Re: [SQLObject] Looking towards 0.5

[Ian B. <ia...@co...>]

On Thu, 2003-05-22 at 10:18, Luke Opperman wrote:
> My thoughts for post-0.4:
> 
> * MemoryTransaction or an equivalent.

Sure.

> * Some decision on concurrency control/resolution (should SQLObject offer any,
> and if so then let's do it.)

If we can come to decisions ;)  And implementations, of course.

> * Actual enforcement of a constraints/validators system, at least providing
> toSQL/toPython functionality. (Ian, I know you've been busy on FormEncode, and
> that will be a help here :)

Yes, that's distracted me from SQLObject stuff recently.


I'd also like to do some cleanup of names:

sqlRepr -> __sqlrepr__
*Connection -> *Store
Maybe others.  _columns -> __columns__ ?
I don't like the _setting style, but maybe lots of __setting__ names
just get excessive, and they shouldn't be public (i.e., no
underscores).  The old names would still work for a version or two, with
warnings being printed maybe in 0.6.  sqlRepr might be more quickly
deprecated, but none of the others.

The constraints/conversion stuff is getting closer.  I want to do one
more renaming on Validator, and change the class hierarchy slightly (so
there's an option for a more trimmed-down validators), but that's it. 
So it should be able to go in fine.

Joins will be redone, definitely.  They'll be put in their own module,
like Col, and a Join/SOJoin distinction will be made.  That should make
the code significantly cleaner -- not so much shorter, but it will be
clear what code gets invoked at what point of the process.  I'll
probably add a one-to-one join, and maybe a more generic join for
annotated relations (like the Worker-Customer + rate example).  Maybe
implement orderBy in the database too.  Open to consideration is join
results becomimg iterators, ala select results...

I think that's all that's been on my mind, at least short-term. 
FormEncode will start to become more involved at some point, but the
coupling will be loose -- I'm going to try to create an interface for
publishing objects (and classes) via FormEncode, and SQLObject should be
able to implement that fairly trivially (once I figure out what "that"
is).

  Ian