Re: [SQLObject] Strange ForeignKey behavior...

On Wed, Dec 23, 2015 at 08:35:10PM +0100, Oleg Broytman <ph...@ph...> wrote:
> Hi!
> On Wed, Dec 23, 2015 at 12:09:27PM -0500, Nathan Edwards <nj...@ge...> wrote:
> > 
> > >> I'll see if I can figure out a minimalist patch to
> > >> propose...
> > > 
> > >    Eagerly waiting! But if you don't succeed notify me -- I'll at least
> > > add a TODO item.
> > 
> > Not sure if this will be up to your coding standards, but here is an
> > attempt - it uses the validator infrastructure to attempt conversion
> > using the idType function of the foreign table.
> > 
> > Patch for col.py and a new test module attached.
> 
>    Thank you! I need to work on it: the test doesn't work with Postgres;
> I'd like to rename 'to_python' to 'from_python' and remove 'from_python'
> assignment (I think there is no need to validate foreign keys when
> they're fetched from the database); and I want to fix style.

   I fixed all that but in required a few major changes.

1. Postgres. Until now SQLObject used to create ForeignKeys of type
idType of the current table, i.e. usually INT. I fixed that (by adding a
lookup of idType of the referenced table), but that a big
backward-incompatible change and I don't want to apply it to bugfix
releases. So I think the patches will go to 2.2.0b2 and 3.0.0a2.

2. 'to_python' => 'from_python': works.

3. Test applied, but I have to create a helper function to setup
mutually referenced classes; before that test framework didn't work with
cyclic references (but see test_cyclic_reference.py).

4. Style mostly fixed. I'll do more fixes when merge the code into
branch 3.0.

> > Cheers!
> > 
> > - n
> > 
> > -- 
> > Dr. Nathan Edwards                      nj...@ge...
> > Department of Biochemistry and Molecular & Cellular Biology
> >             Georgetown University Medical Center
> >                 Room 1217, Harris Building,
> >         3300 Whitehaven St, NW, Washington DC 20007
> >            Phone: 202-687-7042, Fax: 202-687-0057

Oleg.
-- 
     Oleg Broytman            http://phdru.name/            ph...@ph...
           Programmers don't die, they just GOSUB without RETURN.