When using a stored procedure to dynamically generate a table based on configuration and return a result set (SELECT) with the records in that table, the second call to that procedure to generate a different table structure returns no records and it reports a missing column from a previous temporary table of the same name.
I tried this with MariaDB 10.3 and 10.1.21 and received the same result. I have minimized my code here to the minimum to demonstrate the error after trying several variations of single and multiple sub-procedures.
I also tried using some transaction control with COMMITS after executing the process, before trying to start the process with a different parameter, but got the same results.
DROP PROCEDURE IF EXISTS CreateATable;
DELIMITER $$
CREATE PROCEDURE CreateATable( _TableType tinyint )
BEGIN
DROP TEMPORARY TABLE IF EXISTS aTable;
IF _TableType = 1 THEN
SET #SQL_Statement :=
CONCAT(
'CREATE TEMPORARY TABLE aTable (',
'the_id bigint, ',
'the_column varchar(100) ',
') engine=INNODB',
';');
ELSE
SET #SQL_Statement :=
CONCAT(
'CREATE TEMPORARY TABLE aTable (',
'the_id bigint, ',
'the_other_column varchar(100) ',
') engine=INNODB',
';');
END IF;
PREPARE stmtCreateTable FROM #SQL_Statement;
EXECUTE stmtCreateTable;
DEALLOCATE PREPARE stmtCreateTable;
SET #SQL_Statement := NULL;
END$$
DELIMITER ;
DROP PROCEDURE IF EXISTS GetATable;
DELIMITER $$
CREATE PROCEDURE GetATable()
BEGIN
CALL CreateATable( 1 );
SELECT * FROM aTable;
CALL CreateATable( 2 );
SELECT * FROM aTable;
END$$
DELIMITER ;
DROP PROCEDURE IF EXISTS GetATable2;
DELIMITER $$
CREATE PROCEDURE GetATable2(_TableType tinyint)
BEGIN
CALL CreateATable( _TableType );
SELECT * FROM aTable;
END$$
DELIMITER ;
/*
Test execution script starts here
*/
-- Just CALL Create for one and Select
CALL CreateATable( 1 );
DESCRIBE aTable;
SELECT * FROM aTable;
CALL CreateATable( 2 );
DESCRIBE aTable;
SELECT * FROM aTable;
-- -> no errors
-- now CALL procedure to Create and Select from two different temp tables
CALL GetATable();
-- -> no errors
-- now CALL procedure to CREATE AND SELECT from ONE temp table definition using a parameter to select
CALL GetATable2(1);
CALL GetATable2(2);
-- Error Code: 1054. Unknown column 'mySchema.aTable.the_column' in 'field list'
I would expect that I can pass a parameter to a stored procedure to generate a temporary table, and return the records of that temporary table. Even if I call that same procedure multiple times with different parameters on the same session.
The actual results are that when the stored procedure is called to generate the temporary table with a different table structure, it returns this error complaining about the column missing from the temporary table created in the previous invocation of that same stored procedure.
Error Code: 1054. Unknown column 'mySchema.aTable.the_column' in 'field list'
The only way I have found to prevent this error is
a. ending the jdbc connection and ending the server session
b. recompiling one of the stored procedures in the call stack
Recompiling is not viable. And ending the session seems unreasonable.
This seems like a defect. But would be interested to find if there is some way to get this to work.
This seems like a bug and you can report it directly to the MariaDB team at MariaDB bugs database.
A temporary solution is to use a prepared statement in the stored procedure GetATable2 (my test on MariaDB 10.3.16 to use EXECUTE IMMEDIATE):
...
CREATE PROCEDURE `GetATable2`(`_TableType` TINYINT)
BEGIN
CALL CreateATable(`_TableType`);
-- SELECT * FROM `aTable`;
EXECUTE IMMEDIATE 'SELECT * FROM `aTable`';
END$$
...
See dbfiddle.
Related
When I am running this Netezza stored procedure, I am getting an error
attribute 'SOME_VALUE' not found
As per requirement I have to get value from one table (TABLE_A) and insert into another table (TABLE_B).
This is the procedure:
create or replace procedure my_proc()
returns boolean
execute as owner
language NZPLSQL
as
BEGIN_PROC
declare rec RECORD ;
BEGIN
for rec in SELECT * from TABLE_A loop
EXECUTE IMMEDIATE
'INSERT INTO TABLE_B(COLUMN_B)
values( '|| rec.COLUMN_A_OFTABLE_A || ')';
END LOOP;
END;
END_PROC;
execute my_proc()
Here below, I am able to insert a string. But I need to insert different value depending on other table as I mentioned above.
EXECUTE IMMEDIATE 'INSERT INTO TABLE_B(COLUMN_B) values( ''Y'');';
When building a string that you are going run EXECUTE IMMEDIATE against, you have be careful to have everything quoted properly. In your case it's thinking that it needs to treat SOME_VALUE as an attribute/column, and it can't any column with that name.
Wrap your column reference in quote_literal() and it will interpret the contents of your column and quote-escape it properly for you.
create or replace procedure my_proc()
returns boolean
execute as owner
language NZPLSQL
as
BEGIN_PROC
declare rec RECORD ;
BEGIN
for rec in SELECT * from TABLE_A loop
EXECUTE IMMEDIATE
'INSERT INTO TABLE_B(COLUMN_B)
values( '|| quote_literal(rec.COLUMN_A_OFTABLE_A) || ')';
END LOOP;
END;
END_PROC;
You can find some more information in the documentation here.
Note: I am assuming that you have some more complicated logic to implement in this stored procedure, because looping over row by row will be much, much slower that insert..select. Often by an order of magnitude.
I'm new to PL/SQL. I'm using oracle 11g XE along with sql developer. I'm trying to create to create an after insert trigger as follows
create or replace trigger tr1
after
insert ON
employee
for each row
begin
print :new.emp_id;
end;
The employee table is as follows
create table employee
( emp_id varchar2(5) primary key,
emp_name varchar2(10),
salary number,
company varchar2(10) foreign key references companies(comp_name)
);
When I run the statement I got a 'enter binds' window for the bind variable :new. But I was confused that why do I need to enter the value for :new since it is pseudorecord. Then I entered 'employee' as the values for :new. Now I'm getting the following error.
Error(2,8): PLS-00103: Encountered the symbol "" when expecting one of the following: := . ( # % ; The symbol ":=" was substituted for "" to continue.
Your problem is not in the :new pseudorecord. The error is coming from the usage of print, which is used to print the bind variable used in successful PL/SQL block or used in an EXECUTE command. For example, you can use it this way:
VARIABLE n NUMBER
BEGIN
:n := 1;
END;
/
Then
PRINT n;
But if you want to test the value being inserted, you can use DBMS_OUTPUT.PUT_LINE like this:
create or replace trigger tr1
after
insert ON
employee
for each row
BEGIN
dbms_output.put_line(:new.emp_id);
END;
/
Enable DBMS_OUTPUT window in your SQL Developer, then run
insert into employee values(1, 'empName', 1000, 'ABC');
You'll see 1 printed out.
However, you can always test the value from the table. Because the value should be already inserted into table. You can just query.
Let's have a look on my source code:
CREATE OR REPLACE PROCEDURE MAKE_COPY_OF_CLASSROOMS AUTHID CURRENT_USER AS
TYPE classrooms_table_type IS TABLE OF classrooms%ROWTYPE INDEX BY PLS_INTEGER;
classrooms_backup classrooms_table_type;
CURSOR classrooms_cursor IS
SELECT *
FROM classrooms
WHERE year = 1
ORDER BY name;
v_rowcnt PLS_INTEGER := 0;
BEGIN
OPEN classrooms_cursor;
FETCH classrooms_cursor
BULK COLLECT INTO classrooms_backup;
CLOSE classrooms_cursor;
EXECUTE IMMEDIATE 'CREATE TABLE classrooms_copy AS (SELECT * FROM classrooms WHERE 1 = 2)';
--COPY ALL STORED DATA FROM classrooms_backup TO classrooms_copy
END MAKE_COPY_OF_classrooms;
I'm stucked for hours on trying to insert data from "classrooms_backup" into the table "classrooms_copy", which is created by EXECUTE IMMEDIATE command. It's necessary to create table "classrooms_copy" via EXECUTE IMMEDIATE command. I tried to create another EXECUTE command with for loop in it:
EXECUTE IMMEDIATE 'FOR i IN classrooms_backup.FIRST..classrooms_backup.LAST LOOP
INSERT INTO classrooms_copy(id,room_id,year,name)
VALUES(classrooms_backup(i).id,classrooms_backup(i).room_id,classrooms_backup(i).year,classrooms_backup(i).name);
END LOOP;';
But it's road to the hell. I'm retrieving an invalid SQL statement error.
Thanks for your help!
There's no need for much PL/SQL here. Also, try to avoid the keyword CURSOR - there's almost always a better way to do it.
create or replace procedure make_copy_of_classrooms authid current_user as
begin
execute immediate '
create table classrooms_copy as
select *
from classrooms
where year = 1
order by name
';
end make_copy_of_classrooms;
/
create or replace procedure sample
as
ID VARCHAR(20);
BEGIN
execute immediate
'CREATE GLOBAL TEMPORARY TABLE UPDATE_COLUMN_NO_TP
(
NAME VARCHAR2(256)
)';
INSERT INTO UPDATE_COLUMN_NO_TP
SELECT SRC_PK_COLUMNS.PK_KEY
FROM SRC_PK_COLUMNS
WHERE NOT EXISTS (
SELECT 1
FROM TGT_PK_COLUMNS
WHERE TGT_PK_COLUMNS.ID = SRC_PK_COLUMNS.ID);
END;
Error is:
The table is no exist.
So, I want a best solution for this scenario. In my stored procedure I have 10 temporary tables. All are all dynamic creations and inserts.
Table UPDATE_COLUMN_NO_TP not exists at compile time, so you got the error.
If you created a table dynamically, you should access it dynamically.
And pay attention to Mat's comment about essence of GTT.
execute immediate '
INSERT INTO UPDATE_COLUMN_NO_TP
SELECT SRC_PK_COLUMNS.PK_KEY
FROM SRC_PK_COLUMNS
WHERE NOT EXISTS (
SELECT 1
FROM TGT_PK_COLUMNS
WHERE TGT_PK_COLUMNS.ID = SRC_PK_COLUMNS.ID
)
';
Ok - I have a situation in which I must execute a dynamically built stored procedure against tables that may, or may not be in the database. The data retrieved is then shunted to a VB.Net backed ASP based report page. By design, if the tables are not present in the database, the relevant data is automatically hidden on the report page. Currently, I'm doing this by checking for the inevitable error, and hiding the div in the catch block. A bit kludgy, but it worked.
I can't include the VB code-behind, but the relevant stored procedure is included below.
However, a problem with this method was recently brought to my attention when, for no apparent reason, the div was being hidden even though the proper data was available. As it turned out, the user trying to select the table in the dynamic SQL call didn't have the proper select permissions, an easy enough fix once I could track it down.
So, two fold question. First and foremost - is there a better way to check for a missing table than through catching the error in the VB.Net codebehind? All things considered, I'd rather save the error checking for an actual error. Secondly, is there a preferred method to squirrel out a particular OLE DB error out of the general object caught by the try->catch block other than just checking the actual stack trace string?
SQL Query - The main gist of the code is that, due to the design of the database, I have to determine the name of the actual table being targeted manually. The database records jobs in a single table, but each job also gets its own table for processing data on the items processed in that job, and it's data from those tables I have to retrieve. Absolutely nothing I can do about this setup, unfortunately.
DECLARE #sql NVarChar(Max),
#params NVarChar(Max),
#where NVarChar(Max)
-- Retained for live testing of stored procedure.
-- DECLARE #Table NvarChar(255) SET #Table = N'tblMSGExportMessage_10000'
-- DECLARE #AcctID Integer SET #AcctID = 10000
-- DECLARE #Type Integer SET #Type = 0 -- 0 = Errors only, 1 = All Messages
-- DECLARE #Count Integer
-- Sets our parameters for our two dynamic SQL calls.
SELECT #params = N'#MsgExportAccount INT, #cnt INT OUTPUT'
-- Sets our where clause dependent upon whether we want all results or just errors.
IF #Type = 0
BEGIN
SELECT #where =
N' AND ( mem.[MSGExportStatus_OPT_CD] IN ( 11100, 11102 ) ' +
N' OR mem.[IngestionStatus_OPT_CD] IN ( 11800, 11802, 11803 ) ' +
N' OR mem.[ShortcutStatus_OPT_CD] IN ( 11500, 11502 ) ) '
END
ELSE
BEGIN
SELECT #where = N' '
END
-- Retrieves a count of messages.
SELECT #sql =
N'SELECT #cnt = Count( * ) FROM dbo.' + QuoteName( #Table ) + N' AS mem ' +
N'WHERE mem.[MSGExportAccount_ID] = #MsgExportAccount ' + #where
EXEC sp_executesql #sql, #params, #AcctID, #cnt = #Count OUTPUT
To avoid an error you could query the sysobjects table to find out if the table exists. Here's the SQL (replace YourTableNameHere). If it returns > 0 then the table exists. Create a stores procedure on the server that runs this query.
select count(*)
from sysobjects a with(nolock)
where a.xtype = 'U'
and a.name = 'YourTableNameHere'