I am using mariadb 10.3.9, and have created a user defined aggregate function (UDAF) and placed in a common_schema. This schema contains my utility functions to be used by other schema/databases on the same server.
The issue is that when calling the UDAF while using any other schema, it always return NULL!
The following is to demonstrate the issue:
CREATE SCHEMA IF NOT EXISTS common_schema;
DELIMITER $$
DROP FUNCTION IF EXISTS common_schema.add_ints $$
CREATE FUNCTION common_schema.add_ints(int_1 INT, int_2 INT) RETURNS INT NO SQL
BEGIN
RETURN int_1 + int_2;
END $$
DROP FUNCTION IF EXISTS common_schema.sum_ints $$
CREATE AGGREGATE FUNCTION common_schema.sum_ints(int_val INT) RETURNS INT
BEGIN
DECLARE result INT DEFAULT 0;
DECLARE CONTINUE HANDLER FOR NOT FOUND RETURN result;
LOOP FETCH GROUP NEXT ROW;
SET result = common_schema.add_ints(result, int_val);
END LOOP;
END $$
DELIMITER ;
Now, calling it this way, returns the result as expected:
USE common_schema;
SELECT common_schema.sum_ints(seq)
FROM (SELECT 1 seq UNION ALL SELECT 2) t;
-- result: 3
Calling it using any other schema, it returns NULL:
USE other_schema;
SELECT common_schema.sum_ints(seq)
FROM (SELECT 1 seq UNION ALL SELECT 2) t;
-- result: null
Am I missing something here? Is there any configuration that is missing?
Appreciate your help.
Reported as a Bug https://jira.mariadb.org/browse/MDEV-18100.
As a workaround, create the UDAF in every schema.
Related
Create a procedure named 'payment_count' which accepts one output parameter p_count with number as its data type. This procedure should print the count of successful payments (response_msg as 'Successful') with alias name as 'successful_payment_count'.
Hints: Procedure name : payment_count
Output parameter : p_count
Table used : payment_authorization
Function used : count
create or replace procedure payment_count(p_count output number)
as
begin
select count(response_msg='Successful') into successful_payment_count from payment_authorization;
return successful_payment_count;
end;
/
schema payment_authorization
Is this a task in an exam?
Sounds so... anyway:
You have to change keyword output to OUT, change the COUNT(response_msg='Successful') to COUNT(*) and add response_msg='Successful' to the where clause of the query.
create or replace procedure payment_count(p_count out number)
as
begin
select count(response_msg) as successful_payment_count
into p_count
from payment_authorization
where response_msg = 'Successful';
end;
/
I'm new in POSTGRESQL and I am creating PostgreSQL functions, and now I want to create a function which returns two table output, I created and it does not work
my function
CREATE FUNCTION getData() RETURNS setof refcursor AS
$$
DECLARE c_top_items refcursor;
DECLARE c_shopping_cart refcursor;
BEGIN
OPEN c_top_items FOR
SELECT t.name
FROM employee t
LIMIT 10;
RETURN NEXT c_top_items;
OPEN c_shopping_cart FOR
SELECT c.name
FROM employee c
LIMIT 5;
RETURN NEXT c_shopping_cart;
END;
$$ LANGUAGE plpgsql;
Please help!!
When you call getData you get two cursors, fetch all from them. I modified the example in the documentation and fetched the data with:
select * from getData();
fetch all in "<unnamed portal 1>";
fetch all in "<unnamed portal 2>";
But you would probably be better off by changing the signature of the function according to the return multiple cursors-example in the documentation. This is that example:
CREATE FUNCTION myfunc(refcursor, refcursor) RETURNS SETOF refcursor AS $$
BEGIN
OPEN $1 FOR SELECT * FROM table_1;
RETURN NEXT $1;
OPEN $2 FOR SELECT * FROM table_2;
RETURN NEXT $2;
END;
$$ LANGUAGE plpgsql;
-- need to be in a transaction to use cursors.
BEGIN;
SELECT * FROM myfunc('a', 'b');
FETCH ALL FROM a;
FETCH ALL FROM b;
COMMIT;
See: Returning Cursors
I have to use existing stored procedure which returns REF CURSOR. I need to insert that resultset into a temporary table.
Spec of procedure is:
TYPE cur IS REF CURSOR;
PROCEDURE get(p_one NUMBER ,p_two OUT cur);
How can I insert the resultset of this procedure into a table.
I just re-read the title of your question. Do you actually need to be able to select from a procedure?
If so, this can be achieved by using pipelined functions.
The process for this is:
Create an object type to represent the record-type you require.
Create a nested table type of the object.
Create a pipelined function which returns the nest table.
You can then select from this function.
This example should get you on your way:
create or replace type to_test as object (
val1 varchar2(32),
val2 varchar2(32)
);
create or replace type tt_test as table of to_test;
create or replace function demo_pipe return tt_test pipelined
is
v_test to_test;
begin
for rec in (select * from user_tables) loop
v_test := to_test(rec.table_name, rec.tablespace_name);
pipe row (v_test);
end loop;
end;
/
select * from table(demo_pipe);
input
Package name (IN)
procedure name (or function name) (IN)
A table indexed by integer, it will contain values that will be used to execute the procedure (IN/OUT).
E.g
let's assume that we want to execute the procedure below
utils.get_emp_num(emp_name IN VARCHAR
emp_last_name IN VARCHAR
emp_num OUT NUMBER
result OUT VARCHAR);
The procedure that we will create will have as inputs:
package_name = utils
procedure_name = get_emp_num
table = T[1] -> name
T[2] -> lastname
T[3] -> 0 (any value)
T[4] -> N (any value)
run_procedure(package_name,
procedure_name,
table)
The main procedure should return the same table that has been set in the input, but with the execution result of the procedure
table = T[1] -> name
T[2] -> lastname
T[3] -> 78734 (new value)
T[4] -> F (new value)
any thought ?
You can achieve it with EXECUTE IMMEDIATE. Basically, you build a SQL statement of the following form:
sql := 'BEGIN utils.get_emp_num(:1, :2, :3, :4); END;';
Then you execute it:
EXECUTE IMMEDIATE sql USING t(1), t(2), OUT t(3), OUT t(4);
Now here comes the tricky part: For each number of parameters and IN/OUT combinations you need a separate EXECUTE IMMEDIATE statement. And to figure out the number of parameters and their direction, you need to query the ALL_ARGUMENTS table first.
You might be able to simplify it by passing the whole table as a bind argument instead of a separate bind argument for each table element. But I haven't quite figured out how you would do that.
And the next thing you should consider: the elements of the table T your using will have a type: VARCHAR, NUMBER etc. So the current mixture where you have both numbers and strings won't work.
BTW: Why do you want such a dynamic call mechanism anyway?
Get from the all_arguments table the argument_name, data_type, in_out, and the position
Build the PLSQL block
DECLARE
loop over argument_name and create the declare section
argument_name data_type if in_out <> OUT then := VALUE OF THE INPUT otherwise NULL
BEGIN
--In the case of function create an additional argument
function_var:= package_name.procedure_name( loop over argument_name);
--use a table of any_data, declare it as global in the package
if function then
package_name.ad_table.EXTEND;
package_name.ad_table(package_name.ad_table.LAST):= function_var;
end if
--loop over argument_name IF IN_OUT <> IN
package_name.ad_table.EXTEND;
package_name.ad_table(package_name.ad_table.LAST):=
if data_type = VARCHAR2 then := ConvertVarchar2(argument_name)
else if NUMBER then ConvertNumber
else if DATE then ConvertDate
...
END;
The result is stored in the table.
To get value use Access* functions
I want to create a function that returns the number of rows in a table called Rating with a where clause.Where am i going wrong before the declare statement and the end statement?
create or replace
FUNCTION get_movies(user IN NUMBER) RETURN NUMBER
IS
DECLARE cnt NUMBER;
BEGIN
SELECT count(*)
INTO cnt
FROM rating
where userid= user;
RETURN cnt;
END;
I will appreciate help.Thanks.
You should not have the DECLARE keyword. You only need that for an anonymous block (or a sub-block).
create or replace
FUNCTION get_movies(p_userid IN NUMBER) RETURN NUMBER
IS
cnt NUMBER;
BEGIN
...
user is a reserved word so I'd suggest not using that as your parameter name. In the where clause I'm not sure if it will use your parameter value, or the name of the user executing the function; which would error as that string value couldn't be implicitly converted to a number.