How to append rows to a existing SYS_REFCURSOR? - oracle

Question

I want to know if there is a way append to the results of a cursor that is fetched from inside a loop.
Right now out_cursor contains the results pertaining to the last iteration of the outer for loop. I want to know if it will be possible to append the rows from each iteration to this cursor so that the cursor will contain rows from all iterations of the loop.
For context, this out_cursor is consumed by a Java DAO class.
CREATE OR REPLACE NONEDITIONABLE PROCEDURE testproc (out_cursor OUT SYS_REFCURSOR)
IS
CURSOR actor_cursor IS
SELECT actor_id, first_name, last_name
FROM actor
WHERE actor_id <= 100;
row1 actor_cursor%rowtype;
TYPE MyRec IS RECORD (actor_id film_actor.actor_id%TYPE, film_id film_actor.film_id%TYPE);
rec MyRec;
BEGIN
FOR row1 IN actor_cursor
LOOP
dbms_output.put_line('actor_id:'||row1.actor_id ||' ---- '|| row1.first_name || row1.last_name);
OPEN out_cursor FOR
SELECT actor_id, film_id
FROM film_actor
WHERE actor_id = row1.actor_id;
END LOOP;
LOOP -- this loop is here just to print out_cursor for testing
FETCH out_cursor INTO rec;
EXIT WHEN out_cursor%NOTFOUND;
dbms_output.put_line('actor_id:'||rec.actor_id||', film_id:'||rec.film_id);
END LOOP;
END;
The output of the script is something like this
actor_id:98 ---- CHRIS BRIDGES
actor_id:99 ---- JIM MOSTEL
actor_id:100 ---- SPENCER DEPP
actor_id:100, film_id:17
actor_id:100, film_id:118
I understand that only rows from the last iteration of the for loop are there in the out_cursor.
Is there any way to make out_cursor return the results from all the iterations of the outer FOR LOOP, essentially aggregate the results of all the iterations. To produce result like this?
actor_id:98 ---- CHRIS BRIDGES
actor_id:99 ---- JIM MOSTEL
actor_id:100 ---- SPENCER DEPP
actor_id:98, film_id:77 // results from iteration #1
actor_id:98, film_id:43
actor_id:99, film_id:67 // results from iteration #2
actor_id:99, film_id:90
actor_id:100, film_id:17 // results from iteration #3
actor_id:100, film_id:118
I am aware that I can easily archive the same results using JOINS. But I am not allowed to modify the SQL queries (actual SQLs are very complex) - I can only run them and use the results produced from one SQL as parameters (in SELECT or WHERE clause) for the next SQL.
(UPDATE)
The General requirement is - there are a sequence of queries, resultset of each is used in the constraints of the next one. I cannot modify the queries themselves - so using a join or sub-query is out of question.
I was hoping to open a cursor for each query and then iterate through the values in the cursor - using the cursor attributes in the constraint of the next query, for which I will open a cursor inside the for loop. Exactly as shown in the sample. But the problem was the cursor inside the loop only contained the selections made in the last iteration - during each iteration a new cursor was getting created. So, I am getting only a subset of the results I wanted in the out_cursor.
I think i found a way to do this -
CREATE OR REPLACE PACKAGE PKG_BIDS_REPORTS_TEST as
TYPE Q2DATA IS RECORD (
actor_id film_actor.actor_id%TYPE,
film_id film_actor.film_id%TYPE
);
TYPE Q2DATA_TAB IS TABLE OF Q2DATA INDEX BY BINARY_INTEGER;
Q2DATA_REC Q2DATA_TAB;
PROCEDURE pqr_reports_test(l_out_data OUT sys_refcursor);
end PKG_BIDS_REPORTS_TEST;
/
create or replace NONEDITIONABLE PACKAGE BODY PKG_BIDS_REPORTS_TEST as
PROCEDURE pqr_reports_test(l_out_data OUT sys_refcursor) AS
CURSOR actor_cursor IS
SELECT actor_id, first_name, last_name
FROM actor
WHERE actor_id <= 100;
row1 actor_cursor%rowtype;
CURSOR film_actor_cursor(actorid film_actor.actor_id%TYPE) IS
SELECT actor_id, film_id
FROM film_actor
WHERE actor_id = actorid;
row2 film_actor_cursor%rowtype;
V_CNT NUMBER:=0;
BEGIN
V_CNT := q2data_rec.COUNT;
FOR row1 IN actor_cursor
LOOP
--dbms_output.put_line(row1.actor_id ||' ---- '|| row1.first_name ||' ---- '|| row1.last_name);
FOR row2 IN film_actor_cursor(row1.actor_id)
LOOP
V_CNT := V_CNT + 1;
q2data_rec(V_CNT).actor_id := row2.actor_id;
q2data_rec(V_CNT).film_id := row2.film_id;
--dbms_output.put_line(row2.actor_id||','||row2.film_id);
END LOOP;
END LOOP;
OPEN L_OUT_DATA FOR SELECT DISTINCT datarec.actor_id, datarec.film_id from TABLE(q2data_rec) datarec;
END;
End PKG_BIDS_REPORTS_TEST;
Essentially create a PLSQL table and fill it in the loop with a counter like a Java array. It works. Not the most elegant, definitely.
So, I would definitely appreciate any optimizations or alternates.

Answer 1

In order to get the output you are expecting for, you cannot close the first cursor before open the second one. It is the only way to have both iterations combined.
Although it would be inefficient, a way would be
BEGIN
FOR row1 IN actor_cursor
LOOP
dbms_output.put_line(row1.actor_id ||' ---- '|| row1.first_name || row1.last_name);
OPEN out_cursor FOR
SELECT actor_id, film_id
FROM film_actor
WHERE actor_id = row1.actor_id;
END LOOP;
FOR row1 IN actor_cursor
LOOP
OPEN out_cursor FOR
SELECT actor_id, film_id
FROM film_actor
WHERE actor_id = row1.actor_id;
LOOP
FETCH out_cursor INTO rec;
EXIT WHEN out_cursor%NOTFOUND;
dbms_output.put_line(row1.actor_id ||','|| rec.film_id);
END LOOP;
END LOOP;
END;

Answer 2

As I indicated before you cannot combine the result of separate cursors. However, you can (at least in this case) combine them into a single select and return that result. And since you are returning a reference cursor you cannot loop through it unless you close and re-open it. So in this case your procedure should consist of a single open cursor statement.
create or replace
procedure testproc (out_cursor out sys_refcursor)
is
begin
open out_cursor for
select a.actor_id, a.first_name, a.last_name, fa.film_id
from actor a
join film_actor fa
on fa.actor_id = a.actor_id
where a.actor_id <= 100;
end testproc ;
------- test ------
declare
actor_id actor.actor_id%type
fname actor.first_name%type
l_name actor,last_name%type
film_id film.film_id%type;
rec sys_refcursor ;
begin
testproc(rec);
loop
fetch rec
into actor_id, fname, lname, film_id;
exit when rec%notfound;
dbms_output.put_line('actor_id:' || actor_id ||
' name: ' || fname || ' ' || lname
' film_id:' || film_id
);
end loop;
end ;