Re: INSERT and .refresh() problems

On Mon, Jun 27, 2005 at 03:20:02AM +0200, Faber wrote:
> In previous version of PyDO2 that would work, but in the latest I receive 
> following exception on a.refresh() call:
> 
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
>   File "pydo/base.py", line 626, in refresh
>     raise ValueError, "current object doesn't exist in database!"
> ValueError: current object doesn't exist in database!
> 
> Should I do a Relation.commit() before the a.refresh() call or it's a bug 
> and I can continue to use this code, waiting for the fix?  ;)

I'm trying to reproduce this bug, but so far, it I'm not able to
(using the svn head, which I don't think is different from the beta in
this respect).  I just checked in a test case to tests/test_base.py,
which you can test there if you are able to follow along in
subversion, that tries to do what you are doing:


class test_refresh2(base_fixture):
    usetables=('A', 'B')
    tags=alltags

    def pre(self):
        self.B.new(x=4)

    def run(self):
        obj=self.B.getUnique(id=1)
        obj.refresh()
        assert obj.id==1
        assert obj.x==4
        n=self.A.new(name='ho hum',
                     b_id=obj.id,
                     ta=0,
                     x=0,
                     y=0,
                     z=0)

        assert n.id==1
    
The foreign key relationship is actually irrelevant, since presumably
you are getting the failure at the refresh itself.  But I'm not; this
test passes for me. I'm also using psycopg:

  $ python -c 'import psycopg;print psycopg.__version__'
  2.0b3 (dt ext pq3)

So your failure is probably still a PyDO bug, but not the one I'm
testing for -- or our environments are behaving differently.  If this
test works for you but you still have an application error, could you
try turning into a test case?  (Naked unittest greed.)

One other thing, btw -- you do not need to refresh to get an
autoincrement or sequence value.  With sequences, PyDO fetches the
sequence first and explicitly inserts it; with autoincrement, the
autoincrement value is found during the new operation.  If this were
not the case, refresh would be impossible when the unique constraint
is a sequence/autoincrement field.

Cheers,

js

-- 
Jacob Smullyan

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCv15auqamFyFXXLIRAm9GAJ9a77wVwKEdc05JMBFZS6nSWOeTzwCgyCUQ
RS5VILeY2gA/aJN5xaSjesA=
=l6Ub
-----END PGP SIGNATURE-----