Generic cursors and generic result sets - How to?

Question

Hi all,
I'm currently developing some kind of table export util. All it should do is to take a list of table names, determine the table structure (aka fields), fetch all data from this table and store a flat textfile containing all table data as ready-to-use INSERT statements.
Please do not tell me that I could use DataPump or imp/exp, I know this but in this specific project this is not possible. I simply need one textfile per Table, that's it.
So far my approach is the following:
- Developed a package which contains a hard-coded list of tables as "TABLE OF VARCHAR2".
- One procedure inside this package determines the list of fields for every passed tablename and their corresponding fieldtypes and generates a SELECT and INSERT query.
- Another procedure will get the created SELECT query (which is valid, I've tested this) and performs a "OPEN <generic_cursor> FOR <select_query>";
And here is also my problem. After "OPEN" the generic cursor I need a record to fetch the data row into. However as the list of tables is dynamic, so must be the records where to fetch into. I did not find any possible way of dynamically creating a record for the "FETCH INTO" statement. So far the only solution I've found was to hardcode the results of the tables and use IF/ELSE or CASE logic inside the data retrival procedure to fetch always into the right record:
PROCEDURE DUMP_TABLES(P_TABLENAME VARCHAR2, P_SELECT VARCHAR2)
G_CV genericcurtyp;
T_MC_USERS MC_USERS%ROWTYPE;
T_CDALBUM CDALBUM%ROWTYPE;
BEGIN
OPEN G_CV FOR P_SELECT;
LOOP
IF P_TABLENAME = 'MC_USERS' THEN
FETCH G_CV INTO T_MC_USERS;
ELSE
FETCH G_CV INTO T_CDALBUM;
END IF;
EXIT WHEN G_CV%NOTFOUND;
END LOOP;
END;
What I really would like to use is something like this:
PROCEDURE DUMP_TABLES(P_TABLENAME VARCHAR2, P_SELECT VARCHAR2)
G_CV genericcurtyp;
TYPNAME VARCHAR2(32):='T_'||P_TABLENAME;
MYRES TYPNAME%ROWTYPE;
BEGIN
OPEN G_CV FOR P_SELECT;
LOOP
FETCH G_CV INTO MYRES;
EXIT WHEN G_CV%NOTFOUND;
-- Do something with the row data in MYRES
END LOOP;
CLOSE G_CV;
END;
Of course I could hard-code everything, however I would like to keep the possibility to extend the list of tables simply by adding another entry to my list of tablenames without having to add additional code whenever the table list changes.
I've tried also the "FOR i IN (<SELECT>)", however this won't accept a variable for the <SELECT>.
Maybe someone here as an idea how to solve this problem without having to hard-code the required result sets of all affected tables? Or maybe I'm simply going the wrong way and someone could show me the right one?
Any help would be appreciated.
Best regards,
Sascha

Answer 1

You could use DBMS_SQL instead. DBMS_SQL provides DESCRIBE procedure so you can get count and datatypes of select list elements. If you are on 11g you can convert ref cursor to DBMS_SQL cursor or vice versa. But I would use a different solution. First of all, what are you creating "export" for? If you are planning to use it to load data into another Oracle database I'd simply use external tables. If you really need CSV, I'd use DBMS_SCHEDULER and run SQL*Plus task spooling output to a file.
SY.