Note
In case the table doesn't have a PK, you can apply the technique using any candidate key of table1. There will always be one, in the extreme case it's the set of all columns.
This situation may be the right time to improve the db design and add a (synthetic) pk column to table1 ( and any other tables in the system that lack it).
I have an interval partitioned table. I'd like to create a local index unusable by default to make all inserts work fast. I want manually turn indexes usable after insert. Currently every insert which creates new partition makes correspondend local index usable. How can I prevent it?
UPDATE: I'm working with Oracle 11.2
create table T (
part integer
,nm number
)
partition by range (part)
interval (1)
(partition empty values less than (1));
create index t_loc_idx on T (part, nm) local unusable
insert into T
select 1, rownum from user_tables
insert into T
select 2, rownum from user_tables
select index_name, partition_name, status from user_ind_partitions
where index_name = 'T_LOC_IDX'
What we can observe:
INDEX_NAME PARTITION_NAME STATUS
T_LOC_IDX EMPTY UNUSABLE
T_LOC_IDX SYS_P8744 USABLE
T_LOC_IDX SYS_P8745 USABLE
Keep interval partition local indexes unusable by creating a partial index and setting indexing off:
create table T (
part integer
,nm number
)
partition by range (part)
interval (1)
(partition empty values less than (1))
INDEXING OFF /* <-- New */;
create index t_loc_idx on T (part, nm) INDEXING PARTIAL /* <-- New */ local unusable ;
insert into T
select 1, rownum from user_tables where rownum <= 10;
insert into T
select 2, rownum from user_tables where rownum <= 10;
select index_name, partition_name, status from user_ind_partitions
where index_name = 'T_LOC_IDX';
New results:
INDEX_NAME PARTITION_NAME STATUS
========== ============== ======
T_LOC_IDX EMPTY UNUSABLE
T_LOC_IDX SYS_P643 UNUSABLE
T_LOC_IDX SYS_P644 UNUSABLE
I'm currently migrating data from legacy system to the current system.
I have this INSERT statement inside a stored procedure.
INSERT INTO TABLE_1
(PRIMARY_ID, SEQUENCE_ID, DESCR)
SELECT LEGACY_ID PRIMARY_ID
, (SELECT COUNT(*) + 1
FROM TABLE_1 T1
WHERE T1.PRIMARY_ID = L1.LEGACY_ID) SEQUENCE_ID
, L1.DESCR
FROM LEGACY_TABLE L1;
However, whenever I have multiple values of LEGACY_ID from LEGACY_TABLE, the query for the SEQUENCE_ID doesn't increment.
Why is this so? I can't seem to find any documentation on how the INSERT INTO SELECT statement works behind the scenes. I am guessing that it selects all the values from the table you are selecting and then inserts them simultaneously after, that's why it doesn't increment the COUNT(*) value?
What other workarounds can I do? I cannot create a SEQUENCE because the SEQUENCE_ID must be based on the number of PRIMARY_ID that are present. They are both primary ids.
Thanks in advance.
Yes, The SELECT will be executed FIRST and only then the INSERT happens.
A Simple PL/SQL block below, will be a simpler approach, though not efficient.
DECLARE
V_SEQUENCE_ID NUMBER;
V_COMMIT_LIMIT:= 20000;
V_ITEM_COUNT := 0;
BEGIN
FOR REC IN (SELECT LEGACY_ID,DESCR FROM LEGACY_TABLE)
LOOP
V_SEQUENCE_ID:= 0;
SELECT COUNT(*)+1 INTO V_SEQUENCE_ID FROM TABLE_1 T1
WHERE T1.PRIMARY_ID = REC.LEGACY_ID
INSERT INTO TABLE_1
(PRIMARY_ID, SEQUENCE_ID, DESCR)
VALUES
(REC.LEGACY_ID,V_SEQUENCE_ID,REC.DESCR);
V_ITEM_COUNT := V_ITEM_COUNT + 1;
IF(V_ITEM_COUNT >= V_COMMIT_LIMIT)
THEN
COMMIT;
V_ITEM_COUNT := 0;
END IF;
END LOOP;
COMMIT;
END;
/
EDIT: Using CTE:
WITH TABLE_1_PRIM_CT(PRIMARY_ID, SEQUENCE_ID) AS
(
SELECT L1.LEGACY_ID,COUNT(*)
FROM LEGACY_TABLE L1
LEFT OUTER JOIN TABLE_1 T1
ON(L1.LEGACY_ID = T1.PRIMARY_ID)
GROUP BY L1.LEGACY_ID
)
INSERT INTO TABLE_1
(SELECT L1.LEGACY_ID,
CTE.SEQUENCE_ID+ (ROW_NUMBER() OVER (PARTITION BY L1.LEGACY_ID ORDER BY null)),
L1.DESCR
FROM TABLE_1_PRIM_CT CTE, LEGACY_TABLE L1
WHERE L1.LEGACY_ID = CTE.PRIMARY_ID);
PS: With your Millions of Rows, this is going to create a temp table
of same size, during execution. Do Explain Plan before actual execution.
I have a table ( table1 ) that represents item grouping and another table ( table2 ) that represents the items themselves.
table1.id is foreign key to table2 and in every record of table1 I also collect information like the total number of records in table2 associated with that particular record and the sum of various fields so that I can show the grouping and a summary of what's in it without having to query table2.
Usually items in table2 are added/removed one at a time, so I update table1 to reflect the changes in table2.
A new requirement arose, choosen items in a group must be moved to a new group. I thought of it as a 3 step operation:
create a new group in table1
update choosen records in table2 to point to the newly created rec in table1
the third step would be to subtract to the group the number of records / the sum of those other fields I need do show and add them to the new group, data that I can find simply querying table2 for items associated with the new group.
I came up with the following statement that works.
update table1 t1 set
countitems = (
case t1.id
when 1 then t1.countitems - ( select count( t2.id ) from table2 t2 where t2.id = 2 )
when 2 then ( select count( t2.id ) from table2 t2 where t2.id = 2 )
end
),
sumitems = (
case t1.id
when 1 then t1.sumitems - ( select sum( t2.num ) from table2 t2 where t2.id = 2 )
when 2 then ( select sum( t2.num ) from table2 t2 where t2.id = 2 )
end
)
where t1.id in( 1, 2 );
is there a way to rewrite the statement without having to repeat the subquery every time?
thanks
Piero
You can use a cursor and a bulk collect update statement on the rowid. That way you can simply write the join query with the desired result and update the table with those values. I always use this function and make slight adjustments each time.
declare
cursor cur_cur
IS
select ti.rowid row_id
, count(t2.id) countitems
, sum(t2.num) numitems
from table t1
join table t2 on t1.id = t2.t1_id
order by row_id
;
type type_rowid_array is table of rowid index by binary_integer;
type type_countitems_array is table of table1.countitems%type;
type type_numitems_array is table of table1.numitems%type;
arr_rowid type_rowid_array;
arr_countitems type_countitems_array;
arr_numitems type_numitems_array;
v_commit_size number := 10000;
begin
open cur_cur;
loop
fetch cur_cur bulk collect into arr_rowid, arr_countitems, arr_numitems limit v_commit_size;
forall i in arr_rowid.first .. arr_rowid.last
update table1 tab
SET tab.countitems = arr_countitems(i)
, tab.numitems = arr_numitems(i)
where tab.rowid = arr_rowid(i)
;
commit;
exit when cur_cur%notfound;
end loop;
close cur_cur;
commit;
exception
when others
then rollback;
raise_application_error(-20000, 'ERROR updating table1(countitems,numitems) - '||sqlerrm);
end;