Dynamically spilt rows in to mulitple comma-separated lists - sql

Question

I have a table that contains a list of users.
USER_TABLE
USER_ID DEPT
------- ----
USER1 HR
USER2 FINANCE
USER3 IT`
Using a SQL statement, I need to get the list of users as a delimited string returned as a varchar2 - this is the only datatype I can use as dictated by the application I'm using, e.g.
USER1, USER2, USER3
The issue I have is the list will exceed 4000 characters. I have the following which will manually chunk up the users in to lists of 150 users at a time (based on user_id max size being 20 characters plus delimiters safely fitting in to 4000 characters).
SELECT LISTAGG(USER_ID, ',') WITHIN GROUP (ORDER BY USER_ID)
FROM (SELECT DISTINCT USER_ID AS USER_ID, ROW_NUMBER() OVER (ORDER BY USER_ID) RN FROM TABLE_NAME)
WHERE RN <= 150
START WITH RN = 1
CONNECT BY PRIOR RN = RN - 1
UNION
SELECT LISTAGG(USER_ID, ',') WITHIN GROUP (ORDER BY USER_ID)
FROM (SELECT DISTINCT USER_ID AS USER_ID, ROW_NUMBER() OVER (ORDER BY USER_ID) RN FROM TABLE_NAME)
WHERE RN > 150 AND RN <= 300
START WITH RN = 1
CONNECT BY PRIOR RN = RN - 1
This is manual and would require an additional UNION for each chunk of 150 users and the total number of users could increase at a later date.
Is it possible to do this so the delimited strings of user_ids are generated dynamically so they fit in to multiple chunks of 4000 characters and no user_ids are split over multiple strings?
Ideally, I'd want the output to look like this:
USER1, USER2, USER3 (to) USER149
USER150, USER151, USER152 (to) USER300
USER301, USER302, USER303 (to) USER450`
The solution needs to be a SELECT statement as the schema is read-only and we aren't able to create any objects on the database. We're using Oracle 11g.

Answer 1

You can do this with a pipelined function:
create or replace function get_user_ids
return sys.dbms_debug_vc2coll pipelined
is
rv varchar2(4000) := null;
begin
for r in ( select user_id, length(user_id) as lng
from user_table
order by user_id )
loop
if length(rv) + r.lng + 1 > 4000
then
rv := rtrim(rv, ','); -- remove trailing comma
pipe row (rv);
rv := null;
end if;
rv := rv || r.user_id || ',';
end loop;
return;
end;
/
You would call it like this:
select column_value as user_id_csv
from table(get_user_ids);

Answer 2

Alternate way using below function :
create or replace FUNCTION my_agg_user
RETURN CLOB IS
l_string CLOB;
TYPE t_bulk_collect_test_tab IS TABLE OF VARCHAR2(4000);
l_tab t_bulk_collect_test_tab;
CURSOR user_list IS
SELECT USER_ID
FROM USER_TABLE ;
BEGIN
OPEN user_list;
LOOP
FETCH user_list
BULK COLLECT INTO l_tab LIMIT 1000;
FOR indx IN 1 .. l_tab.COUNT
LOOP
l_string := l_string || l_tab(indx);
l_string := l_string || ',';
END LOOP;
EXIT WHEN user_list%NOTFOUND;
END LOOP;
CLOSE user_list;
RETURN l_string;
END my_agg_user;
After function created ,
select my_agg_user from dual;

Answer 3

I believe the SQL I have below should work in most cases. I've hard-coded the SQL to break the strings up in to 150 entries of user id, but the rest is dynamic.
The middle part produces duplicates, which requires an additional distinct to eliminate, but I'm not sure if there is a better way to do this.
WITH POSITION AS ( SELECT ((LEVEL-1) * 150 + 1) FROM_POS, LEVEL * 150 TO_POS
FROM DUAL
CONNECT BY LEVEL <= (SELECT COUNT(DISTINCT( USER_ID)) / 150 FROM TABLE_NAME)
)
SELECT DISTINCT
LISTAGG(USER_ID, ',') WITHIN GROUP (ORDER BY USER_ID) OVER (PARTITION BY FROM_POS, TO_POS)
FROM
(SELECT DISTINCT USER_ID AS USER_ID, ROW_NUMBER() OVER (ORDER BY USER_ID) RN FROM TABLE_NAME) V0 ,
POSITION
WHERE V0.RN >= POSITION.FROM_POS
AND V0.RN <= POSITION.TO_POS