handling runtime exception for sql functions - function

Question

I use db2 9.7.6 full edition for windows. I need develop functions which can be use on sql select expression.
Functions must contain modifying sql data and handling runtime exceptions. There are two variants, but I have problems of implementations all requirments in ever variants.
The first variant is implementation of sql table function, e.g.
CREATE FUNCTION func1 (val CHAR(20))
RETURNS table(result varchar(1000))
LANGUAGE SQL
MODIFIES SQL DATA
BEGIN atomic
insert into a values(val);
return (select result from a);
END
It works, but I can't implement handling exception how in sql procedures. When I tried to use block "declare exit handler", I got a syntax errors.
The second variant is implementation of pl/sql function, e.g.
CREATE OR REPLACE FUNCTION bb
RETURN varchar2
MODIFIES SQL DATA
AS
BEGIN
insert into st values ('a');
return 0;
END bb;
But when I tried to execute this function, I got error "SQLCODE=-740, SQLSTATE=51034 is defined with the MODIFIES SQL DATA option, which is not valid in the context where the routine is invoked". Help me please.
best regards, Turkin Andrew.

Answer 1

There are many differences between inlined SQL compound and compiled compound. I wrote the following script, and it runs ok in db2 10.1 for LUW.
Script.sql
CREATE or replace FUNCTION func1a (val CHAR(20))
RETURNS varchar(20)
LANGUAGE SQL
READS SQL DATA
BEGIN
declare ret varchar(20);
declare exit handler for sqlstate '02000' resignal sqlstate '08888';
select C1 into ret from T1 fetch first 1 row only;
return ret;
END#
CREATE or replace FUNCTION func1b (val CHAR(20))
RETURNS table(result varchar(20))
LANGUAGE SQL
MODIFIES SQL DATA
BEGIN atomic
insert into T1 values(val);
return (select C1 from T1);
END#
The execution (In Windows client, but it does not matter)
db2 CREATE TABLE T1 (C1 CHAR(20))
db2 -td# -vf Script.sql
db2 "values func1a('s')"
db2 "SELECT * FROM TABLE (FUNC1b('A'))"
As you can see there are many difference in DB2 10.1 for inlined and compiled SQL:
- Single value return vs table or row return.
- Reads vs modified data.
- Condition handler vs nothing.
Due to the function definition, the calling method is different, in one case it is a scalar value, in the other is a function table.

Answer 2

It just doesn't work.
List of maximum access levels
Routine type Default SQL access level Maximum allowed SQL access level
SQL procedures MODIFIES SQL DATA MODIFIES SQL DATA
SQL functions (scalar functions) READS SQL DATA READS SQL DATA
SQL functions (table functions) READS SQL DATA MODIFIES SQL DATA
External procedures MODIFIES SQL DATA MODIFIES SQL DATA
External functions (scalar functions) READS SQL DATA READS SQL DATA
External functions (table functions) READS SQL DATA READS SQL DATA
For scalar functions the maximum level is "READS SQL DATA". There is probably now way around that, so you have to reconsider your coding.