Executing Dynamic Native SQL with Oracle Table type gives invalid identifier error - sql

Question

We have created Oracle Table type. And we have created an array of it. This is done as we dont know how many values can come which may be too much for a IN clause in sql query.
--Code Snippet -----
create or replace
TYPE "INPUTCODE"
as object
(pc varchar2(100) )
create or replace
TYPE "INPUTCODEARR"
IS TABLE OF inputcode;
create or replace
PROCEDURE "TEST_PROC" (testCodes IN inputcodeArr, timeHorizon IN NUMBER, p_recordset OUT SYS_REFCURSOR)
AS
var_sqlStmt VARCHAR2(4096);
BEGIN
var_sqlStmt := 'select t.a,t.b, t.c';
var_sqlStmt := var_sqlStmt || 'from test t';
if testCodes is not null then
var_sqlStmt := var_sqlStmt || ', table(testCodes) tc';
var_sqlStmt := var_sqlStmt || 'where tc.pc = t.name';
end if;
dbms_output.put_line('Final SQL Statement::' || var_sqlStmt);
open p_recordset for var_sqlStmt;
END TEST_PROC;
All the above ones are compiles and when you run the TEST_PROC procedure with few testCode values it will fail with error Invalid identifier for testCodes.
In the procedure, final sql statement which is printing in my console is correct and when you run this as a static sql statement inside procedure it runs without any error. But inside the dynamic sql it fails.
I tried executing using DYNAMIC_SQL package, but it results in same error.
Also, i tried giving it as a bind variable for 'table(testCodes)'. That also failed.
Please suggest.

Answer 1

You are using dynamic SQL, so you must tell Oracle which word is an identifier and which word is a variable.
Consider the following statement running directly in SQLPlus:
select t.a,t.b, t.c from test t, table(testCodes) tc
It will fail because no object is named testCodes in your DB. You have to tell the SQL engine that testCodes is in fact a variable. You have to do this because you have chosen to use dynamic SQL whereas variable binding is automatic in static SQL.
In most cases, you can bind "object" variables in the same way as standard variables. In PL/SQL there are several ways to do this, for instance with cursors you would use USING:
SQL> DECLARE
2 l_cur SYS_REFCURSOR;
3 l_tab inputcodeArr := inputcodeArr(INPUTCODE('A'), INPUTCODE('B'));
4 l_obj varchar2(100);
5 BEGIN
6 OPEN l_cur FOR 'SELECT pc FROM TABLE(:my_variable)' -- notice the ":"
7 USING l_tab; -- binding by position
8 LOOP
9 FETCH l_cur
10 INTO l_obj;
11 EXIT WHEN l_cur%NOTFOUND;
12 dbms_output.put_line(l_obj);
13 END LOOP;
14 CLOSE l_cur;
15 END;
16 /
A
B
PL/SQL procedure successfully completed
In your case however I wouldn't bother with dynamic SQL since you can open a cursor conditionally:
CREATE OR REPLACE PROCEDURE "TEST_PROC"(testCodes IN inputcodeArr,
timeHorizon IN NUMBER,
p_recordset OUT SYS_REFCURSOR) AS
BEGIN
IF testCodes IS NOT NULL THEN
OPEN p_recordset FOR
SELECT t.a, t.b, t.c FROM test t, TABLE(testCodes) tc
WHERE tc.pc = t.NAME;
ELSE
OPEN p_recordset FOR
SELECT t.a, t.b, t.c FROM test t;
END IF;
END TEST_PROC;
My advice would be to stick with static SQL as long as possible, as it's a lot easier to make mistakes with dynamic SQL.
Update following comment:
If your number of input is not constant and you have to use dynamic SQL because there are many combinations of filters, you can use the following strategy:
CREATE OR REPLACE PROCEDURE "TEST_PROC"(testCodes IN inputcodeArr,
timeHorizon IN NUMBER,
p_recordset OUT SYS_REFCURSOR) AS
l_sql LONG := 'SELECT t.a, t.b, t.c FROM test t WHERE';
BEGIN
-- filter #1
IF testCodes IS NOT NULL THEN
l_sql := l_sql || ' t.name IN (SELECT pc FROM TABLE(:filter1))';
ELSE
l_sql := l_sql || ' :filter1 IS NULL';
END IF;
-- filter #2
IF timeHorizon IS NOT NULL THEN
l_sql := l_sql || ' AND t.horizon = :filter2';
ELSE
l_sql := l_sql || ' AND :filter2 IS NULL';
END IF;
-- open cursor
OPEN p_recordset FOR l_sql USING testCodes, timeHorizon;
END TEST_PROC;
/
I'm making sure that the final SQL will always have the same number of variables in the same order, however each condition where the filter is NULL will be a tautology (NULL IS NULL).