How to return less than 4000 characters from pl/sql function in SQL call?

Question

Hello,
Is there a way to limit length for varchar when calling pl/sql function from SQL? No matter how I write it it always returns 4000 bytes.
If there is none, then does it make sense ever to specify lenght of the return variable?
My goal is to encapsulate business rules within pl/sql functions. But if all varchar columns are returned as 4000 it is not feasible solution. Not only this is a performance issue in a data warehousing environment, bet when using those rules within SQL views user experiance would suffer as well. Are we left with the rule hardcoding solution? Also, I think that using SUBSTRING or TRUNC functions on top of business rules function defeats the purpose.
Please see my attempt below. Your thoughts are appreciated.
Thank you.
/* Formatted on 06/11/2009 2:26:41 PM (QP5 v5.126.903.23003) */
CREATE OR REPLACE FUNCTION mytest (myvar_in VARCHAR2)
RETURN VARCHAR2
AS
l_return VARCHAR2 (15);
BEGIN
l_return := 'TEST_' || myvar_in;
RETURN l_return;
END mytest;
CREATE TABLE TEST_ME
AS
SELECT mytest ('ME') AS VERYLONG FROM DUAL;
describe TEST_ME;
RUN ABOVE CODE:
Function created.
Table created.
TABLE TEST_ME
Name Null? Type
VERYLONG VARCHAR2(4000)
Edited by: Ilmars2 on Jun 11, 2009 2:46 PM

Answer 1

Pointless,
Thanks for jumping in on this and I am glad you asked :).
I do not doubt that it is an architectural challenge. Otherwise it would have been done already! I am struggling with the fact that SQL knows what data type the function will return, but does not know the length of it, precision or scale. I will leave it at that.
I will go with some high level requirements to allow for alternative thoughts:
1)     Business defined rules. There are multiple types of business rules. Simple lookups, bucketing, complex calculations, data retrieved from other tables etc. We have about 500 different rules. Some of them are even overloaded – different inputs will produce the same output.
Some simple examples are:
Rule1 - Fruit
     when ‘A’ then ‘Apple’
     When ‘O’ then ‘Orange’
     Else ‘N/A’ end
Rule2 – Bonus
     when STATUS =’Active’ and LEVEL=’CEO’ then bonus=salary*1.0
     when STATUS =’Active’ and LEVEL=’nobody’ then bonus=salary*0.01
     else bonus=0
Rule 3 – Income Bracket
     When more than 0 and less or equal to 30000 then INC_B=’LOW’
     When more than 30000 and less of equal to 60000 then INC_B=’MIDDLE’
Etc.
2)     Challenge: All data users in an organization need to use the same rules (let’s assume data source is Oracle database). How to expose all the rules to different types of users in manageable way? Types of users – analysts, application/web developers, data warehouse teams etc.
3)     Current system: Not only each user has coded their own rules (luckily based on the common specification), but hard-coding is taking place for each query within the confines of one project. The project I just looked at had about 12 modules with 30 hardcoded queries. Oops! Few rules just changed.
My take: I was leaning toward encapsulating business logic within UDF’s. UDF’s provide all the flexibility we need + overloading. All the functions could be consumed by data warehouse team (building summary tables, cubes etc.) and application developers. For power users we could build views by applying the same functions on top of the source data. Thus avoid data duplication. It seemed win-win until this 4000 issue :).
Your thoughts on alternative approaches are appreciated.
Thank you.