It looks like you and Fedor had similar ideas.  

I wonder if maybe we should do something like this:

Keep the current classes for backwards compatibility, but start reworking these 
classes to start using the crossdb framework.  It looks like these classes could use 
crossdb to get the same results.  

At the same time, start deprecating these things and start moving towards crossdb 
usage which should do away with most of the limitations and problems that I've been 
reading about with the current Criterion.  

Travis

---- Original Message ----
From: Eric Dobbs <[EMAIL PROTECTED]>
Sent: 2002-04-24
To: Turbine Torque Developers List <[EMAIL PROTECTED]>
Subject: Re: Torque subproject - crossdb

Welcome!

On Wednesday, April 24, 2002, at 10:37  AM, [EMAIL PROTECTED] 
wrote:

> Just to fill you in on what this is about, I submitted a proposal to 
> the jakarta general list and you can read the discussions here:
> http://www.mail-archive.com/[email protected]/
>
> But basically, I am wanting to bring the crossdb project to Jakarta and 
> making it a subproject of Torque seems to be a good idea because of the 
> relations.
>
> So I don't know how this works, but I guess it would be up to you guys?

I read some of the discussion on general yesterday and
briefly browsed through the code.  About a year ago
Fedor and I discussed at length the need for a new
query model, but we never settled our disagreements and
both dropped the issue.  crossdb looks to be similar in
many respects to various things Fedor and I proposed in
our debate.

Below are links to Fedor's and my code and links into
the debate.  The code and discussion might be worth a
little of your attention so you can see what we've
discussed before.  I wouldn't suggest you read all of
it (that would probably be pretty boring 8^).

Fedor's proposed code is here:
<http://cvs.apache.org/viewcvs/jakarta-turbine-
torque/proposals/fedor/nqm/>

And mine is here:
<http://cvs.apache.org/viewcvs/jakarta-turbine-
torque/proposals/eric/statement/>

Discussion about the options...
The last option I presented is here:
<http://www.mail-archive.com/turbine-
dev%40jakarta.apache.org/msg01147.html>

And this search should hit most of our debate:
<http://www.mail-archive.com/cgi-
bin/htsearch?config=turbine-
dev_jakarta_apache_org&words=fedor+eric+new+query+model>

Your code has an immediate advantage because it's in
use whereas Fedor's and my code has collected lots of
dust.

And I just noticed that Jeff Schnitzer has suggested on
turbine-torque-user an api from RogueWave as a model:
<http://www.roguewave.fr/support/docs/dbtug/4-5.cfm#451>

The fact that RogueWave has been making some money on
an api for abstracting sql generation lends support to
our argument on general that there's value in such a
project.

Here's a jumpstart into the Torque objects that deal
with SQL generation:

org.apache.torque.util.BasePeer
org.apache.torque.util.SqlExpression
org.apache.torque.util.Criteria
org.apache.torque.adaptor.DB*

And village is also used in BasePeer:
http://share.whichever.com/village

I think crossdb fits somewhere in there.  But I haven't
given it enough attention just yet to know exactly
where.

BasePeer and Criteria are used extensively by Torque
applications.  So those APIs need to be preserved.  I
mostly loathe the Criteria object, but there's no quick
escape.  I haven't actually looked recently, but I
suspect that SqlExpression and the adaptors could be
replaced.  Dan has cautioned in the past that the
database adaptors serve other purposes besides handling
the dialects of different databases.  So those uses
would need to be factored out of the adaptors before
replacing them.

-Eric

--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>



--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to