[RFC, RFA] ndoutils for Oracle integration into CVS

Bernhard Reutner-Fischer rep.dot.nop at gmail.com
Tue Apr 21 16:41:06 CEST 2009


On Mon, Apr 13, 2009 at 01:59:59PM +0200, Hendrik Baecker wrote:
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>Bernhard Reutner-Fischer schrieb:
>> David Schmidt wrote support for using an Oracle DB as backend for ndo2db.
>> 
>> I'd like to ask for assistance to include this Oracle support into the CVS tree.
>> 
>Some time ago I discusesd the oracle thing with David.
>Since he has to trim some of the table names cause of length restriction
>I would see a "change" to the database scheme which, IIRC isn't wanted
>for the moment.

Well yes, but there is no way around the length restriction of table names in
oracle, isn't it?

But you're right, changing ndo2db_db_rawtablenames in general is not a good idea,
I will make that dependant on oracle ASAP (thanks for the reminder!).

>> The code is up to current CVS, a huge part of a direct patch would come from
>> the fact that we changed the directory layout in db/ to cater for
>> per-database directories. We did:
>> $ mv db/queries db/x
>> $ mkdir db/queries
>> $ mv db/x db/queries/mysql
>> and then populated db/queries/oracle with the respective queries.
>> 
>Seperation of the normal behavior for MySQL Clients is a good thing
>until we're missing a overall solution for those feature requests.
>Maintaing the code for Oracle is another one.
>I don't know if Ethan has the time to maintain it.

Maintaining the Oracle parts shouldn't be a big deal, I'm willing to keep it
up to date (since i need it anyway).

>> The rest of the patch can be separated into these hunks:
>> - configury & README: Check for Oracle libs, add section for Oracle to README
>> - liboci/*: Simple OCI (Oracle C Interface or the like) helpers like ora_oci_init
>>   for initializing the OCI env, ora_oci_query(), ora_oci_insert() and the like.
>> - src/*: Handle USE_ORACLE path
>> 
>> Thoughts?
>A huge benefit for the NDOUtils to get access to the 'bigger' databases.
>Did you thought about linking against libdi as a DBI abstraction layer
>for C?

Never heared of it, no. dpkg -l libdi comes back empty. Do you have an URL
at hand?

>That wouldn't solve the lenght restriction of table names but it might
>blow away the great liboci work from David.

As long as it supports parameter binding and is not too much of a pain to
switch over, it's fine with me, sure.

Thanks!

------------------------------------------------------------------------------
Stay on top of everything new and different, both inside and 
around Java (TM) technology - register by April 22, and save
$200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco.
300 plus technical and hands-on sessions. Register today. 
Use priority code J9JMT32. http://p.sf.net/sfu/p




More information about the Developers mailing list