Compound Triggers - plsql

Problem - Want to avoid the problem of mutating triggers by using the compound trigger. But unable to do so
Background -
I want to insert data in new table " Tracking Table " whenever there is change in Main table "CUSTOM_ITEM"
Design is such that, everytime a row is created in table an ITEM_ID is generated but there is a column FIRST_ITEM_ID that remains same in some cases.
So whenever a new row is added, I want to check its FIRST_ITEM_ID and then check the whole table and find out all the ITEM_IDs having that same FIRST_ITEM_ID.
And I want to insert all those rows in the New table using trigger.
Is it even possible ?
Attaching the trigger :
CREATE OR REPLACE TRIGGER APP.TEST_TRG
FOR DELETE OR INSERT OR UPDATE
ON APP.CUSTOM_ITEM
COMPOUND TRIGGER
TYPE t_change_tab IS TABLE OF APP.TEST_TRG.OBJECT_ID%TYPE;
g_change_tab t_change_tab := t_change_tab();
BEFORE EACH ROW IS
BEGIN
Select item_id bulk collect into g_change_tab from CUSTOM_ITEM where first_item_id =
(Select first_item_id from CUSTOM_ITEM where item_id = :NEW.item_id);
For i in 1 .. g_change_tab.COUNT()
LOOP
g_change_tab.extend;
END LOOP;
END BEFORE EACH ROW;
AFTER STATEMENT IS
BEGIN
For i in 1 .. g_change_tab.COUNT()
LOOP
app.bc_acs_pkg.populate_TEST_TRG /* Package Inserts data */
(p_object_type => 'ITEM',
p_object_id => g_change_tab(i));
END LOOP;
g_change_tab.delete;
END AFTER STATEMENT;
END ;
/

You can do what you want just not with your current approach. Let's take a step back. What is a mutating table exception (ORA-04091). It is thrown when you attempt to access the table on which the trigger fired in a row level event, that is not permitted. Just creating a compound trigger does not remove that restriction. So in your Before Row segment the statement
Select item_id
bulk collect into g_change_tab
from CUSTOM_ITEM where first_item_id =
(Select first_item_id from CUSTOM_ITEM where item_id = :NEW.item_id);
is invalid, and results in raising ORA-04091. What you need is to just build your collection with the necessary ids. Then process them in the After statement segment.
create or replace trigger test_trg
for delete or insert or update
on custom_item
compound trigger
type t_change_tab is
table of custom_item.first_item%type;
g_change_tab t_change_tab := t_change_tab();
before each row is
l_first_item_exists boolean := false;
indx integer;
begin
indx := g_change_tab.first;
while not l_first_item_exists
and indx is not null
loop
l_first_item_exists := (g_change_tab(indx) = :new.first_item);
if not l_first_item_exists
then
indx := g_change_tab.next(indx);
end if;
end loop;
if not l_first_item_exists
then
g_change_tab.extend;
g_change_tab(g_change_tab.last) := :new.first_item;
end if;
end before each row;
after statement is
begin
for indx in g_change_tab.first .. g_change_tab.last
loop
insert into tracking_table(item_id, first_item)
select item_id, first_item
from custom_item
where first_item = g_change_tab(indx);
end loop;
end after statement;
end test_trg;
The issue here is the loops, always the slowest processing, and very bad in triggers. Below is an approach which avoids them totally. It does however require creating your type array at the schema level.
create or replace type custom_item_change_t is
table of integer ; --custom_item.first_item%type;
create or replace trigger test_trg
for insert
on custom_item
compound trigger
g_change_tab custom_item_change_t := custom_item_change_t();
before each row is
begin
g_change_tab.extend;
g_change_tab(g_change_tab.last) := :new.first_item;
end before each row;
after statement is
begin
insert into tracking_table(item_id, first_item)
select item_id, first_item
from custom_item ci
where first_item in (select distinct column_value
from table(g_change_tab)
)
and not exists
( select null
from tracking_table tt
where ci.item_id = tt.item_id
and ci.first_item = tt.first_item
);
end after statement;
end test_trg;

Related

How to use the for loop in fetching Id's from a rows in a table to be used by a procedure in PLSQL?

This is my code below I get this error(Error at line 24/8: ORA-06550: line 20, column 12:PLS-00201: identifier 'A.ID' must be declared) as shown in the image below when I try running the code. Please how can I write the plsql code properly(using for loop) to fetch each row ID and pass them to the procedure?
BEGIN
DECLARE
p_id number(30);
p_status varchar(20);
BEGIN
for c in (
SELECT
a.ID,
a.STATUS
INTO
p_id,
p_status
from USER_COMMISSIONS a,
order_line b where a.order_line_id=b.id and a.status= 'unconfirmed'
)
LOOP
begin
p_id := a.ID;
p_status := a.STATUS;
EXCEPTION
WHEN NO_DATA_FOUND THEN
NULL;
end;
-- update pstk_payload set status = 'done' where id = pyld_id;
dbms_output.put_line(p_id);
-- PSTK_PAYMENT_PACKAGE.add_payment(p_amt, p_user_id, p_reference, p_name, p_narration, p_payment_date, p_net_amt, p_payment_type_id, p_transaction_type_id, p_payment_id, p_status);
END LOOP;
end;
END;
There's nothing to declare, actually - everything you need (at least, in code you posted and that's not commented) is contained in cursor itself.
As William commented, you need to reference columns with the cursor name (not tables that are their source).
Also, no need for any exception handler; cursor certainly won't return no_data_found; if its select doesn't return anything the only "consequence" will be that none of commands within the loop will be executed.
If you're joining tables, then use JOIN; leave where clause for conditions (if any).
Therefore:
begin
for c in (select a.id,
a.status
from user_commissions a join order_line b on a.order_line_id = b.id
where a.status= 'unconfirmed'
)
loop
dbms_output.put_line(c.id ||', '|| c.status);
end loop;
end;

Creating a Database Trigger that checks if more than one record was added on a date?

CREATE OR REPLACE TRIGGER POSITION_NUMBER
BEFORE UPDATE OR INSERT OR DELETE ON APPLIES
DECLARE
PRAGMA AUTONOMOUS_TRANSACTION;
NUMBER_OF_POSITIONS NUMBER;
BEGIN
SELECT count(pnumber) INTO NUMBER_OF_POSITIONS
FROM APPLIES WHERE anumber = :NEW.anumber;
IF( NUMBER_OF_POSITIONS > 2 AND count(APPDATE) > 2 )
THEN
RAISE_APPLICATION_ERROR(-20000,'an Employee cannot apply for
more than two positions');
END IF;
END;
/
Im attemtping to create a trigger that goes off if an Applicant applys for more than two Positions on the Same Day, but im not sure how i would implement the Date side of it. Below is the set of relational Schemeas
You can use the TRUNC function to remove the time portion and then see if the application date matches today's date, regardless of time.
Also, there is no need for the autonomous transaction pragma. You are not executing any DML.
CREATE OR REPLACE TRIGGER position_number
BEFORE UPDATE OR INSERT OR DELETE
ON applies
DECLARE
number_of_positions NUMBER;
BEGIN
SELECT COUNT (pnumber)
INTO number_of_positions
FROM applies
WHERE anumber = :new.anumber AND TRUNC (appdate) = TRUNC (SYSDATE);
IF number_of_positions > 2
THEN
raise_application_error (
-20000,
'An Employee cannot apply for more than two positions on the same day');
END IF;
END;
/

PL/SQL Trigger gets a mutating table error

My trigger wants to check if a 'new' manager supervises no more than 5 employees.
Manager supervising only 5 people are in BLOCKED_MANAGER table(ssn,numberofemployees).
Finally, every update is recorded in SUPERLOG table(date,user,old_manager,new_manager).
I get no compiling error about the trigger, but when I update a superssn I get this error:
SQL> update employee set superssn='666666607' where ssn='111111100';
update employee set superssn='666666607' where ssn='111111100'
*
ERROR at line 1:
ORA-04091: Table FRANK.EMPLOYEE is mutating, the trigger/function
can't read it
ORA-06512: a "FRANK.TLOG", line 20
ORA-04088: error during execution of trigger 'FRANK.TLOG'
How can I solve this trigger? Thank you
create or replace trigger tlog
before update of superssn on employee
for each row
declare
t1 exception;
n number:=0;
cont number:=0;
empl varchar2(16);
cursor cur is (select ssn from blocked_manager where ssn is not null);
begin
open cur;
loop
fetch cur into empl;
exit when cur%notfound;
if(:new.superssn = empl) then
n:=1;
end if;
end loop;
close cur;
if n=1 then
raise t1;
end if;
select count(*) into cont from employee group by superssn having superssn=:new.superssn;
if(cont=4) then
insert into blocked_manager values(:new.superssn,5);
end if;
insert into superlog values(sysdate,user,:old.superssn, :new.superssn );
exception
when t1 then
raise_application_error(-20003,'Manager '||:new.superssn||' has already 5 employees');
end;
Probably the quickest way around this is to use a carefully constructed statement trigger instead of a row trigger. Row triggers have the phrase FOR EACH ROW in them, are invoked for each row which is modified (based on the BEFORE/AFTER INSERT, BEFORE/AFTER UPDATE, and BEFORE/AFTER DELETE constraints on the trigger), can see the appropriate :NEW and :OLD values, and are subject to the "can't look at the table on which the trigger is defined" rule. Statement triggers are invoked at the appropriate time for each statement which is executed, can't see row values, but aren't subject to the limits on looking at the particular table on which they're defined. So for the portions of your logic which don't need to work with :NEW or :OLD values a trigger such as this might prove useful:
CREATE OR REPLACE TRIGGER EMPLOYEE_S_BU
BEFORE UPDATE ON EMPLOYEE
-- Note: no BEFORE EACH ROW phrase, so this is a statement trigger
BEGIN
-- The following FOR loop should insert rows into BLOCKED_MANAGER for all
-- supervisors which have four or more employees under them and who are not
-- already in BLOCKED_MANAGER.
FOR aRow IN (SELECT e.SUPERSSN, COUNT(e.SUPERSSN) AS EMP_COUNT
FROM EMPLOYEE e
LEFT OUTER JOIN BLOCKED_MANAGER b
ON b.SSN = e.SUPERSSN
WHERE b.SSN IS NULL
GROUP BY e.SUPERSSN
HAVING COUNT(e.SUPERSSN) >= 4)
LOOP
INSERT INTO BLOCKED_MANAGER
(SSN, EMPLOYEE_COUNT)
VALUES
(aRow.SUPERSSN, aRow.EMP_COUNT);
END LOOP;
-- Remove rows from BLOCKED_MANAGER for managers who supervise fewer
-- than four employees.
FOR aRow IN (SELECT e.SUPERSSN, COUNT(e.SUPERSSN) AS EMP_COUNT
FROM EMPLOYEE e
INNER JOIN BLOCKED_MANAGER b
ON b.SSN = e.SUPERSSN
GROUP BY e.SUPERSSN
HAVING COUNT(e.SUPERSSN) <= 3)
LOOP
DELETE FROM BLOCKED_MANAGER
WHERE SSN = aRow.SUPERSSN;
END LOOP;
-- Finally, if any supervisor has five or more employees under them,
-- raise an exception. Note that we go directly to EMPLOYEE to determine
-- the number of employees supervised.
FOR aRow IN (SELECT SUPERSSN, COUNT(*) AS EMP_COUNT
FROM EMPLOYEE
GROUP BY SUPERSSN
HAVING COUNT(*) >= 5)
LOOP
-- If we get here we've found a supervisor with 5 (or more) employees.
-- Raise an exception
RAISE_APPLICATION_ERROR(-20000, 'Found supervisor ' || aRow.SUPERSSN ||
' supervising ' || aRow.EMP_COUNT ||
' employees');
END LOOP;
END EMPLOYEE_S_BU;
Note that if you get rid of the BLOCKED_MANAGER table (which this trigger still maintains, although I don't know if it's truly necessary) the logic gets cut down considerably.
You'll still need a row trigger to handle the logging, but as that's just a matter of cutting down your existing trigger I'll leave that to you. :-)
Share and enjoy.
As you have discovered, you cannot select from the same table that a row-level trigger is defined against; it causes a table mutating exception.
In order to properly create this validation using a trigger a procedure should be created to obtain user-specified locks so the validation can be correctly serialized in a multi-user environment.
PROCEDURE request_lock
(p_lockname IN VARCHAR2
,p_lockmode IN INTEGER DEFAULT dbms_lock.x_mode
,p_timeout IN INTEGER DEFAULT 60
,p_release_on_commit IN BOOLEAN DEFAULT TRUE
,p_expiration_secs IN INTEGER DEFAULT 600)
IS
-- dbms_lock.allocate_unique issues implicit commit, so place in its own
-- transaction so it does not affect the caller
PRAGMA AUTONOMOUS_TRANSACTION;
l_lockhandle VARCHAR2(128);
l_return NUMBER;
BEGIN
dbms_lock.allocate_unique
(lockname => p_lockname
,lockhandle => p_lockhandle
,expiration_secs => p_expiration_secs);
l_return := dbms_lock.request
(lockhandle => l_lockhandle
,lockmode => p_lockmode
,timeout => p_timeout
,release_on_commit => p_release_on_commit);
IF (l_return not in (0,4)) THEN
raise_application_error(-20001, 'dbms_lock.request Return Value ' || l_return);
END IF;
-- Must COMMIT an autonomous transaction
COMMIT;
END request_lock;
This procedure can then be used in a compound trigger (assuming at least Oracle 11, this will need to be split into individual triggers in earlier versions)
CREATE OR REPLACE TRIGGER too_many_employees
FOR INSERT OR UPDATE ON employee
COMPOUND TRIGGER
-- Table to hold identifiers of inserted/updated employee supervisors
g_superssns sys.odcivarchar2list;
BEFORE STATEMENT
IS
BEGIN
-- Reset the internal employee supervisor table
g_superssns := sys.odcivarchar2list();
END BEFORE STATEMENT;
AFTER EACH ROW
IS
BEGIN
-- Store the inserted/updated supervisors of employees
IF ( ( INSERTING
AND :new.superssn IS NOT NULL)
OR ( UPDATING
AND ( :new.superssn <> :old.superssn
OR :new.superssn IS NOT NULL AND :old.superssn IS NULL) ) )
THEN
g_superssns.EXTEND;
g_superssns(g_superssns.LAST) := :new.superssn;
END IF;
END AFTER EACH ROW;
AFTER STATEMENT
IS
CURSOR csr_supervisors
IS
SELECT DISTINCT
sup.column_value superssn
FROM TABLE(g_superssns) sup
ORDER BY sup.column_value;
CURSOR csr_constraint_violations
(p_superssn employee.superssn%TYPE)
IS
SELECT count(*) employees
FROM employees
WHERE pch.superssn = p_superssn
HAVING count(*) > 5;
r_constraint_violation csr_constraint_violations%ROWTYPE;
BEGIN
-- Check if for any inserted/updated employee there exists more than
-- 5 employees for the same supervisor. Serialise the constraint for each
-- superssn so concurrent transactions do not affect each other
FOR r_supervisor IN csr_supervisors LOOP
request_lock('TOO_MANY_EMPLOYEES_' || r_supervisor.superssn);
OPEN csr_constraint_violations(r_supervisor.superssn);
FETCH csr_constraint_violations INTO r_constraint_violation;
IF csr_constraint_violations%FOUND THEN
CLOSE csr_constraint_violations;
raise_application_error(-20001, 'Supervisor ' || r_supervisor.superssn || ' now has ' || r_constraint_violation.employees || ' employees');
ELSE
CLOSE csr_constraint_violations;
END IF;
END LOOP;
END AFTER STATEMENT;
END;
You do not need the blocked_manager table to manage this constraint. This information can be derived from the employee table.
Or in versions earlier than Oracle 11i:
CREATE OR REPLACE PACKAGE employees_trg
AS
-- Table to hold identifiers of inserted/updated employee supervisors
g_superssns sys.odcivarchar2list;
END employees_trg;
CREATE OR REPLACE TRIGGER employee_biu
BEFORE INSERT OR UPDATE ON employee
IS
BEGIN
-- Reset the internal employee supervisor table
employees_trg.g_superssns := sys.odcivarchar2list();
END;
CREATE OR REPLACE TRIGGER employee_aiur
AFTER INSERT OR UPDATE ON employee
FOR EACH ROW
IS
BEGIN
-- Store the inserted/updated supervisors of employees
IF ( ( INSERTING
AND :new.superssn IS NOT NULL)
OR ( UPDATING
AND ( :new.superssn <> :old.superssn
OR :new.superssn IS NOT NULL AND :old.superssn IS NULL) ) )
THEN
employees_trg.g_superssns.EXTEND;
employees_trg.g_superssns(employees_trg.g_superssns.LAST) := :new.superssn;
END IF;
END;
CREATE OR REPLACE TRIGGER employee_aiu
AFTER INSERT OR UPDATE ON employee
IS
DECLARE
CURSOR csr_supervisors
IS
SELECT DISTINCT
sup.column_value superssn
FROM TABLE(employees_trg.g_superssns) sup
ORDER BY sup.column_value;
CURSOR csr_constraint_violations
(p_superssn employee.superssn%TYPE)
IS
SELECT count(*) employees
FROM employees
WHERE pch.superssn = p_superssn
HAVING count(*) > 5;
r_constraint_violation csr_constraint_violations%ROWTYPE;
BEGIN
-- Check if for any inserted/updated employee there exists more than
-- 5 employees for the same supervisor. Serialise the constraint for each
-- superssn so concurrent transactions do not affect each other
FOR r_supervisor IN csr_supervisors LOOP
request_lock('TOO_MANY_EMPLOYEES_' || r_supervisor.superssn);
OPEN csr_constraint_violations(r_supervisor.superssn);
FETCH csr_constraint_violations INTO r_constraint_violation;
IF csr_constraint_violations%FOUND THEN
CLOSE csr_constraint_violations;
raise_application_error(-20001, 'Supervisor ' || r_supervisor.superssn || ' now has ' || r_constraint_violation.employees || ' employees');
ELSE
CLOSE csr_constraint_violations;
END IF;
END LOOP;
END;

nested table element assignment in pl/sql

My requirement in pl/sql nested table is the following:
I have a nested table collection type declared and I populate the elements based on a lookup from a table.
In cases where the lookup yields more than one row(more than one value for code), then add all those multiple values in the nested table and proceed. Here is where i am stuck.
I am not able to increment that parent counter "indx" inside the exception to process those multiple rows. Since i am not, it only stores the latest data in the nested table and not all of them.
declare
TYPE final_coll_typ IS TABLE OF varchar2(100);
l_final_coll final_coll_typ;
MULTI_FETCH EXCEPTION;
PRAGMA EXCEPTION_INIT(MULTI_FETCH, -1422); -- this is an ora error for exact fetch returns more than the required number of rows
begin
for indx in 1..<count> loop
<some processing logic here>
select code into l_final_coll(indx) from lookup_tbl where <some filter>;
exception
when MULTI_FETCH then
for p in (select code from lookup_tbl where <some filter>)
loop
l_final_coll(indx) := p.code;
dbms_output.put_line(l_final_coll(indx));
end loop;
continue; -- this is for further processing after the loop
end loop;
end;
Lets say, the first iteration of the counter indx produced only one row data for code. That gets stored in l_final_coll(indx).
Lets say the next iteration of indx i the main for loop produces 2 rows of values for code. My thought was to catch the exception (ORA-01422) and keep adding these 2 code values in the existing nested table.
So, in effect, my nested table should now have 3 values of code in its element. But, currently, I can only get it to populate 2 of them (the single value from first itreration and the latest value from the next)
Any pointers would be appreciated on how I can accomplish this.
PS: Tried manipulating the counter variables indx and p. But, obviously pl/sql does not allow it for a "for loop".
You don't need to do a single select at all, just use the cursor loop to start with, and append to the collection (which you'd have to initialise):
declare
type final_coll_typ is table of varchar2(100);
l_final_coll final_coll_typ;
begin
l_final_coll := final_coll_typ();
for indx in 1..<count> loop
<some processing logic here>
for p in (select code from lookup_tbl where <some filter>) loop
l_final_coll.extend(1);
l_final_coll(l_final_coll.count) := p.code;
end loop;
end loop;
dbms_output.put_line('Final size: ' || l_final_coll.count);
end;
/
For each row found but the cursor, the collection is extended by one (which isn't very efficient), and the cursor value is put in the last, empty, row; which is found from the current count.
As a demo, if I create a dummy table with a duplicate value:
create table lookup_tbl(code varchar2(100));
insert into lookup_tbl values ('Code 1');
insert into lookup_tbl values ('Code 2');
insert into lookup_tbl values ('Code 2');
insert into lookup_tbl values ('Code 3');
... then with a specific counter and filter:
declare
type final_coll_typ is table of varchar2(100);
l_final_coll final_coll_typ;
begin
l_final_coll := final_coll_typ();
for indx in 1..3 loop
for p in (select code from lookup_tbl where code = 'Code ' || indx) loop
l_final_coll.extend(1);
l_final_coll(l_final_coll.count) := p.code;
end loop;
end loop;
dbms_output.put_line('Final size: ' || l_final_coll.count);
end;
/
... I get:
anonymous block completed
Final size: 4
As a slightly more complicated option, you could bulk-collect all the matching data into a temporary collection, then loop over that to append those values into the real collection. Something like:
declare
type final_coll_typ is table of varchar2(100);
l_final_coll final_coll_typ;
l_tmp_coll sys.dbms_debug_vc2coll;
begin
l_final_coll := final_coll_typ();
for indx in 1..<count> loop
<some processing logic here>
select code bulk collect into l_tmp_coll from lookup_tbl where <some filter>;
for cntr in 1..l_tmp_coll.count loop
l_final_coll.extend(1);
l_final_coll(l_final_coll.count) := l_tmp_coll(cntr);
end loop;
end loop;
end;
/
There may be a quicker way to combine two collections but I'm not aware of one. Bulk-collect has to be into a schema-level collection type, so you can't use your local final_coll_typ. You can create your own schema-level type, and then use it for both the temporary and final collection variables; but I've used a built-in one, sys.dbms_debug_vc2coll, which is defined as table of varchar2(1000).
As a demo, with the same table/data as above, and the same specific count and filter:
declare
type final_coll_typ is table of varchar2(100);
l_final_coll final_coll_typ;
l_tmp_coll sys.dbms_debug_vc2coll;
begin
l_final_coll := final_coll_typ();
for indx in 1..3 loop
select code bulk collect into l_tmp_coll
from lookup_tbl where code = 'Code ' || indx;
for cntr in 1..l_tmp_coll.count loop
l_final_coll.extend(1);
l_final_coll(l_final_coll.count) := l_tmp_coll(cntr);
end loop;
end loop;
dbms_output.put_line('Final size: ' || l_final_coll.count);
end;
/
... I again get:
anonymous block completed
Final size: 4

How to insert into a table correctly using table of records and forall in pl/sql

I want to insert records into MY_TABLE using forall. But the no. of records dat gets inserted keeps on changing with each test run! I think it has something to do with loop counter but I am not able to figure out. Here's the code snippet.
DECLARE
TYPE l_rec_type IS RECORD (
datakey SOURCE_TABLE.datakey%TYPE,
sourcekey SOURCE_TABLE.sourcekey%TYPE,
DESCRIPTION SOURCE_TABLE.DESCRIPTION%TYPE,
dimension_name SOURCE_TABLE.dimension_name%TYPE ,
data_type SOURCE_TABLE.data_type%TYPE
);
TYPE l_table_type IS TABLE OF l_rec_typeINDEX BY PLS_INTEGER;
l_table l_table_type;
l_cntr NUMBER;
BEGIN
FOR rec_dimname IN (SELECT dimension_name FROM dimension_table) LOOP
l_cntr1 := 1
FOR rec_source IN (SELECT * FROM source_table WHERE data_type IS NOT NULL) LOOP
l_table(l_ctr1).datakey := rec_source.datakey;
l_table(l_ctr1).sourcekey := rec_source.sourcekey;
l_table(l_ctr1).DESCRIPTION := rec_source.DESCRIPTION;
l_table(l_ctr1).dimension_name := rec_source.dimension_name;
l_table(l_ctr1).data_type := rec_source.data_type;
l_cntr1 := l_cntr1+1;
END LOOP
FORALL j IN l_table.FIRST..l_table.LAST
INSERT INTO my_table VALUES(l_table(j).datakey,
l_table(j).sourcekey,
l_table(j).DESCRIPTION,
l_table(j).dimension_name,
l_table(j).data_type,
1,
SYSDATE,
login_id
);
END LOOP;
END;
What am I doing wrong? Normal insert using for loop is inserting 5000 records. Another problem that I am facing is how to handle WHEN DUP_VAL_ON_INDEX and WHEN OTHERS exception using forall. In nornal for loop its easy. But I have to use FORALL for fast inserts. Please help!
Looking at your code I can see that you not delete the data stored in the pl/table inside your loop and you don't have a order by to your query's. So if the first iteration have more data then the second you will have duplicate data.
So after initializing your l_cntr1 var (l_cntr1 := 1) you must clear your pl/table:
l_table.delete;
Hope that helps.
Here's the fixed code. Plus SAVE EXCEPTIONS really saved my day!. Here is how I implemented the solution. Thank you all for your valuable time and suggestions.
DECLARE
TYPE l_rec_type IS RECORD (
datakey SOURCE_TABLE.datakey%TYPE,
sourcekey SOURCE_TABLE.sourcekey%TYPE,
DESCRIPTION SOURCE_TABLE.DESCRIPTION%TYPE,
dimension_name SOURCE_TABLE.dimension_name%TYPE ,
data_type SOURCE_TABLE.data_type%TYPE
);
TYPE l_table_type IS TABLE OF l_rec_typeINDEX BY PLS_INTEGER;
l_table l_table_type;
l_cntr NUMBER;
ex_dml_errors EXCEPTION;
PRAGMA EXCEPTION_INIT(ex_dml_errors, -24381);
login_id NUMBER := -1;
errm VARCHAR2(512);
err_indx NUMBER
BEGIN
FOR rec_dimname IN (SELECT dimension_name FROM dimension_table) LOOP
l_cntr1 := 1;
l_table.DELETE; -- Added
FOR rec_source IN (SELECT * FROM source_table WHERE data_type IS NOT NULL) LOOP
l_table(l_ctr1).datakey := rec_source.datakey;
l_table(l_ctr1).sourcekey := rec_source.sourcekey;
l_table(l_ctr1).DESCRIPTION := rec_source.DESCRIPTION;
l_table(l_ctr1).dimension_name := rec_source.dimension_name;
l_table(l_ctr1).data_type := rec_source.data_type;
l_cntr1 := l_cntr1+1;
END LOOP
FORALL j IN l_table.FIRST..l_table.LAST SAVE EXCEPTIONS
INSERT INTO my_table VALUES(l_table(j).datakey,
l_table(j).sourcekey,
l_table(j).DESCRIPTION,
l_table(j).dimension_name,
l_table(j).data_type,
1,
SYSDATE,
login_id
);
END LOOP;
END LOOP;
EXCEPTION
WHEN ex_dml_errors THEN
l_error_count := SQL%BULK_EXCEPTIONS.count;
DBMS_OUTPUT.put_line('Number of failures: ' || l_error_count);
errm := SQLERRM(-SQL%BULK_EXCEPTIONS(i).ERROR_CODE);
err_indx := SQL%BULK_EXCEPTIONS(i).error_index
FOR i IN 1 .. l_error_count LOOP
DBMS_OUTPUT.put_line('Error: ' || i ||
' Array Index: ' || SQL%BULK_EXCEPTIONS(i).error_index ||
' Message: ' || SQLERRM(-SQL%BULK_EXCEPTIONS(i).ERROR_CODE));
IF errm LIKE '%unique%constraint%violated' THEN -- Insert into my_multiple_entries_tbl on duplicate value on index DATAKEY
INSERT INTO my_multiple_entries_tbl(my_multiple_entries_tbl_seq.NEXTVAL,
l_table(err_indx).datakey,
l_table(err_indx).sourcekey,
l_table(err_indx).data_type,
SYSDATE,
login_id );
ELSE -- Insert into my_other_errors_tbl on other errors
INSERT INTO my_other_errors_tbl ( my_other_errors_tbl_seq.NEXTVAL,
l_table(err_indx).datakey,
l_table(err_indx).sourcekey,
l_table(err_indx).data_type,
SYSDATE,
login_id );
END IF;
END;
Your seem to be inserting exactly the same thing multiple times - you're solely looping through the count of dimension_table, which means it can be simplified to the following, which will be faster. At the bottom is a forall version.
You can't use exception when dup_val_on_index with either version, you have to do it row by row. Judging solely by what you've posted I suspect that you can actually achieve what you're trying to do in a single query and save all this problem completely ( including dealing with duplicate values ).
declare
i integer;
begin
select count(*)
into i
from dimension_table;
for j in 1 .. i loop
insert into my_table (datakey, sourcekey, description
, dimension_name, someother_column
, some_date_column, login_id
select datakey, sourcekey, description, dimension_name
, data_type, 1, sysdate, login_id -- previously missing
from source_table
where data_type is not null;
end loop;
commit;
end;
/
If, however, you really want to use forall you can do something like this:
declare
cursor c_src is
select datakey, sourcekey, description, dimension_name
, data_type, 1, sysdate, login_id -- previously missing
from source_table
where data_type is not null;
type t__src is table of c_src%rowtype index by binary_integer;
t_src t__src;
i integer;
begin
select count(*)
into i
from dimension_table;
for j in 1 .. i loop
open c_src;
loop
fetch c_src bulk collect into t_src;
forall k in t_src.first .. t_src.last
insert into my_table (datakey, sourcekey, description
, dimension_name, someother_column
, some_date_column, login_id
values t_src;
end loop;
close c_src;
end loop;
commit;
end;
/

Resources