OLE DB and Oracle

Question

I am presently struggling to connect a Powerbuilder App to an Oracle Database, details:
Development environment
Win2000 latest SP etc
OracleXE 10g (v10.?) - single user dev system
Powerbuilder 10.2 Enterprise
Using the Powerbuilder driver (direct connection not OLE DB) it connects perfectly and all looks OK, but
Using the MS or Oracle supplied OLE DB driver I can connect, read data from and save data to the database but I have a problem running Stored Procedures/Functions. I can run simple Stored Procedures/Functions but if they contain any SQL then they will not execute. With the PB direct driver they work perfectly.
On my Client's installation, which is:
Win XP
Oracle 10.2 - part of a very large installation, that uses TNS rather then specifying a server ip address which is what I do on my setup here. But one of there guys was doing the testing with me and he knew all about it.
The direct driver would not connect at all - could be versioning, could be something missing - for example I have found docs that say we need SQL*Net????
The Oracle supplied OLE DB driver would not connect at all.
The MS supplied OLE DB driver connected fine but I could not test my software with it. When I say connected, what I mean is that the Test Connection button on the UDL editor reports that it connects OK.
My preferred mechanism would now be to use OLE DB, it is a proprietary piece of software and getting the latest (and working versions) should be easier than getting the required DLL from Sybase. if my client changes the version of software on the server we have been assigned to then all that is needed is the correct OLE DB driver. I have used OLE DB between Powerbuilder and MS SQL Server for a number of years and so I am used to it's quirks.
But using either of the OLE DB drivers (on my development system) gives me the strange Stored Procedure behaviour, described above. I can work around this - it's main use being to service various PKs which are required to be auto-incrementing. Using transaction control will give me the same functionality, storing the current index of a procedure in a parameter table, instead of using Oracle sequences. But if I am getting one problem, I am afraid that I might get more, why should the issue be just with stored funcs/procs? Is there some sort of SET statement I need, I have tried SET TRANSACTION ..... but no joy?
Any help would be greatly appreciated.
Bill Mac

Answer 1

Hi,
assuming you called the function from a sql select statement (select gf_test from dual;) then the error is as expected, you can't do DML inside a select. Parallel DML is unlikely to be your problem.
I'm a little unclear what you are trying to do with this function though, you could hack it by declaring this an autonomous transaction, but it's unlikely that that is the best way to solve the business problem. If you are trying to find the value of a primary key that is generated from a sequence then the normal way of doing this is to have a before insert trigger that generates the pk for you and then have a pl/sql procedure to do the insert and return the pk to the calling application (which can be powerbuilder or any other language that can call pl/sql.
Niall
Bill Mac wrote:
OK I have managed to get some code to return an error number and this is it:
ORA-14551: cannot perform a DML operation inside a query
Cause: DML operation like insert, update, delete or select-for-update cannot be performed inside a query or under a PDML slave.
Action: Ensure that the offending DML operation is not performed or use an autonomous transaction to perform the DML operation within the query or PDML slave.
The code was:
create or replace function GF_TEST return NUMBER is
retval number;
begin
SET TRANSACTION READ WRITE;
UPDATE parameters
SET last_parent_co = last_parent_co + 1
WHERE (row_no = 1)
RETURNING last_parent_co INTO retval;
COMMIT WORK;
retval := 1001;
return retval;
exception when others then
return sqlcode;
end;
Does that now make any sense to any one? Am I a pdml slave?
Edited by: Bill Mac on 07-Apr-2009 07:38