V8.3.0: PQprepare()/PQexecPrepared() and DECLARE CURSOR

From: Sebastien FLAESCH <sf(at)4js(dot)com>
To: pgsql-hackers(at)postgresql(dot)org
Subject: V8.3.0: PQprepare()/PQexecPrepared() and DECLARE CURSOR
Date: 2007-12-18 14:35:44
Message-ID: 4767DAC0.6050408@4js.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi All,

I am new to this mailing list and want to participate to the 8.3.0 beta program.
(Sorry to be late BTW)

My name is Sebastien FLAESCH and I am in charge of the database interfaces at Four J's Development Tools.

Our product is a Informix 4gl compatible compiler / runtime system.

I wrote all the database interfaces to:

- Oracle (OCI),
- DB2 UDB (CLI),
- SQL Server (ODBC and Native Client),
- PostgreSQL (libpq),
- MySQL (libmysqlclient),
- Sybase ASA (dblib*),
- ANTs (ODBC).

Understand it's about a real database driver for our virtual machine (kind of php db or jdbc driver).

We have a large customer base using Informix and some of them have migrated / want to migrate to PostgreSQL.

We support a libpq-based driver for PostgreSQL since version 7, we support currently 8.1, 8.2 and now I am working in the 8.3 driver.

I do use prepared statements with the PQprepare() / PQexecPrepared() API since first version 8 - thanks for that by the way.

Now I want to take benefit of server cursors, using the DECLARE/FETCH/CLOSE instructions.

8.3 also introduced positioned update/deletes (WHERE CURRENT OF), so we do not more need to emulate this with oids.

...

The problem: It appears that the server gets confused when doing PQprepare("DECLARE...) followed by several PQexecPrepared().

Basically I do libpq API calls like this:

For SQL that does not return a result set:

PQprepare(... "cu1", "INSERT INTO ..." );
PQexecPrepared( ... "cu1" ... );
PQexecPrepared( ... "cu1" ... );
PQexecPrepared( ... "cu1" ... );
PQexec( "DEALLOCATE cu1" );

For SQL producing a result set:

PQprepare(... "cu1", "DECLARE cu1 CURSOR WITH HOLD ..." );
PQexecPrepared( ... "cu1" ... ); -- opens the cursor...
PQexec( "FETCH NEXT FROM cu1 ..." ... );
PQexec( "FETCH NEXT FROM cu1 ..." ... );
PQexec( "FETCH NEXT FROM cu1 ..." ... );
PQexec( "CLOSE cu1" ); -- frees cursor resources (need to re-execute DECLARE)
PQexecPrepared( ... "cu1" ... ); -- opens the cursor...
PQexec( "FETCH NEXT FROM cu1 ..." ... );
PQexec( "FETCH NEXT FROM cu1 ..." ... );
PQexec( "CLOSE cu1" ); -- frees cursor resources (need to re-execute DECLARE)
PQexecPrepared( ... "cu1" ... ); -- Here I get error: [42P03][cursor "cu1" already exists]

I wonder why the second PQexecPrepare() executes and the third fails...

To make this work, I need to de-allocate the statement and re-prepare with PQprepare() ...

I will try to provide you with a little sample to reproduce, but wanted to post this early to let you known.

Best regards,
Sebastien FLAESCH

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Andrew Dunstan 2007-12-18 15:00:29 Re: V8.3.0: PQprepare()/PQexecPrepared() and DECLARE CURSOR
Previous Message peter.trautmeier 2007-12-18 13:48:23 Re: ecxt_scantuple has wrong TupleDesc