sqlobject-discuss — All things SQLObject-related

[SQLObject] SQLObject & Webware FormKit Integration

[Ian S. <Ian...@et...>]

Has anyone given any more thought to SO -> Webware (esp Fun/FormKit) =
integration?

Seems like it should be possible to generate a simple SO management =
application from an SO model?

[SQLObject] Re: Oracle

[Dirk E. <Dir...@Ce...>]

Hi,

Mark Mueller asked about the Oracle version of SQLObject a few days ago.

The state of affair's is:
I moved to a new job (university).
I do not use Oracle here.
I have not set up any DB related stuff here yet.

John Dell'Aquila and myself have a both ported SQLObject to Oracle.
Our (untested) first versions where very similar. That's a good sign. :-)
The only real issue with SQLObject and Oracle is the slice notation
part where you retrieve rows at specific table index positions.
Some DBs have LIMIT and OFFSET keywords for this. Oracle has no notion
of this. John wrote a kludge version using rownum to retrieve all rows
up to the highest index (LIMIT + OFFSET + 1) and then remove the indices
smaller than the lowest index (OFFSET) on the python side of things.

The version I have somewhere in a tarball is against SQLObject 0.4x.
I would be happy to give you a copy. Maybe I'll install Oracle
here for test purposes some day. But I can't make any promises...

Andy Todd said that he would be happy to test any versions we come
up with.

John are you still reading this list?

Regards,
Dirk

-- 
 Dirk J. Evers                      dir...@ce...
 NRW Graduate School in                             +49 (0)521/106-3793
 Bioinformatics and Genome Research, CeBiTec - Center for Biotechnology
 University of Bielefeld,                    D-33594 Bielefeld, Germany

[SQLObject] related join design?

[Tracy S. R. <tr...@re...>]

This may be a broader question about relational information design, but 
I thought I'd put this out anyway.

I would like to build an object, say a 'node', that has a number of 
related 'items' based upon a particular 'context'.  In one context, for 
example, one particular node may 'contain' a certain of items, but 
contain other items in a different context.

Here's how I have it defined in SQLObject.  Does anyone see a better 
way to do this? Or, is there a better way to design associations 
between two sets of things that depends on the settings of a 3rd thing?

	class node(SQLObject):
		name = StringCol(length=32)
		itemAssociations = MultipleJoin('itemAssociation')
	
		def getItems(self, context):
			return [ia.item for ia in self.itemAssociations if ia.context == 
context]
		
		def getItems_ALT(self, context):
			# note: 'getItems' is 5-10% faster for small sets of items
			return [iaq.item for iaq in itemAssociation.select(AND(
					itemAssociation.q.nodeID==self.id,
					itemAssociation.q.contextID==context.id))]
	
		def associateItem(self, item, context):
			return itemAssociation.new(node=self, item=item, context=context)
	
	class itemAssociation(SQLObject):
		node = ForeignKey('node')
		item = ForeignKey('item')
		context = ForeignKey('context')
	
	class item(SQLObject):
		name = StringCol(length=32)
	
	class context(SQLObject):
		name = StringCol(length=32)

For now, I have a 'compile' method in 'node' that takes a context 
object as its first arg and returns a node and its information/items 
into a dictionary where its 'items' key is a list of item dictionaries, 
too.

Thanks,
--T

Re: [SQLObject] BoolCol with SQLite

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

On Jan 8, 2004, at 10:33 PM, Javier Ruere wrote:
>   I was asking because in test_converters.ConvertersTest.test_bool,
> PostgreSQL and MySQL are tested but bot SQLite (besides my code not
> working :). Also there is special code in BoolCol for the first two DBs
> but not for the last one, IIRC.
>   Has anyone used this type with SQLite successfully?

An oversight then.  SQLite should work just like MySQL.  I thought 
there were unit tests that would have caught that, but apparently not.

--
Ian Bicking | ia...@co... | http://blog.ianbicking.org

Re: [SQLObject] BoolCol with SQLite

[Javier R. <jav...@Ho...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ian Bicking wrote:
> On Dec 28, 2003, at 1:35 PM, Javier Ruere wrote:
>
>>     I've been away for a while but I see that BoolCol has been added. I
>> have been unable to use it with SQLite. After looking the code and the
>> tests, I think it is not supported for SQLite. Am I right? It would be
>> really useful to me so if I can help with it, count on me.
>
> It should be working -- at least, I thought it was, but I don't have the
> code in front of me at the moment (SF CVS and my laptop are
> uncooperative).  Non-Boolean-supporting databases (everything but
> Postgres) just use integers (1/0).  There's other details to how BoolCol
> works, but those are database-agnostic.

  I was asking because in test_converters.ConvertersTest.test_bool,
PostgreSQL and MySQL are tested but bot SQLite (besides my code not
working :). Also there is special code in BoolCol for the first two DBs
but not for the last one, IIRC.
  Has anyone used this type with SQLite successfully?

Javier

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQE//i8g8XQC840MeeoRAgeJAKCFoy18b6oGvhjxORpQrPDGsuRPKQCfeefT
gdO7gGezukq/kp3CTc8rxzI=
=XCsl
-----END PGP SIGNATURE-----

Re: [SQLObject] BoolCol with SQLite

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

On Dec 28, 2003, at 1:35 PM, Javier Ruere wrote:
> 	I've been away for a while but I see that BoolCol has been added. I
> have been unable to use it with SQLite. After looking the code and the
> tests, I think it is not supported for SQLite. Am I right? It would be
> really useful to me so if I can help with it, count on me.

It should be working -- at least, I thought it was, but I don't have 
the code in front of me at the moment (SF CVS and my laptop are 
uncooperative).  Non-Boolean-supporting databases (everything but 
Postgres) just use integers (1/0).  There's other details to how 
BoolCol works, but those are database-agnostic.

--
Ian Bicking | ia...@co... | http://blog.ianbicking.org

Re: [SQLObject] select distinct; unique composite columns

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

On Dec 27, 2003, at 12:10 PM, David M. Cook wrote:
> * Is it possible to do a select distinct?  I want to do something like
>
>      Person.select(Address.q.personID=Person.q.id, distinct=1)
>
>  (I guess I could just do:
>
>     people_with_adresses = [p for p in Person.select() if p.addresses]
>
>   Gotta stop thinking "relationally"...

No, though perhaps that could be added.  It would be a rather small 
change to SelectResults, and DBConnection.queryForSelect.

> * Is there a way to set a unique constraint on multiple columns (with
>   autogeneration?)  Something like:
>
>       class Person(SQLOBject):
>
>   	  lastName = StringCol()
> 	  firstName = StringCol()
> 	  UNIQUE('lastName', 'firstName')

No, there's no way to do this.  In general there's few ways to add 
things to the CREATE TABLE statement that SQLObject creates.  Things 
like Type=InnoDB, unique constraints, and indexes would all be useful 
to add.  I have some other thoughts about changing the API, and this 
might fit in somehow.  Right now it's all done with separate magic 
instance variables, like _idName -- there could be a _uniqueCols (like 
_uniqueCols = [('lastName', 'firstName')]), _tableType, _indexes, 
etc... but that seems like too much, all these variables need to be 
separated off into some other kind of description.

> * How do I access the dbapi Exception classes?  Right now I do an
>
>       import sqlite as dbapi
>
>   to access them, but is there a way do dig down into the connection 
> object
>   to get them?

They aren't exposed in any way right now.  It would be nice if the 
appropriate exceptions were available as part of the connection.  One 
nuisance is the different ways database adapters use the exceptions.  
But it would still be helpful

--
Ian Bicking | ia...@co... | http://blog.ianbicking.org

[SQLObject] Oracle

[Mark M. <mar...@ya...>]

Happy New Year To All --

I have several projects that will be starting soon
that will need Oracle support - both  8i and 9i.

I've seen some discussion about adding Oracle support
to SQLObject - which would be great.  Can anyone help
me find some add-on code that would help me, otherwise
I'll have to code sql the old way.

Thanks so much!!

Mark C. Mueller


__________________________________________________
Do you Yahoo!?
New DSL Internet Access from SBC & Yahoo!
http://sbc.yahoo.com

[SQLObject] BoolCol with SQLite

[Javier R. <jav...@Ho...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi all,
	I've been away for a while but I see that BoolCol has been added. I
have been unable to use it with SQLite. After looking the code and the
tests, I think it is not supported for SQLite. Am I right? It would be
really useful to me so if I can help with it, count on me.

Thanks,
	Javier
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQE/7zBj8XQC840MeeoRApcZAJ9l03/UKlyOnP2n5KjCwhF/wt11nACdGgyS
MdDEMa0cIRwud5NFFmBBnH4=
=nHHJ
-----END PGP SIGNATURE-----

[SQLObject] select distinct; unique composite columns

[David M. C. <da...@da...>]

* Is it possible to do a select distinct?  I want to do something like

     Person.select(Address.q.personID=Person.q.id, distinct=1)
     
 (I guess I could just do:
  
    people_with_adresses = [p for p in Person.select() if p.addresses]
    
  Gotta stop thinking "relationally"...

* Is there a way to set a unique constraint on multiple columns (with
  autogeneration?)  Something like:
  
      class Person(SQLOBject):
  
  	  lastName = StringCol()
	  firstName = StringCol()
	  UNIQUE('lastName', 'firstName')

* How do I access the dbapi Exception classes?  Right now I do an
  
      import sqlite as dbapi
  
  to access them, but is there a way do dig down into the connection object
  to get them?
  

Thanks,
Dave Cook
  

Re: [SQLObject] import problems (revisited)

[Scott R. <sc...@to...>]

Looks like I may be unfairly blaming SQLObject here.  I think it's
something wierd in the bowels of WebKit's ImportSpy stuff, coupled with
a sensitive assert...  WebKit does seem to actually re-import
sub-imports of __import__ed modules.  Not sure why.

[SQLObject] import problems (revisited)

[Scott R. <sc...@to...>]

I'm converting a mid-size project from webware's middlekit to sqlobject.

One problem that I've run out of workarounds for is an import issue,
similar to one that was last discussed on the mailing list in September.

I'm storing all of my SQLObject classes in a single module called
booking.  At the top of files that use the classes, I include the line

from stuff import *

This seems to work fine, except occasionally, when I get 

  File "/sandbox/Main/eventleaf.py", line 1, in ?
    from stuff import *
  File "/usr/lib/python2.2/ihooks.py", line 397, in import_module
    q, tail = self.find_head_package(parent, name)
  File "/usr/lib/python2.2/ihooks.py", line 433, in find_head_package
    q = self.import_it(head, qname, parent)
  File "/usr/lib/python2.2/ihooks.py", line 486, in import_it
    m = self.loader.load_module(fqname, stuff)
  File "./WebKit/ImportSpy.py", line 30, in load_module
  File "/usr/lib/python2.2/ihooks.py", line 270, in load_module
    m = self.hooks.load_source(name, filename, file)
  File "/usr/lib/python2.2/ihooks.py", line 168, in load_source
    return imp.load_source(name, filename, file)
  File "/sandbox/Main/stuff.py", line 5, in ?
    class Person(SQLObject):
  File "/usr/lib/python2.2/site-packages/SQLObject/SQLObject.py", line 151, in __new__
    assert not classRegistry.get(registry, {}).has_key(className), "A database object by the name %s has already been created" % repr(className)
AssertionError: A database object by the name 'Person' has already been created

I'm not changing the definition, just importing the SQLObj classes into
a module that is imported by a module that has already imported the
SQLObj classes.

This seems intermittent, and I have not yet tracked down the situation
that makes it happen, but I suspect it may only be happening to modules
loaded via this code:

    res = imp.find_module(req + "leaf")
    mod = imp.load_module(req + "leaf", res[0], res[1], res[2])

This may be a red herring, but it's starting to smell like the cause.

Could someone who understands python and sqlobject internals better than
I do help me solve this puzzle?

  - Scott

[SQLObject] value cache and validators

[alexander s. <al...@an...>]

hello!

i think there is a bug in SQLObject attribute value cache: when 
'_cacheValues' is set, an attribute getter is replaced by value written 
to data base.  if attribute is converted by Validator (fromPython), 
later access will return DB value for the attribute, and not it's 
"pythonic" representation.

merry Christmas!
alex.

[SQLObject] Re: autoCommit, PostgresConnection problems

[Guenther S. <gs...@sy...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sunday 21 December 2003 14:09, Guenther Starnberger wrote:

hi,

> I digged a little bit around in the SQLObject source & found some things
> which look like bugs to me (altough i'm not really sure :). Maybe someone
> who understands the framework better than me can review these / resolve
> them.

just noticed another problem:

The CacheFactory only creates self.cache if caching is enabled. When=20
'rollback' is called on a Transaction object it does (among other things):

subCaches =3D [(sub, sub.allIDs()) for sub in self.cache.allSubCaches()]

but the allIDs() method of the CacheFactory just returns self.cache.keys()=
=20
without checking if caching is enabled which results in an AttributeError=20
beeing raised.

cu
/gst

=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/5b6OZtF7I/+gjcERAvu/AJ4i/oDsYbmKQkIdpBhgbRpJDsJgpACg1QF4
zIuOzPcRrBpGjtN+fOBGJJ0=3D
=3DI234
=2D----END PGP SIGNATURE-----

[SQLObject] autoCommit, PostgresConnection problems

[Guenther S. <gs...@sy...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi all,

I digged a little bit around in the SQLObject source & found some things wh=
ich=20
look like bugs to me (altough i'm not really sure :). Maybe someone who=20
understands the framework better than me can review these / resolve them.

=2D ---

DBConnection.py / PostgresConnection / __init__:

=2D -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------
def __init__(self, dsn=3DNone, host=3DNone, db=3DNone,
___ user=3DNone, passwd=3DNone, autoCommit=3D1,
___ usePygresql=3DFalse,
___ **kw):

[...]

self.autoCommit =3D autoCommit

[...]

DBAPI.__init__(self, **kw)
=2D -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------

When you run PostgresConnection with the argument 'autoCommit =3D 0' in the=
=20
constructor it sets self.autoCommit to 0 too.

But afterwards it calls DBAPI.__init__ only with **kw, but not with the oth=
er=20
named arguments. DBAPI.__init__ calls DBConnection.__init__ which doesn't g=
et=20
the autoCommit argument (because it isn't stored in **kw) and which default=
s=20
to 'autoCommit=3DTrue'.

Therefore autoCommit is always True, no matter which value is passed to the=
=20
PostgresConnection Constructor.

=2D ---

DBConnection.py / class DBAPI / releaseConnection:

=2D -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------
elif self.autoCommit:
___ if self.debug:
_______ self.printDebug(conn, 'auto', 'COMMIT')
___ conn.commit()
=2D -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------

when autoCommit is active the PgSQL adapter throws an 'Commit failed -=20
autocommit is on.' exception when commit() is called.

Shouldn't this be a 'elif not self.autoCommit' instead (without an else=20
statement afterwards)? Imo the connection only needs the 'COMMIT' if it isn=
't=20
set to autoCommit.

There is another similiar Problem with the 'pool' argument which is added b=
y=20
PostgresConnection to **kw, but which isn't expected by DBConnection. (I=20
posted bug #859999 some time ago on the SourceForge project page).

=2D ---

btw, I have used my PgSQL adapter (i posted a patch some time ago) now for=
=20
some time without any major problems.

As (AFAIK) the only Python PostgreSQL Library PgSQL supports Unicode (my=20
adapter should support this, though i have to admit that i haven't really=20
tested it yet). As PgSQL Code is in its own class (inherited from=20
PostgresConnection), it shouldn't affect any other code.

If there's a chance of getting PgSQL Support into SQLObject i'll repost my=
=20
patch to it.

cu
/gst

=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/5ZumZtF7I/+gjcERAl/rAJ42A0m6XKIHqxi4WHaXuxwVPIIT2wCghx6Q
9Iglk+dsGtB5Usn1giTfUsw=3D
=3Dmp5M
=2D----END PGP SIGNATURE-----

Re: [SQLObject] bug: connection/transaction leak?

[Guenther S. <gs...@sy...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Friday 19 December 2003 22:05, Ian Bicking wrote:

hi,

> Each new Account is going to have a reference to the connection, so
> that's to be expected.  Hmm... and the accounts won't be deleted
> because they are cached.  That's the problem, then.  Maybe the
> transaction simply needs to clear its cache on a commit/rollback (well,
> you can't really clear the cache, but you can make all the objects
> GCable).

hmm..  shouldn't objects in the cache automatically expire 'faster' when i =
set=20
the cullFrequency to a lower value? i tried setting the cullFrequency to 5=
=20
and running gc.collect() after every iteration, but the refcount to the=20
connection still stays the same (resp. +1 at every iteration).

on the other hand when i disable caching by passing "cache =3D False" to th=
e=20
connection & run the gc it works fine (the refcount to 'conn' doesn't=20
increase).

cu
/gst

=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/5GWeZtF7I/+gjcERAva+AJ9H/M2TjDu4GeP+ypEreDrud/zOVQCgiNdp
hZ13/zMOlWTM+0/bkJzXYHA=3D
=3D3rEe
=2D----END PGP SIGNATURE-----

Re: [SQLObject] bug: connection/transaction leak?

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

> using the following while block:
>
> while 1:
> ___ b = Account.new(name = "FOO" + str(time.time()), password = "bar", 
> connection = conn)
> ___ print sys.getrefcount(conn)
> ___ del(b)
>
> shows an increasing refcount of conn with every iteration. maybe the 
> leak
> is somewhere in the sqlobject class?

Each new Account is going to have a reference to the connection, so 
that's to be expected.  Hmm... and the accounts won't be deleted 
because they are cached.  That's the problem, then.  Maybe the 
transaction simply needs to clear its cache on a commit/rollback (well, 
you can't really clear the cache, but you can make all the objects 
GCable).

--
Ian Bicking | ia...@co... | http://blog.ianbicking.org

Re: [SQLObject] bug: connection/transaction leak?

[Guenther S. <gs...@sy...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thursday 18 December 2003 00:50, you wrote:

hi,

> Hmm.. that shouldn't happen.  Though looking at the code, I don't do a
> whole lot to return connections.  The transaction should get garbage
> collected, and __del__ releases the connection, but I can imagine
> there'd be lots of reasons a transaction wouldn't get collected (though
> in this case it looks like it should be).
>
> Probably on commit()/rollback() the transaction should become invalid,
> and the connection should be returned.  Or maybe you could run
> trans.begin() to revalidate it (at which point it'd grab another
> connection).

using the following while block:

while 1:
___ b = Account.new(name = "FOO" + str(time.time()), password = "bar", connection = conn)
___ print sys.getrefcount(conn)
___ del(b)

shows an increasing refcount of conn with every iteration. maybe the leak
is somewhere in the sqlobject class?

cu
/gst

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/42OUZtF7I/+gjcERAqPsAKCuapSeoG3udyvRrytIRu67bqxRzACgzO5F
FblDKu3mUSZvZ0RNtfO57Yo=
=IX1s
-----END PGP SIGNATURE-----

[SQLObject] Patch to Support psycopg.Binary

[J-P L. <sql...@si...>]

<mailto:sql...@li...>This is a patch against 
v0.5.1.  I'm using bytea field types in my tables and this patch makes 
life easier.  In my SQLObject classes, I do:

def _set_data(self, val):
    self._SO_set_data(psycopg.Binary(val))

Seems to work with some preliminary testing.


Index: SQLObject/Converters.py
===================================================================
RCS file: /usr/local/cvs/depot/sqlobject/SQLObject/Converters.py,v
retrieving revision 1.1
retrieving revision 1.2
diff -u -5 -r1.1 -r1.2
--- SQLObject/Converters.py    17 Nov 2003 22:36:02 -0000    1.1
+++ SQLObject/Converters.py    18 Dec 2003 15:03:27 -0000    1.2
@@ -135,10 +135,19 @@
     return "(%s)" % ", ".join([sqlrepr(v, db) for v in value])
 
 registerConverter(type(()), SequenceConverter)
 registerConverter(type([]), SequenceConverter)
 
+try:
+    import psycopg
+except ImportError: pass
+else:
+    def PsycopgBinaryConverter(value, db):
+        return str(value)
+
+    registerConverter(type(psycopg.Binary('')), PsycopgBinaryConverter)
+
 if hasattr(time, 'struct_time'):
     def StructTimeConverter(value, db):
         return time.strftime("'%Y-%m-%d %H:%M:%S'", value)
 
     registerConverter(time.struct_time, StructTimeConverter)

Re: [SQLObject] bug: connection/transaction leak?

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

On Dec 17, 2003, at 3:24 PM, Guenther Starnberger wrote:
> from SQLObject import *
> import time
>
> class Account(SQLObject):
> ___ name = StringCol(alternateID = True)
> ___ password = StringCol()
>
> conn = PostgresConnection('user=zope dbname=WebspaceNG', autoCommit = 
> False)
>
> while 1:
> ___ trans = conn.transaction()
> ___ b = Account.new(name = "FOO" + str(time.time()), password = "bar", 
> connection = trans)
> ___ trans.commit()
> ___ raw_input()
> - -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------
>
> if you run this code, with every iteration an additional connection 
> gets
> opened to the postgresql server (until the max. connection limit is 
> reached &
> postgresql refuses new connections).

Hmm.. that shouldn't happen.  Though looking at the code, I don't do a 
whole lot to return connections.  The transaction should get garbage 
collected, and __del__ releases the connection, but I can imagine 
there'd be lots of reasons a transaction wouldn't get collected (though 
in this case it looks like it should be).

Probably on commit()/rollback() the transaction should become invalid, 
and the connection should be returned.  Or maybe you could run 
trans.begin() to revalidate it (at which point it'd grab another 
connection).

--
Ian Bicking | ia...@co... | http://blog.ianbicking.org

[SQLObject] bug: connection/transaction leak?

[Guenther S. <gs...@sy...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

hi,

i noticed that when using sqlobject together with zope after some time more
and more connections are open to my postgresql server.

first i thought that it was a zope bug (i am binding sqlobject transactions=
 to
zope transactions and assumed, that the transaction objects were referenced
from somewhere in the ZODB), but now it seems that it is a bug in sqlobject.
  =20
=2D -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------
from SQLObject import *
import time

class Account(SQLObject):
___ name =3D StringCol(alternateID =3D True)
___ password =3D StringCol()
                                                                           =
                                                                     =20
conn =3D PostgresConnection('user=3Dzope dbname=3DWebspaceNG', autoCommit =
=3D False)
                                                                           =
                                                                     =20
while 1:
___ trans =3D conn.transaction()
___ b =3D Account.new(name =3D "FOO" + str(time.time()), password =3D "bar"=
, connection =3D trans)
___ trans.commit()
___ raw_input()
=2D -------- 8< -------- 8< -------- 8< -------- 8< -------- 8< --------

if you run this code, with every iteration an additional connection gets
opened to the postgresql server (until the max. connection limit is reached=
 &
postgresql refuses new connections).

cu
/gst

=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/4Ml1ZtF7I/+gjcERAkT8AJ0bvPL8+SoglxVTWbVrAdPmO/qRUACZAc9y
lAIFFBN8XaQRxA0btc4E8iM=3D
=3DXoeD
=2D----END PGP SIGNATURE-----

Re: [SQLObject] overriding new and calling the original method doesn't work

[Guenther S. <gs...@sy...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Tuesday 16 December 2003 01:28, Ian Bicking wrote:

Hi,

> def new(cls, **kw):
>      yada yada
>      return super(MyClass, cls).new(**kw)
> new = classmethod(new0

Yop - thanks Ian & Edmund.

In this way it works. :)

cu
/gst

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/3rC6ZtF7I/+gjcERAv6gAKDGheG83UiTzorkNjT2hNggaK7kqwCgvDR8
C+n+l+MfQIj3WgSTrx6LCz0=
=vCMh
-----END PGP SIGNATURE-----

Re: [SQLObject] Patch: Database Errors not Releasing Pooled Connections

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

On Dec 13, 2003, at 11:04 PM, Ken Kinder wrote:
> Upon exploring the source code, I think I have isolated the problem. 
> Your
> DBAPI.runWithConnection method from SQLObject/DBConnection.py has the
> following definition:

Good catch.  Applied to CVS.

--
Ian Bicking | ia...@co... | http://blog.ianbicking.org

Re: [SQLObject] overriding new and calling the original method doesn't work

[Edmund L. <el...@in...>]

On Dec 16, 2003, at 10:33, Edmund Lian wrote:

> I started that thread. Which version of Python are you using? Turns 
> out that there is a bug in the classmethod code that causes 
> inheritance problems in Python < 2.2.3 (I think, can't quite remember 
> now). That was one problem.

The bug was but 535444, where super() was broken for classmethods in 
Python < 2.2.1

...Edmund.

Re: [SQLObject] overriding new and calling the original method doesn't work

[Edmund L. <el...@in...>]

On Dec 15, 2003, at 18:26, Guenther Starnberger wrote:

> some time ago there was a thread on this mailinglist, about overiding 
> the
> new() method in classes inherited from SQLObject.
> (http://thread.gmane.org/gmane.comp.python.sqlobject/906)
>
> according to a posting in this thread it is possible to call the 
> 'original'
> SQLObject new() method by: 'SQLObject.new(cls, **kw)'.
>
> but somehow this doesn't work for me :/
>
> class Account(SQLObject):
> __name = StringCol(alternateID = True)
> __password = StringCol()
>
> __def new(cls, **kw):
> ____print "FOO"
> ____SQLObject.new(cls, **kw)
>
> __new = classmethod(new)

I started that thread. Which version of Python are you using? Turns out 
that there is a bug in the classmethod code that causes inheritance 
problems in Python < 2.2.3 (I think, can't quite remember now). That 
was one problem.

I override new as follows:

def new(cls, **kw):
         """
         Overrides class method .new() to automatically add mandatory
         sections to request object
         """
         obj = super(Request, cls).new(**kw)
         return obj
new = classmethod(new)


...Edmund.