sqlobject-discuss

[SQLObject] Is this a bug?

[Iwan V. <iw...@re...>]

Hi there,

I am on sqlobject 0.9.1.

This looks like a bug - correct me if I'm wrong.  Transaction.close()
does not close the underlying connection. Seems like the __getattr__ of
Transaction is to blame if I understand correctly.  Here is some code
showing the problem (fill in the xxx es):

import sqlobject

connectionURI = 'postgres://xxx:xxx@localhost/xxx'
conn = sqlobject.connectionForURI(connectionURI).transaction()


conn.close()

assert conn._connection.closed, 'The underlying connection is open'

If you follow this with:
conn._connection.close()

Then things seems to work.

Regards
- Iwan

Re: [SQLObject] Is this a bug?

[Oleg B. <ph...@ph...>]

On Mon, Mar 30, 2009 at 04:16:16PM +0200, Iwan Vosloo wrote:
> This looks like a bug - correct me if I'm wrong.  Transaction.close()
> does not close the underlying connection.

   You can do it as tnx.commit(close=True)

Oleg.
-- 
     Oleg Broytmann            http://phd.pp.ru/            ph...@ph...
           Programmers don't die, they just GOSUB without RETURN.

Re: [SQLObject] Is this a bug?

[Oleg B. <ph...@ph...>]

On Mon, Mar 30, 2009 at 04:16:16PM +0200, Iwan Vosloo wrote:
> This looks like a bug - correct me if I'm wrong.  Transaction.close()
> does not close the underlying connection. Seems like the __getattr__ of
> Transaction is to blame if I understand correctly.

   I think .close() on a Transaction must be forbidden. I am going to add a
.close() method to Transaction that raises an exception saying "Do not call
close() - call either commit(), commit(close=True) or rollback()". Ok?

Oleg.
-- 
     Oleg Broytmann            http://phd.pp.ru/            ph...@ph...
           Programmers don't die, they just GOSUB without RETURN.

Re: [SQLObject] Is this a bug?

[Iwan V. <iw...@re...>]

On Wed, 2009-04-08 at 00:40 +0400, Oleg Broytmann wrote:
> On Mon, Mar 30, 2009 at 04:16:16PM +0200, Iwan Vosloo wrote:
> > This looks like a bug - correct me if I'm wrong.  Transaction.close()
> > does not close the underlying connection. Seems like the __getattr__ of
> > Transaction is to blame if I understand correctly.
> 
>    I think .close() on a Transaction must be forbidden. I am going to add a
> .close() method to Transaction that raises an exception saying "Do not call
> close() - call either commit(), commit(close=True) or rollback()". Ok?

Sounds good to me Oleg.  

Only thing is that if one is familiar with the DBAPI, one intuitively
expects a sqlobject Transaction so work like a DBAPI Connection (or is
it just me?).  If so, you sortof expect it to have .commit and .close
methods.

But having it complain with a nice error message like that make the
difference with the DBAPI explicit.  Which is also good.

Thanks
-i

Re: [SQLObject] Is this a bug?

[Oleg B. <ph...@ph...>]

On Mon, Apr 13, 2009 at 09:51:47AM +0200, Iwan Vosloo wrote:
> On Wed, 2009-04-08 at 00:40 +0400, Oleg Broytmann wrote:
> > On Mon, Mar 30, 2009 at 04:16:16PM +0200, Iwan Vosloo wrote:
> > > This looks like a bug - correct me if I'm wrong.  Transaction.close()
> > > does not close the underlying connection. Seems like the __getattr__ of
> > > Transaction is to blame if I understand correctly.
> > 
> >    I think .close() on a Transaction must be forbidden. I am going to add a
> > .close() method to Transaction that raises an exception saying "Do not call
> > close() - call either commit(), commit(close=True) or rollback()". Ok?
> 
> Sounds good to me Oleg.  
> 
> Only thing is that if one is familiar with the DBAPI, one intuitively
> expects a sqlobject Transaction so work like a DBAPI Connection (or is
> it just me?).  If so, you sortof expect it to have .commit and .close
> methods.
> 
> But having it complain with a nice error message like that make the
> difference with the DBAPI explicit.  Which is also good.

   Calling .close() on a transaction instance means the user has messed up
connections and transaction instances so I decided the exception is
TypeError. Revisions 3856-3858. Thank you!

Oleg.
-- 
     Oleg Broytmann            http://phd.pp.ru/            ph...@ph...
           Programmers don't die, they just GOSUB without RETURN.