I'm trying to learn how to create a user defined aggregate function. So far, I've been able to create one that compiles fine, but calling it gives an unexpected result. The function is a very simple test function that looks through a number of rows that are either set to 'Y' or 'N' and returns 'Y' if all are set to 'Y' and otherwise returns 'N'. I'm running it on a single row and getting back a blank varchar 2 instead.
I'm not sure what is the procedure to go through with debugging this. I've tried using DBMS_OUTPUT.PUT_LINE(), but I cannot see anything on the database output. The largest problem is that it is creating the function fine, and most of the code is in an object type. Thus, if I were to try to debug the select statement, it is calling code on the database that has already been compiled.
Below is the code for the function, but I don't want to know why this isn't working as much as I want to know how to debug so I can solve these issues myself, especially when more complex aggregate functions are involved.
CREATE OR REPLACE TYPE MYSCHEMA.ALL_TRUE_T AS OBJECT
(
TRUE_SO_FAR VARCHAR2(1),
STATIC FUNCTION ODCIAggregateInitialize(sctx IN OUT ALL_TRUE_T) RETURN NUMBER,
MEMBER FUNCTION ODCIAggregateIterate(self IN OUT ALL_TRUE_T, value IN VARCHAR2) RETURN NUMBER,
MEMBER FUNCTION ODCIAggregateTerminate(self IN ALL_TRUE_T, returnValue OUT VARCHAR2, flags IN NUMBER) RETURN NUMBER,
MEMBER FUNCTION ODCIAggregateMerge(self IN OUT ALL_TRUE_T, ctx2 IN ALL_TRUE_T) RETURN NUMBER
);
CREATE OR REPLACE TYPE BODY MYSCHEMA.ALL_TRUE_T IS
STATIC FUNCTION ODCIAggregateInitialize(sctx IN OUT ALL_TRUE_T)
RETURN NUMBER IS
BEGIN
sctx := ALL_TRUE_T('Y');
return ODCIConst.Success;
END;
MEMBER FUNCTION ODCIAggregateIterate(self IN OUT ALL_TRUE_T, value IN VARCHAR2)
RETURN NUMBER IS
BEGIN
IF value <> 'Y' OR self.TRUE_SO_FAR <> 'Y' THEN
self.TRUE_SO_FAR := 'N';
END IF;
RETURN ODCIConst.Success;
END;
MEMBER FUNCTION ODCIAggregateTerminate(self IN ALL_TRUE_T, returnValue OUT VARCHAR2, flags IN NUMBER)
RETURN NUMBER IS
BEGIN
returnValue := self.TRUE_SO_FAR;
RETURN ODCIConst.Success;
END;
MEMBER FUNCTION ODCIAggregateMerge(self IN OUT ALL_TRUE_T, ctx2 IN ALL_TRUE_T)
RETURN NUMBER IS
BEGIN
IF ctx2.TRUE_SO_FAR = 'N' THEN
self.TRUE_SO_FAR := 'N';
END IF;
RETURN ODCIConst.Success;
END;
END;
CREATE OR REPLACE PACKAGE MYSCHEMA.ALL_TRUE_PKG IS
FUNCTION ALL_TRUE (input VARCHAR2) RETURN VARCHAR2;
END;
CREATE OR REPLACE PACKAGE BODY MYSCHEMA.ALL_TRUE_PKG IS
FUNCTION ALL_TRUE (input VARCHAR2) RETURN VARCHAR2
AGGREGATE USING ALL_TRUE_T;
END;
And here is how I call it. YN_TEST_TABLE currently has a single row with an 'N' in it.
SELECT
MYSCHEMA.ALL_TRUE_PKG.ALL_TRUE(YN)
FROM
MYSCHEMA.YN_TEST_TABLE
Finally, I'm not sure if this is relevant, but I'm using Toad 11.6.
Edit:
So I've tried inserting into a temp log table and that didn't work either.
I added the following
MEMBER FUNCTION ODCIAggregateIterate(self IN OUT ALL_TRUE_T, value IN VARCHAR2)
RETURN NUMBER IS
BEGIN
BEGIN
INSERT INTO MYSCHEMA.LAWTONFOGLES_TEMP_LOG
(
ID,
Message,
Time
)
VALUES
(
'all_true',
'test1',
systimestamp
);
END;
IF value <> 'Y' OR self.TRUE_SO_FAR <> 'Y' THEN
self.TRUE_SO_FAR := 'N';
END IF;
RETURN ODCIConst.Success;
END;
There was nothing in the temp log, but also no error message. It is as if none of the 4 aggregate function parts are even being called.
EDIT2:
So, to make things more interesting, this works when it is not in a package.
I did the following
CREATE OR REPLACE FUNCTION MYSCHEMA.LAWTONFOGLES_ALL_TRUE (input VARCHAR2) RETURN VARCHAR2
AGGREGATE USING ALL_TRUE_T;
and then ran this
SELECT
MYSCHEMA.LAWTONFOGLES_ALL_TRUE(YN)
FROM
MYSCHEMA.YN_TEST_TABLE
and got the results I expected. It seems that the code itself isn't a problem, but putting it in a package causes it to break. Thursday my Oracle DBA will be opening a ticket up with oracle, so I'll be sure to update with why does putting this in a package break it but leaving it as just a function doesn't when they get back with us. Until then I may just have to keep this outside of a package.
Also, I tried to add a put_line on it when it was working and still did not get an output. I think that the way user defined aggregate functions work prevent put_line from working.
If you're using TOAD, be sure to turn on DBMS_OUTPUT recording before you run your proc so you can see your outputs. It should be on the bottom DBMS tab (if you have it open). Typically you'll see a red circle since it's defaulted as off. Click the circle so that it's green.
See this link as an example: http://geekbrigade.wordpress.com/2009/04/09/how-to-set-and-view-dbms-output-of-oralce-in-toad/
Related
Can you call a PL/SQL procedure from inside a function?
I haven't come across with the practical example.So if anyone has come across with this please share.
Yes. You can call any pl/sql program from inside any other pl/sql program. A function can call a function, a procedure can call a procedure which calls a function, a function can invoke a TYPE BODY...do an INSERT..which causes a TRIGGER to fire.
A simple (not really practical) example.
Using the HR schema where you have an EMPLOYEES table which includes columns EMPLOYEE_ID, FIRST_NAME, and LAST_NAME.
I have a function that takes in an INTEGER which we use to look up an EMPLOYEE record. We take their first and last names, and concat them together, and return the value back in UPPERCASE text.
Before we do that however, we call a procedure which does nothing but take a 5 second nap using the DBMS_LOCK package.
The code:
create or replace procedure do_nothing_comments (x in integer, y in integer)
is
begin
null;
-- yeah, this is a dumb demo
dbms_lock.sleep(5);
end;
/
create or replace FUNCTION upper_name (
x IN INTEGER
) RETURN VARCHAR2 IS
upper_first_and_last VARCHAR2 (256);
BEGIN
SELECT upper (first_name)
|| ' '
|| upper (last_name)
INTO upper_first_and_last
FROM employees
WHERE employee_id = x;
do_nothing_comments (1, 2); -- here we are calling the procedure
RETURN upper_first_and_last;
END;
/
Now let's invoke the function.
DECLARE
X NUMBER;
v_Return VARCHAR2(200);
BEGIN
X := 101;
v_Return := UPPER_NAME(
X => X
);
:v_Return := v_Return;
END;
/
I'm going to do this in SQL Developer using the Execute feature with the function open:
I get the answer back...it just takes 5 seconds longer than it really needed to.
Here you go:
create or replace function demo
return varchar2
as
begin
dbms_output.put_line('Hello');
return 1;
end demo;
Hello I am attempting to do the following problem and keep receiving errors. Create a user-defined function called DOLLARS_BY_VEHICLE_TYPE that receives an input parameter of a concatenated make and model and then queries the VEHICLES and SALES_FACTS tables to return the total gross sales amount of the sales by that combination. My code is as follows:
CREATE OR REPLACE FUNCTION DOLLARS_BY_VEHICLE_TYPE (V_AUTODESC VARCHAR2)
RETURN VARCHAR2
IS
V_AMT VARCHAR2 (50);
BEGIN
SELECT NVL (SUM (GROSS_SALES_AMOUNT), 0)
INTO V_AMT
FROM VEHICLES v, SALES_FACTS s
WHERE V.Vehicle_Code = S.Vehicle_Code AND V.DESCRIPTION= V_AUTODESC;
RETURN V_AMT;
EXCEPTION
WHEN OTHERS
THEN
RETURN 0;
END;
/
However, when I run it I get:
Warning: Function created with compilation errors.
ERROR at line 1:
ORA-06575: Package or function DOLLARS_BY_VEHICLE_TYPE is in an invalid state
Please see that you are trying to return the sum from your function and that can be only number/int/float. So the return type of your function should be number not varchar2. Also note that you are using a deprecated sql syntax which can be changed. See below:
CREATE OR REPLACE FUNCTION DOLLARS_BY_VEHICLE_TYPE (V_AUTODESC VARCHAR2)
-- RETURN VARCHAR2
RETURN NUMBER
IS
--V_AMT VARCHAR2 (50);
V_AMT NUMBER;
BEGIN
SELECT NVL (SUM (GROSS_SALES_AMOUNT), 0)
INTO V_AMT
FROM VEHICLES v
inner join SALES_FACTS s
ON V.Vehicle_Code = S.Vehicle_Code
where V.DESCRIPTION = V_AUTODESC;
RETURN V_AMT;
EXCEPTION
WHEN OTHERS
THEN
RETURN 0;
END;
/
I apologize ahead of time for the probably basic question. I am student and it is crunch time!
I am using Oracle 10g Express.
I created a function:
create or replace FUNCTION test_glaccounts_description
(
account_description_param VARCHAR2
)
RETURN NUMBER
AS
description_dup_var NUMBER;
BEGIN
SELECT 1
INTO description_dup_var
FROM general_ledger_accounts
WHERE account_description = account_description_param;
RETURN description_dup_var;
EXCEPTION
WHEN NO_DATA_FOUND THEN
RETURN 0;
END;
And would like to use that function in a procedure.
I've tried:
PROCEDURE insert_gla_with_test
(
account_number_param NUMBER,
account_description_param VARCHAR2
)
AS
BEGIN
IF test_glaccounts_description = 1 THEN
INSERT INTO general_ledger_accounts
VALUES (account_number_param, account_description_param);
ELSE raise_application_error (-20001, 'Duplicate account description');
END IF;
END;
But it doesn't like the "test_gla_accounts" line.....what am I doing wrong?
To my understanding, the function returns a value of 1 or 0, in the procedure, if the function returned a 1, I would like the param fields added to the table. If the function returned a 0, I would like the procedure to raise the error.
test_glaccounts_description takes a parameter (account_description_param) and returns a NUMBER. In order to call the function, therefore, you need to pass in a parameter. Assuming that you want to pass in the account_description_param that is passed in to the insert_gla_with_test procedure
CREATE OR REPLACE PROCEDURE insert_gla_with_test
(
account_number_param NUMBER,
account_description_param VARCHAR2
)
AS
BEGIN
IF test_glaccounts_description( account_description_param ) = 1 THEN
INSERT INTO general_ledger_accounts
VALUES (account_number_param, account_description_param);
ELSE
raise_application_error (-20001, 'Duplicate account description');
END IF;
END;
I'm having this annoying issue with aggregate udfs where it always fails to invoke the iteration method. My code is just like the other examples around the web (including the ones on oracle docs and asktom). I tried to change the UDF type and the same happens everytime. It says:
ORA-29925: cannot execute DBAODB.WMCONCATROUTINES.ODCIAGGREGATEITERATE
ORA-06553: PLS-306: wrong number or types of arguments in call to 'ODCIAGGREGATEITERATE'
Oracle version is 11.1.0.7.0 and Here's my code:
CREATE OR REPLACE TYPE WmConcatRoutines as object (
tmpData number,
STATIC FUNCTION ODCIAggregateInitialize( wmInst IN OUT WmConcatRoutines) RETURN number,
MEMBER FUNCTION ODCIAggregateIterate(wmInst IN OUT WmConcatRoutines, value number) return number,
MEMBER FUNCTION ODCIAggregateMerge(wmInst IN OUT WmConcatRoutines, wmInst2 IN WmConcatRoutines) return number,
MEMBER FUNCTION ODCIAggregateTerminate(wmInst IN WmConcatRoutines, returnValue OUT number, flags IN number) return number
);
/
CREATE OR REPLACE TYPE BODY WmConcatRoutines IS
STATIC FUNCTION ODCIAggregateInitialize( wmInst IN OUT WmConcatRoutines) RETURN number IS
BEGIN
wmInst := WmConcatRoutines(0);
return ODCIConst.Success;
END;
MEMBER FUNCTION ODCIAggregateIterate(wmInst IN OUT WmConcatRoutines, value number) return number IS
BEGIN
wmInst.tmpData := wmInst.tmpData + value;
return ODCIConst.Success;
END;
MEMBER FUNCTION ODCIAggregateTerminate(wmInst IN WmConcatRoutines, returnValue OUT number, flags IN number) return number IS
BEGIN
returnValue := wmInst.tmpData;
return ODCIConst.Success;
END;
MEMBER FUNCTION ODCIAggregateMerge(wmInst IN OUT WmConcatRoutines, wmInst2 IN WmConcatRoutines) return number IS
BEGIN
wmInst.tmpData := wmInst.tmpData + wmInst2.tmpData;
return ODCIConst.Success;
END;
END;
/
CREATE OR REPLACE FUNCTION WM_CONCAT_test (input number) RETURN number
parallel_enable
AGGREGATE USING WmConcatRoutines;
/
Any ideas on what may be causing this? thanks in advance
The parameter names in your implementations of the ODCI functions have to match those in the documentation; so you have to use sctx, self, and ctx2 rather than wminst and wminst2.
Edit As APC briefly mentioned, it only seems to be self that is required; you can use something else in place of sctx and ctx2. I can't find a documentation reference...
Edit 2 Yes, I can... APC's reference to it being an object thing led me to this.
I'm aware that Oracle does not have a boolean type to use for parameters, and am currently taking in a NUMBER type which would have 1/0 for True/False (instead of the 'Y'/'N' CHAR(1) approach).
I'm not a very advanced Oracle programmer, but after doing some digging and reading some ASKTOM posts, it seems like you can restrict a field using a format for the column like:
MyBool NUMBER(1) CHECK (MyBool IN (0,1))
Is there a way to apply the same sort of a check constraint to an input parameter to a stored procedure? I'd like to restrict the possible inputs to 0 or 1, rather than checking for it explicitly after receiving the input.
You can use Booleans as parameters to stored procedures:
procedure p (p_bool in boolean) is...
However you cannot use Booleans in SQL, e.g. select statements:
select my_function(TRUE) from dual; -- NOT allowed
For a number parameter there is no way to declaratively add a "check constraint" to it, you would have to code some validation e.g.
procedure p (p_num in number) is
begin
if p_num not in (0,1) then
raise_application_error(-20001,'p_num out of range');
end if;
...
Yes and no.
You can do..
create or replace package t_bool is
subtype t_bool_num IS PLS_INTEGER RANGE 0..1;
function f_test (i_bool_num t_bool_num) return varchar2;
end t_bool;
/
create or replace package body t_bool is
function f_test (i_bool_num t_bool_num) return varchar2 is
begin
if i_bool_num = 0 then
return 'false';
elsif i_bool_num = 1 then
return 'true';
elsif i_bool_num is null then
return 'null';
else
return to_char(i_bool_num);
end if;
end;
end t_bool;
/
The good news is that, if you do
exec dbms_output.put_line(t_bool.f_test(5));
it reports an error.
The bad news is that if you do
select t_bool.f_test(5) from dual;
then you don't get an error