I am trying to use Bulk all and Forall in Oracle database:
Original code from Procedure is as below:
IF NVL(v_mc,0) != 0 THEN
FOR rec IN
(SELECT a.testid,
SUM(pct * NVL(cap,0))/v_mc lead1
BULK COLLECT INTO testids1, testids2
FROM testtable a
WHERE a.id = n_id
AND a.type =n_type
GROUP BY a.testid;
)
LOOP
UPDATE testtable
SET LEAD1 =ROUND(testids2(i),2)
WHERE tid = n_id
AND type = n_type
AND testid =testids1(i);
END LOOP;
END IF;
So In select statement , I am using Sum function and also using aliasing here .
Code , I have written which use Bulk collect and Forall is as follows:
PROCEDURE test
IS
TYPE test1Tab IS TABLE OF sh_rpt_temp_peer_wip.test1%TYPE;
TYPE test2Tab IS TABLE OF testtable.lead1%TYPE;
testids1 testidTab; --Error 1 and Error 2
testids2 LeadTab;
BEGIN
IF NVL(v_mc,0) != 0 THEN
SELECT testid,
SUM(pct * NVL(cap,0))/v_mc lead1
BULK COLLECT INTO testids1, testids2
FROM testtable a --Error 3
WHERE a.id = n_id
AND a.type =n_type
GROUP BY a.testid ORDER BY a.testid;
FORALL i IN testids1.FIRST..testids1.LAST
UPDATE testtable
SET LEAD1 =ROUND(testids2(i),2)
WHERE tid = n_id --Error 3
AND type = n_type
AND testid =testids1(i);
END IF;
END;
But while I am compiling procedure , I am getting multiple errors. I am very new to PL/SQL. Please let me know if I can retrieve calculated value as a Column in Bulk Collect?
I am getting below errors in procedure:
Error 1) PL/SQL: Item ignored
Error 2) component 'LEAD' must be declared
Error 3) expression is of wrong type
Please let me know what is wrong here
Thanks
As I identified that the collection type that you are referring is not in scope in the procedure, might be you have declared globally. I modified your code get a try it once, hopefully, it works for you.
PROCEDURE test
IS
TYPE test1Tab IS TABLE OF testtable.testid%TYPE;
TYPE test2Tab IS TABLE OF number;
testids1 test1Tab; //Error 1 and Error 2
testids2 test2Tab;
BEGIN
IF NVL(v_mc,0) != 0 THEN
SELECT testid,
SUM(pct * NVL(cap,0))/v_mc lead
BULK COLLECT INTO testids1, testids2
FROM testtable a //Error 3
WHERE a.id = n_id
AND a.type =n_type
GROUP BY a.testid ORDER BY a.testid;
FORALL i IN testids1.FIRST..testids1.LAST
UPDATE testtable
SET LEAD = ROUND(testids2(i),2)
WHERE tid = n_id //Error 3
AND type = n_type
AND testid = testids1(i);
END IF;
END;
Related
I need to check a condition. i.e:
if (condition)> 0 then
update table
else do not update
end if
Do I need to store the result into a variable using select into?
e.g:
declare valucount integer
begin
select count(column) into valuecount from table
end
if valuecount > o then
update table
else do
not update
You cannot directly use a SQL statement in a PL/SQL expression:
SQL> begin
2 if (select count(*) from dual) >= 1 then
3 null;
4 end if;
5 end;
6 /
if (select count(*) from dual) >= 1 then
*
ERROR at line 2:
ORA-06550: line 2, column 6:
PLS-00103: Encountered the symbol "SELECT" when expecting one of the following:
...
...
You must use a variable instead:
SQL> set serveroutput on
SQL>
SQL> declare
2 v_count number;
3 begin
4 select count(*) into v_count from dual;
5
6 if v_count >= 1 then
7 dbms_output.put_line('Pass');
8 end if;
9 end;
10 /
Pass
PL/SQL procedure successfully completed.
Of course, you may be able to do the whole thing in SQL:
update my_table
set x = y
where (select count(*) from other_table) >= 1;
It's difficult to prove that something is not possible. Other than the simple test case above, you can look at the syntax diagram for the IF statement; you won't see a SELECT statement in any of the branches.
Edit:
The oracle tag was not on the question when this answer was offered, and apparently it doesn't work with oracle, but it does work with at least postgres and mysql
No, just use the value directly:
begin
if (select count(*) from table) > 0 then
update table
end if;
end;
Note there is no need for an "else".
Edited
You can simply do it all within the update statement (ie no if construct):
update table
set ...
where ...
and exists (select 'x' from table where ...)
not so elegant but you dont need to declare any variable:
for k in (select max(1) from table where 1 = 1) loop
update x where column = value;
end loop;
I have a store procedure that will
Update maximum 500 rows status from 0 to 1
Return those rows to program via cursor
Here is my store procedure code
PROCEDURE process_data_out (
o_rt_cursor OUT SYS_REFCURSOR
) IS
v_limit NUMBER;
l_data_ids VARCHAR2(32000);
BEGIN
v_limit := 500; -- limit 500
l_data_ids := '';
-- Create loop to get data
FOR i IN (
SELECT *
FROM
(
SELECT id FROM
TBL_DATA a
WHERE
a.created_at BETWEEN SYSDATE - 0.5 AND SYSDATE + 0.1
AND a.status = 0
AND a.phone NOT IN (SELECT phone FROM TBL_BIG_TABLE_1)
AND a.phone NOT IN (SELECT phone FROM TBL_BIG_TABLE_2 WHERE IS_DENY = 1)
ORDER BY
priority
)
WHERE
ROWNUM <= v_limit
) LOOP
BEGIN
-- Build string of ids like id1,id2,id3,
l_data_ids := l_data_ids
|| i.id
|| ',';
-- update row status to prevent future repeat
UPDATE TBL_DATA
SET
status = 1
WHERE
id = i.id;
END;
END LOOP;
COMMIT;
-- If string of ids length >0 open cursor to take data
IF ( length(l_data_ids) > 0 )
THEN
-- Cut last comma id1,id2,id3, --> id1,id2,id3
l_data_ids := substr(l_data_ids,1,length(l_data_ids) - 1);
-- open cursor
OPEN o_rt_cursor FOR
SELECT
id,
phone
FROM
TBL_DATA a
WHERE
a.id IN (
SELECT
to_number(column_value)
FROM
XMLTABLE ( l_data_ids )
);
END IF;
EXCEPTION
WHEN OTHERS THEN
ROLLBACK;
END process_data_out;
I want to optimize this performance and here is my question
Should I replace in by exists
Replace
AND a.phone NOT IN (SELECT phone FROM TBL_BIG_TABLE_1)
AND a.phone NOT IN (SELECT phone FROM TBL_BIG_TABLE_2 WHERE IS_DENY = 1)
by
AND NOT Exists (SELECT phone FROM TBL_BIG_TABLE_1 where TBL_BIG_TABLE_1.phone = a.phone)
AND NOT Exists (SELECT phone FROM TBL_BIG_TABLE_2 WHERE TBL_BIG_TABLE_2.phone = a.phone and IS_DENY = 1)
Is there a better way than
Save a string of ids like id1,id2,id3 after update row status
Open cursor by select from string of ids
I appreciate for any suggestion.
Thank for your concern
Row-by-row processing is always slower and you are also creating the string of ids, which again takes time so overall performance is going down.
You can use the collection DBMS_SQL.NUMBER_TABLE to store the updated ids from the UPDATE statement using the RETURNING clause and use it in the cursor query.
Also, I have changed your update statement so that it does not use NOT IN and uses the LEFT JOINS and ROW_NUMBER analytical function for increasing the performance as follows:
CREATE OR REPLACE PROCEDURE PROCESS_DATA_OUT (
O_RT_CURSOR OUT SYS_REFCURSOR
) IS
V_LIMIT NUMBER;
L_DATA_IDS DBMS_SQL.NUMBER_TABLE;
BEGIN
V_LIMIT := 500; -- limit 500
UPDATE TBL_DATA A
SET A.STATUS = 1
WHERE A.ID IN (
SELECT ID
FROM ( SELECT ID,
ROW_NUMBER() OVER(ORDER BY PRIORITY) AS RN
FROM TBL_DATA B
LEFT JOIN TBL_BIG_TABLE_1 T1 ON T1.PHONE = B.PHONE
LEFT JOIN TBL_BIG_TABLE_2 T2 ON T2.IS_DENY = 1 AND T2.PHONE = B.PHONE
WHERE B.CREATED_AT BETWEEN SYSDATE - 0.5 AND SYSDATE + 0.1
AND B.STATUS = 0
AND T1.PHONE IS NULL
AND T2.PHONE IS NULL)
WHERE RN <= V_LIMIT ) RETURNING ID BULK COLLECT INTO L_DATA_IDS;
OPEN O_RT_CURSOR FOR SELECT ID, PHONE
FROM TBL_DATA A
WHERE A.ID IN (SELECT COLUMN_VALUE FROM TABLE ( L_DATA_IDS ));
EXCEPTION
WHEN OTHERS THEN
ROLLBACK;
END PROCESS_DATA_OUT;
/
I wrote a procedure where i am trying to insert value from source to destination table, i used bulk collect in order to execute the data for large amount of data.
create or replace PROCEDURE TEST2 (
p_array_size IN NUMBER
) IS
CURSOR cur1 IS SELECT DISTINCT
*
FROM
test;
CURSOR cur3 IS SELECT * FROM test;
CURSOR cur2( BND_TYPE d.bndng_typ%TYPE, BND_VAL d.bndng_val%TYPE, FINANCIAL_INST_ID d.financial_institution_id%TYPE, PRDCT_ID d.prdct_id%TYPE,
PRDCT_SUB_ID d.prdct_sub_id%TYPE, INSTRUMENT_ID d.instrmnt_id%TYPE) IS SELECT
d.*
FROM
test1 d,
(
SELECT
b.prdct_id,
FROM
test2 a,
test1 b
WHERE
a.ir_id = b.ir_id
AND a.price_component_id = b.price_component_id
AND a.financial_institution_id = b.financial_institution_id
GROUP BY
b.prdct_id,
) e
WHERE
d.prdct_id = e.prdct_id
AND d.bndng_typ = BND_TYPE
AND d.bndng_val = BND_VAL
AND d.financial_institution_id = FINANCIAL_INST_ID
AND d.prdct_id = PRDCT_ID
AND d.prdct_sub_id = PRDCT_SUB_ID
AND d.instrmnt_id = INSTRUMENT_ID ;
TYPE loan_data_tbl IS TABLE OF cur1%rowtype INDEX BY PLS_INTEGER;
loan_data loan_data_tbl;
TYPE loanrate_tbl IS TABLE OF cur2%rowtype INDEX BY BINARY_INTEGER;
loan_rate loanrate_tbl;
BEGIN
DECLARE
v_noofDays NUMBER:=0;
currentDt DATE;
BEGIN
SELECT * INTO currentDt FROM dt;
BEGIN
IF cur1%Isopen Then
Close cur1;
End IF;
IF cur2%Isopen Then
Close cur2;
End IF;
OPEN cur1;
LOOP
FETCH cur1 BULK COLLECT INTO loan_data LIMIT p_array_size;
EXIT WHEN loan_data.COUNT = 0;
FOR i IN 1..loan_data.COUNT
LOOP
OPEN cur3;
OPEN cur2(loan_data(i).bndng_typ, loan_data(i).bndng_val,loan_data(i).financial_institution_id,
loan_data(i).prdct_id, loan_data(i).prdct_sub_id, loan_data(i).instrmnt_id);
loop
FETCH cur2 BULK COLLECT INTO loan_rate LIMIT p_array_size;
EXIT WHEN loan_rate.COUNT = 0;
FOR j IN 1..loan_rate.COUNT
LOOP
IF(cur3.POS_NUM = loan_data(i).POS_NUM AND cur3.POS_TYPE = loan_data(i).POS_TYPE
AND cur3.PRICE_COMPONENT_ID = loan_rate(j).PRICE_COMPONENT_ID
AND cur3.RPRTD_TILL_DT = loan_data(i).RPRTD_TILL_DT) THEN
update test SET SEQ_NUM=1,
WHERE SEQ_NUM=2;
ELSE
INSERT INTO test VALUES (
....
....
);
END IF;
COMMIT;
END LOOP;
END LOOP;
CLOSE cur2;
CLOSE cur1;
END LOOP;
END LOOP;
CLOSE cur1;
END;
END;
End ;
/
In above procedure, i removed some column names for security purpose.
I am getting two errors one is
PLS-00201: identifier 'D.BNDNG_TYP' must be declared and
PLS-00225: subprogram or cursor 'cur3' reference is out of scope
if any one can help to solve this.
In cursor, You cannot provide the type referring to alias from cursor query. You need to just provide the table name test1 instead of alias d.
CURSOR cur2( BND_TYPE test1.bndng_typ%TYPE,
BND_VAL test1.bndng_val%TYPE,
FINANCIAL_INST_ID test1.financial_institution_id%TYPE,
PRDCT_ID test1.prdct_id%TYPE,
PRDCT_SUB_ID test1.prdct_sub_id%TYPE,
INSTRUMENT_ID test1.instrmnt_id%TYPE)
IS SELECT
.....
This is untested as I don't have your tables and data, but as a first refactoring I would start with this type of structure:
create or replace procedure test2
as
cursor loan_data_cur
( bnd_type cr_loan_prima_rate_orig.bndng_typ%type
, bnd_val cr_loan_prima_rate_orig.bndng_val%type
, financial_inst_id cr_loan_prima_rate_orig.financial_institution_id%type
, prdct_id cr_loan_prima_rate_orig.prdct_id%type
, prdct_sub_id cr_loan_prima_rate_orig.prdct_sub_id%type
, instrument_id cr_loan_prima_rate_orig.instrmnt_id%type )
is
select d.*
from test1 d
join ( select b.prdct_id
from test2 a
join test1 b
on b.ir_id = a.ir_id
and b.price_component_id = a.price_component_id
and b.financial_institution_id = a.financial_institution_id
group by b.prdct_id ) e
on d.prdct_id = e.prdct_id
where d.bndng_typ = loan_data_cur.bnd_type
and d.bndng_val = loan_data_cur.bnd_val
and d.financial_institution_id = loan_data_cur.financial_inst_id
and d.prdct_id = loan_data_cur.prdct_id
and d.prdct_sub_id = loan_data_cur.prdct_sub_id
and d.instrmnt_id = loan_data_cur.instrument_id;
begin
for loan_data in (
select distinct *
from test
)
loop
for loan_rate in loan_data_cur
( loan_data.bndng_typ
, loan_data.bndng_val
, loan_data.financial_institution_id
, loan_data.prdct_id
, loan_data.prdct_sub_id
, loan_data.instrmnt_id )
loop
update test t set t.seq_num = 1
where t.seq_num = 2
and t.pos_num = loan_data.pos_num
and t.pos_type = loan_data.pos_type
and t.price_component_id = loan_rate.price_component_id
and t.rprtd_till_dt = loan_data.rprtd_till_dt;
if sql%rowcount = 0 then
insert into test values (x, y, z);
end if;
end loop;
end loop;
commit;
end test2;
I would also looking at doing the join from test to test1 as a single cursor instead of explicitly with two separate cursors, as the optimiser might find a more efficient method, such as a hash join.
Probably the update/insert combination could be written as a single merge. Once you have a merge working, you might find you can apply it to the whole table in one shot and not need any cursor loops at all.
Bulk-collecting into an array could be useful if the number of row-by-row updates causes performance problems (if it's more than a few thousand, say). If so, you would want to structure it so that you could apply the updates/inserts using a forall construction, not more loops.
I am trying to write an oracle function that takes in a parentID and returns me a "table" that recursively finds its child and child's child and so on.. Here is my code for it but whenever I try and create the function I get these errors :
[Error] Compilation (17: 10): ORA-00604 : error occurred at recursive SQL level 1
ORA-01422 : exact fetch returns more than requested number of rows
ORA-06512
ORA-06512
And here is code:
CREATE OR REPLACE TYPE TAB_TableOfIDs IS TABLE OF NUMBER;
CREATE OR REPLACE FUNCTION RETURNCHILDIDS (pParentID NUMBER)
RETURN TAB_TableOfIDs
AS
ResultSet TAB_TableOfIDs;
BEGIN
WITH MY_VIEW (ID)
AS (SELECT A.ID
FROM MYTABLE A
WHERE A.ID = pParentID
UNION ALL
SELECT E.ID
FROM MYTABLE E
INNER JOIN MY_VIEW B
ON E.ParentID = B.ID)
select bulk collect into ResultSet from MY_VIEW;
RETURN ResultSet;
END;
/
The interesting part is , the code just works fine in a normal SQL query editor but if you put it in a function it gives the error.. So what am I doing wrong? and how can I fix it?
Thank you in advance.
I had my solution :
CREATE OR REPLACE FUNCTION RETURNCHILDIDS (pParentID NUMBER)
RETURN TAB_TableOfIDs
IS
CURSOR c1
IS
WITH ALT_BIRIM (targetID, ID, parentID)
AS (SELECT A.targetID, A.ID, A.PARENTID
FROM ENTKURUM A
WHERE A.ID = (SELECT ID
FROM ENTKURUM
WHERE targetID = pParentID)
UNION ALL
SELECT E.targetID, E.ID, E.PARENTID
FROM ENTKURUM E
INNER JOIN ALT_BIRIM B
ON E.PARENTID = B.ID)
SELECT targetID
FROM alt_birim;
ResultSet TAB_TableOfIDs := NEW TAB_TableOfIDs ();
BEGIN
FOR mrow IN c1
LOOP
ResultSet.EXTEND ();
ResultSet (ResultSet.COUNT) := mrow.targetID;
END LOOP;
RETURN ResultSet;
END;
Either:
Rewrite your SELECT INTO statement so that only one row is returned.
OR
Create a cursor and retrieve each row if you are unsure of how many records you might retrieve.
CREATE OR REPLACE FUNCTION RETURNCHILDIDS (pParentID IN NUMBER)
RETURN number
IS
cnumber number;
CURSOR c1
IS
[Your query]
...
BEGIN
OPEN c1;
FETCH c1 INTO cnumber;
CLOSE c1;
RETURN cnumber;
END;
create or replace procedure BAS_NUM_UPD is
cursor cur is
select distinct o.oi_b,mpr.pa_ke_i,ltrim(substr(convert_171_to_711(cp.p_t_num),1,7),'0') bs_nbr
from t_obj o, mat_pa_rel mp, cor_pa cp
where o.ob_t = 'something'
and o.oi_b = mp.oi_b
and mp.pa_ke_i = cp.pa_ke_i;
l_ba_num_at_i number(10) := get_attribute_id('Ba timber');
flag1 VARCHAR2(10);
type t1 is table of varchar2(10);
par_k t1;
BEGIN
for x in cur loop
BEGIN
select pa_ke_i into par_k from mat_pa_rel where oi_b=x.oi_b ;
if par_k.count=null then
insert into cs_val (oi_b, at_i, value, flag, ) values (x.oi_b, l_ba_num_at_i, null, 1);
end if;
select flag into flag1 from cs_val where at_i = l_ba_num_at_i and oi_b = x.oi_b
and value = x.bs_nbr;
EXCEPTION
when NO_DATA_FOUND THEN
insert into cs_val (oi_b, at_i, value, flag, )
values (x.oi_b, l_ba_num_at_i, x.bs_nbr, 1);
flag1 :='Nothing';
when OTHERS
then
raise_application_error(-20011,'Unknown Exception in PROCEDURE');
END;
end loop;
end BAS_NUM_UPD;
error:
PLS-00642: local collection types not allowed in SQL statements
You should get it running if you do a bulk collect
select pa_ke_i bulk collect into par_k from mat_pa_rel where oi_b=x.oi_b ;
Then I think the if is not right. I think you need to do
if par_k.count = 0 then
But to be honest you might just make a count
select count(*) into l_cnt from mat_pa_rel where oi_b=x.oi_b;
If l_cnt = 0 then ...
Of course l_cnt has to be defined.
You should create type t1 in the schema and not in the pl/sql block.