I have a custom function in SQL which I want to return value NULL if it reports an error without breaking or stoping the procedure using this function.
How can I do that?
I know that the error handling should be done inside the function it self but what should I write in it?
example:
Procedure name is PROC
, Function Name is FUNC
code:
create dbo.PROC
AS
select column1, column2, FUNC(column3) as funccolumn
from table1
do somthing else
GO
You can try to check if the value you are going to pass to the function is a number:
drop procedure sp_test
go
drop function fn_test
go
drop table t_test
go
create table t_test (id int, name varchar(255))
go
insert into t_test values(1,'coco')
insert into t_test values(2,'cucu')
insert into t_test values(2,'10')
insert into t_test values(2,'10.5')
go
create function fn_test(#p decimal(18,6)) returns int
as
begin
return #p * 10
end
go
create procedure sp_test
as
begin
select id, name,
/* when is a number you call the function */
case when isnumeric(name)<>0 then dbo.fn_test(name)
/* if isn't a number you return NULL */
else null end as call_to_fn_test
from t_test
end
go
sp_test
Related
I am building a function on postgresql, basically send one id from one table then re-build the insert statement of that row and insert it as string column from another table.
I have this table, in insert_query I want to store the insert statement of one row, with his variables:
create table get_insert (tabname varchar(30), insert_query varchar(5000));
I want to store something like this on insert_query column:
Insert into baseball_table (code, name) select '01','Robet';
Currently my function is storing just this, which doesn't work since I need to store the real values:
INSERT INTO baseball_table(code,name) SELECT code,name FROM baseball_table WHERE id=1;
This is my function:
CREATE OR REPLACE FUNCTION get_values(
_id character varying
)
RETURNS boolean
LANGUAGE 'plpgsql'
VOLATILE PARALLEL UNSAFE
AS $function$
DECLARE v_id integer;
DECLARE sql_brand varchar;
BEGIN
sql_query'INSERT INTO baseball_table(code,name) SELECT code,name FROM core.brand WHERE id=' || v_id ||'
';
INSERT INTO core.clone_brand (tabname, insert_query)VALUES ('brand',sql_query);
RETURN true;
END;
$function$;
Which is the best way to get the real values without making variables of each column?
Regards
I want to get the way to get the real values without making variables of each column.
I am trying to fetch data from remote db by using dblink through function but getting an error "query has no destination for result data". I am using plpgsql language to do the same.
Function:
CREATE OR REPLACE FUNCTION fun()
RETURNS text AS
$$
begin
select dblink_connect(
'port=5432 dbname=test user=postgres password=****');
WITH a AS (
SELECT *
FROM dblink(
'SELECT slno,fname,mname,lname
FROM remote_tbl'
) AS t (slno int, fname text, mname text, lname text)
)
, b AS (
INSERT INTO temptab1
SELECT slno, name
FROM a
)
, c AS (
INSERT INTO temptab2
SELECT slno, name
FROM a
)
INSERT INTO temptab3
SELECT slno, name
FROM a;
select dblink_disconnect();
end;
$$
LANGUAGE plpgsql;
Calling Function:
select fun();
Error: query has no destination for result data
The stored procedure won't just return the result of the last SELECT. You need to actually return the value:
CREATE OR REPLACE FUNCTION fun() RETURNS text AS $$
BEGIN
--- ....
RETURN(SELECT dblink_disconnect());
END
$$ LANGUAGE plpgsql;
You're getting the error because Postgres expects the function to return something of type text, but your function doesn't return anything.
Use a plain SQL function instead of PL/PgSQL, or use SELECT INTO and ordinary RETURN.
Reason for the error you're getting is because there is no return in between your BEGIN and END for example:
BEGIN
update mytable set lastcodeused = to_char(cast(lastcodeused as INTEGER)+1, 'FM999999999999999999') where
classid = classIdVar and appid= appIdInt
RETURNING concat(pageUniqueCode,lastcodeused) as pageUniqueCode
into taskcode;
return taskcode;
END;
If you have this error using a pgplsql procedure or function, and you are sure that the return is defined correctly, there exists a different non-intuitive cause. I needed some time to realize this so I think it is worth sharing. I had a function like this:
CREATE OR REPLACE FUNCTION "db".fn_x(
id integer)
RETURNS TABLE(b_val varchar(100), c_val varchar(100))
LANGUAGE 'plpgsql'
AS $BODY$
DECLARE
var_b_val varchar(100);
var_c_val varchar(100);
BEGIN
select var_b, var_c
-- Missing INTO clause was the cause of the error.
var_b_val, var_c_val
from "db".table_y where y_id = id;
return query(select var_b_val, var_c_val);
END;
$BODY$;
Just adding that missing INTO clause made the function work correctly.
In conclusion, this error can also trigger on silent syntax errors.
I have the following simplified testcase:
CREATE TABLE test(id serial PRIMARY KEY, data varchar);
CREATE VIEW test_v AS SELECT * from test;
CREATE FUNCTION insert_test_fn(in_data varchar) RETURNS integer
AS $$
DECLARE
my_id integer;
BEGIN
INSERT INTO test (data) VALUES (in_data)RETURNING id INTO my_id;
-- do things with my_id
RETURN my_id;
END;
$$ LANGUAGE plpgsql;
CREATE RULE _INSERT AS
ON INSERT TO test_v DO INSTEAD
SELECT insert_test_fn(new.data);
INSERT INTO test_v (data) VALUES ('testval');
-- I would like to do this:
-- INSERT INTO test_v (data) VALUES ('testval') RETURNING id;
I have some client code which issues the statement
INSERT INTO test_v (data) VALUES ('testval') RETURNING id;.
This used to insert into test (ON INSERT TO test_v DO INSTEAD INSERT INTO test ... RETURNING...).
Now I need to change part of the behavior and instead of inserting directly into test, I want to call a function where this is done. (Parts of the data to be inserted will have to be calculated first).
Is there a way to define the _INSERT rule so that it still works when called with a RETURNING clause? This would allow me to leave the client code unchanged.
That will work just fine, but it is unnecessarily complicated:
You can directly insert into the view if its definition is as simple as that. But I assume that your actual case is more complicated.
You don't need the function, you can directly put INSERT INTO test ... in the rule definition.
I am trying to fetch data from remote db by using dblink through function but getting an error "query has no destination for result data". I am using plpgsql language to do the same.
Function:
CREATE OR REPLACE FUNCTION fun()
RETURNS text AS
$$
begin
select dblink_connect(
'port=5432 dbname=test user=postgres password=****');
WITH a AS (
SELECT *
FROM dblink(
'SELECT slno,fname,mname,lname
FROM remote_tbl'
) AS t (slno int, fname text, mname text, lname text)
)
, b AS (
INSERT INTO temptab1
SELECT slno, name
FROM a
)
, c AS (
INSERT INTO temptab2
SELECT slno, name
FROM a
)
INSERT INTO temptab3
SELECT slno, name
FROM a;
select dblink_disconnect();
end;
$$
LANGUAGE plpgsql;
Calling Function:
select fun();
Error: query has no destination for result data
The stored procedure won't just return the result of the last SELECT. You need to actually return the value:
CREATE OR REPLACE FUNCTION fun() RETURNS text AS $$
BEGIN
--- ....
RETURN(SELECT dblink_disconnect());
END
$$ LANGUAGE plpgsql;
You're getting the error because Postgres expects the function to return something of type text, but your function doesn't return anything.
Use a plain SQL function instead of PL/PgSQL, or use SELECT INTO and ordinary RETURN.
Reason for the error you're getting is because there is no return in between your BEGIN and END for example:
BEGIN
update mytable set lastcodeused = to_char(cast(lastcodeused as INTEGER)+1, 'FM999999999999999999') where
classid = classIdVar and appid= appIdInt
RETURNING concat(pageUniqueCode,lastcodeused) as pageUniqueCode
into taskcode;
return taskcode;
END;
If you have this error using a pgplsql procedure or function, and you are sure that the return is defined correctly, there exists a different non-intuitive cause. I needed some time to realize this so I think it is worth sharing. I had a function like this:
CREATE OR REPLACE FUNCTION "db".fn_x(
id integer)
RETURNS TABLE(b_val varchar(100), c_val varchar(100))
LANGUAGE 'plpgsql'
AS $BODY$
DECLARE
var_b_val varchar(100);
var_c_val varchar(100);
BEGIN
select var_b, var_c
-- Missing INTO clause was the cause of the error.
var_b_val, var_c_val
from "db".table_y where y_id = id;
return query(select var_b_val, var_c_val);
END;
$BODY$;
Just adding that missing INTO clause made the function work correctly.
In conclusion, this error can also trigger on silent syntax errors.
I have following stored procedure defined:
USE [BcmMetrice]
GO
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
ALTER PROCEDURE [dbo].[ActivityAdd_proc] #Name nvarchar(max),#Description nvarchar(max) =null ,#Users nvarchar(max),#Object_id nvarchar (15) =null, #Source nvarchar(10) =null, #TemplateId bigint =null, #UserID bigint =null
AS
DECLARE activityUsers_cursor CURSOR FOR
select s from dbo.SplitString(#Users, ';')
DECLARE
#new_ActivityId bigint,
#new_CommentId bigint,
#activityUser_l bigint
BEGIN TRY
INSERT INTO [BcmMetrice].[dbo].[Activity]
([Name]
,[Description]
,[Type]
,[Created])
VALUES
(#Name
,#Description
,ISNULL(#TemplateId,0)
,GETDATE())
SET #new_ActivityId = (SELECT SCOPE_IDENTITY())
INSERT INTO [BcmMetrice].[dbo].[Comment] ([UserID],[CommentText],[Timestamp])
VALUES (ISNULL(#UserID,151),'Activity has been created',GETDATE())
SET #new_CommentId = (SELECT SCOPE_IDENTITY())
INSERT INTO [BcmMetrice].[dbo].[ActivityComment] ([ActivityID],[CommentID])
VALUES (#new_ActivityId, #new_CommentId)
INSERT INTO [BcmMetrice].[dbo].[Log]([Timestamp],[Type],[Data],[StackTrace]) VALUES (GETDATE(),'SQL.ActivityAdd_proc','users='+ISNULL(CAST(#Users as varchar(max)),'empty'),null)
OPEN activityUsers_cursor
FETCH NEXT FROM activityUsers_cursor INTO #activityUser_l
WHILE ##FETCH_STATUS = 0
BEGIN
INSERT INTO [BcmMetrice].[dbo].[Log]([Timestamp],[Type],[Data],[StackTrace]) VALUES (GETDATE(),'SQL.ActivityAdd_proc','Inserting users='+ISNULL(CAST(#activityUser_l as varchar(max)),'empty'),null)
INSERT INTO [BcmMetrice].[dbo].[ActivityUser]
([ActivityId]
,[UserId]
,[Role])
VALUES
(#new_ActivityId
,#activityUser_l
,1)
FETCH NEXT FROM activityUsers_cursor INTO #activityUser_l
END
CLOSE activityUsers_cursor
DEALLOCATE activityUsers_cursor
END TRY
BEGIN CATCH
PRINT 'ERROR'
INSERT INTO [BcmMetrice].[dbo].[Log]([Timestamp],[Type],[Data],[StackTrace]) VALUES (GETDATE(),'SQL.ActivityAdd_proc','ERROR CATCHED!'+ERROR_MESSAGE(),null)
END CATCH
select #new_ActivityId
The thing I would like to do is to return from the procedure the id of a newly added activity. That is why at the very and I use line:
select #new_ActivityId
When testing this procedure in SQL Management Studio everything seems to be working fine. Problem starts when I try to use this procedure in my .NET project. I updated my edmx model form database, but when I execute this procedure the return value is invalid.
Procedure execution looks like this:
int ret = dc.Db.ActivityAdd_proc(name, description, users, object_id, source, templateId, userId);
Does anyone have an idea what I might be doing wrong?
I found solution to my problem. When you try to return data from stored procedure via select statement you must create Function Import in model browser of edmx. In a wizard you choose a type of returned collection (in my case it was scalar of type int64).