InheritableSQLObject: Data corruption when child record create fails

Hi all,

I am doing some tests with InheritableSQLobject and have run into a problem 
which I hope to get some feedback on.

The problem is that when creating a new object based on an 
InheritableSQLObject class, and the creation of the child object fails after 
a new parent was created, the parent record will remain in the database 
without the relevant child.

Here is a short example:

~~~~~~~~~~ file: inheritTest3.py ~~~~~~~~~~~~
from sqlobject import *
from sqlobject.inheritance import InheritableSQLObject

__connection__ = "sqlite:/:memory:"

class Vehicle(InheritableSQLObject):
    vid = StringCol(notNone=True, default=None, unique=True)
    wheels = IntCol()

class Bicycle(Vehicle):
    seats = IntCol(notNone=True, default=None)

class Car(Vehicle):
    doors = IntCol(notNone=True, default=None)

def createTables():
    Vehicle.createTable()

Re: InheritableSQLObject: Data corruption when child record create fails

Hi!

On Mon, Dec 07, 2009 at 06:41:54PM +0200, Tom Coetser wrote:
> The problem is that when creating a new object based on an 
> InheritableSQLObject class, and the creation of the child object fails after 
> a new parent was created, the parent record will remain in the database 
> without the relevant child.

   A valid concern, sure. I never stumbled upon the problem because I
always use transactions, and on any exception I just ABORT the transaction;
on ABORT the DB removes the half created parent.

> I'm attaching a patch as a possible fix for this

   Thank you!

> -        super(InheritableSQLObject, self)._create(id, **kw)
> +        # TC: Create this record and catch all exceptions in order to destroy
> +        # TC: the parent if the child can not be created.
> +        try:
> +            super(InheritableSQLObject, self)._create(id, **kw)
> +        except:
> +            # TC: If we are the child, destroy the parent
> +            if self.sqlmeta.parentClass:
> +                self._parent.destroySelf()
> +                self._parent = None
> +            # TC: Reraise the original exception
> +            raise
> +

Changing validator messages

Hi All,

I am using the field validator option to auto validate certain fields,
and return more meaningful error messages than the default ones returned
by formencode. It seems that there is a problem with changing some of
the messages though.

Given this sample file :

~~~~~~~~~~~~~ validatorTest.py ~~~~~~~~~~~~~~~~~~~~
from sqlobject import *
from formencode import validators, Invalid

__connection__ = "sqlite:/:memory:"

validName = validators.String(not_empty=True, min=5,
                     messages={'empty': 'The name may not be empty.',
                               'noneType': 'The name is None',
                               'tooShort': 'The name is too short',
                              }
                             )

class foo(SQLObject):
    name = StringCol(validator=validName)

foo.createTable()

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

When passing a value of None for 'name' when creating a new foo, this is

Re: Changing validator messages

On Thu, Dec 10, 2009 at 11:18:33PM +0200, Tom Coetser wrote:
> >>> vt.foo(name=None)                                                 
>     raise Invalid(self.message('empty', state), value, state)
> formencode.api.Invalid: Please enter a value

   foo(name=None) raises

    raise Invalid(self.message('empty', state), value, state)
formencode.api.Invalid: The name may not be empty.

   for me. I suspect you are using old buggy version of FormEncode.

Oleg.
--

-- 
     Oleg Broytman            http://phd.pp.ru/            phd <at> phd.pp.ru
           Programmers don't die, they just GOSUB without RETURN.

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev

Re: Changing validator messages

On Thursday 10 December 2009 23:56:51 Oleg Broytman wrote:
> On Thu, Dec 10, 2009 at 11:18:33PM +0200, Tom Coetser wrote:
> > >>> vt.foo(name=None)
> >
> >     raise Invalid(self.message('empty', state), value, state)
> > formencode.api.Invalid: Please enter a value
>
>    foo(name=None) raises
>
>     raise Invalid(self.message('empty', state), value, state)
> formencode.api.Invalid: The name may not be empty.
>
>    for me. I suspect you are using old buggy version of FormEncode.

And your suspicion was correct  The stable version in Debian Lenny seems to 
be formencode 1.0.1, but updating to 1.2 seems to have fixed the problem.

Thanks very much.

Cheers,
 Tom

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev

Re: Changing validator messages

On Fri, Dec 11, 2009 at 08:54:16AM +0200, Tom Coetser wrote:
> And your suspicion was correct  The stable version in Debian Lenny seems to 
> be formencode 1.0.1, but updating to 1.2 seems to have fixed the problem.

   SQLObject's setup.py list requirement "FormEncode>=1.1.1", and I think
the bugfix I made between 1.1.0 and 1.1.1 is related to the problem. If you
ran setup.py install it should upgrade FormEncode automatically, shouldn't
it?
   I use Debian too, but I have to install a lot of software manually,
especially Python modules. Debian "stable" is too stable (and I don't have
time to upgrade to "testing" now).

Oleg.
--

-- 
     Oleg Broytman            http://phd.pp.ru/            phd <at> phd.pp.ru
           Programmers don't die, they just GOSUB without RETURN.

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev

bug report: limit(num) method for Firebird

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 

_______________________________________________
sqlobject-discuss mailing list
sqlobject-discuss <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sqlobject-discuss

Re: bug report: limit(num) method for Firebird

On Fri, Dec 18, 2009 at 01:19:25AM +0100, Petr Jake?? wrote:
>   File
> "/usr/lib/python2.5/site-packages/SQLObject-0.10.1-py2.5.egg/sqlobject/firebird/firebirdconnection.py",
> line 120, in _queryAddLimitOffset
>     match = cls.limit_re.match(query)
> AttributeError: class FirebirdConnection has no attribute 'limit_re'

   I'll look at it.

Oleg.
--

-- 
     Oleg Broytman            http://phd.pp.ru/            phd <at> phd.pp.ru
           Programmers don't die, they just GOSUB without RETURN.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 

Re: bug report: limit(num) method for Firebird

Hi!

On Fri, Dec 18, 2009 at 01:19:25AM +0100, Petr Jake?? wrote:
>   File
> "/usr/lib/python2.5/site-packages/SQLObject-0.10.1-py2.5.egg/sqlobject/firebird/firebirdconnection.py",
> line 120, in _queryAddLimitOffset
>     match = cls.limit_re.match(query)
> AttributeError: class FirebirdConnection has no attribute 'limit_re'

   This is the patch that I think fixes the problem:

Index: firebirdconnection.py
===================================================================
--- firebirdconnection.py       (revision 4047)
+++ firebirdconnection.py       (working copy)
 <at>  <at>  -10,6 +10,8  <at>  <at> 
     dbName = 'firebird'
     schemes = [dbName]

+    limit_re = re.compile('^\s*(select )(.*)', re.IGNORECASE)
+
     def __init__(self, host, db, user='sysdba',
                  password='masterkey', autoCommit=1,
                  dialect=None, role=None, charset=None, **kw):
 <at>  <at>  -18,8 +20,6  <at>  <at> 
             import kinterbasdb
         self.module = kinterbasdb

-        self.limit_re = re.compile('^\s*(select )(.*)', re.IGNORECASE)
-
         self.host = host
         self.db = db
         self.user = user

Oleg.
--

-- 
     Oleg Broytman            http://phd.pp.ru/            phd <at> phd.pp.ru
           Programmers don't die, they just GOSUB without RETURN.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 

Adding support for additional database

I would like to add support to SQLObject for the www.rdbhost.com service.

Rdbhost provides Postgresql databases via web-service.   We have a DB
API 2 interface module, called 'Rdbhdb', to facilitate using our
hosted databases from Python, and I would like to provide 'drop-in'
support for SQLObject.

In that pursuit, I have a suggestion and a question or two:

It seems that the scheme from an opener URI is used to select the
driver, but is not provided to that driver's constructor.  For
example, the 'Pygresql' scheme will instantiate a
'pgconnection::PostgresConnection', but the __init__ is not provided
the scheme name; a Pygresql user would have to provide an explicit
backend parameter to identify the Pygresql api module.  I suggest that
dbconnection.py be modified to pass a 'scheme' parameter to the
constructors, where it would appear as part of the **kw param.

In the pgconnection init, the self.use_dsn parameter is set, and then
the self.dsn_dict and self.dsn are both created.  Why are those two
built, when a dsn string is already available?  The makeConnection
method uses only the self.dsn or only the self.dsn_dict, depending on
self.use_dsn.

What is the preferred tool for patch-making under WinXP?

Are there preconditions for adding a new database to SQLObject?

Thank you,
David

--

-- 
dkeeney <at> travelbyroad.net
Rdbhost -> SQL databases as a webservice [www.rdbhost.com]

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev