Re: [ZODB-Dev] Python properties on Persistent objects
> > Properties work with persistent objects with one limitation that I'm aware of. > The __setattribute__ method used by the persistent base class assumes > that any attribute assignment modifies the object's persistent state. > A property > that doesn't modify state or that modifies volatile state will fool > it. Conceivably, > one could override __setattribute__ to handle such properties, but doing so > would require great care. I don't recommend it. Great, thanks! I already assumed so, but now this little feature is officially confirmed. -Mikko ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
Re: [ZODB-Dev] Python properties on Persistent objects
On Wed, Dec 16, 2009 at 7:12 PM, Mikko Ohtamaa wrote: > I need to have little clarification should properties work on > Persistent objects. Properties work with persistent objects with one limitation that I'm aware of. The __setattribute__ method used by the persistent base class assumes that any attribute assignment modifies the object's persistent state. A property that doesn't modify state or that modifies volatile state will fool it. Conceivably, one could override __setattribute__ to handle such properties, but doing so would require great care. I don't recommend it. Jim -- Jim Fulton ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
Re: [ZODB-Dev] Python properties on Persistent objects
> That code is insane. Really insane. It's generating classes and doing > custom adapter factories that inject a _v_ variable into a persistent > object temporarily. I wouldn't try to copy it at all. I would run away. But I thought this was *the* Zope way and the brightest minds in the community have obviously written the code :) In any case, the persistent + property issue has received some awareness until it bits someone's ass again. I am going back to the drawing board and my behaviors. property set = always starts transaction even though nothing is being written -Mikko ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
Re: [ZODB-Dev] Python properties on Persistent objects
Mikko Ohtamaa wrote: >> This isn't right: a z3c.form form is just a view like any other. It is >> looked up on the context and the request. In Dexterity (which I assume >> you're using?), that's going to be the content object. > > Sorry, I think I mixed with zope.formlib. In any case I hope to find > the answer to the orignal question. > > The original pattern which I copied is is here: > http://code.google.com/p/getpaid/source/browse/Products.PloneGetPaid/trunk/Products/PloneGetPaid/preferences.py That code is insane. Really insane. It's generating classes and doing custom adapter factories that inject a _v_ variable into a persistent object temporarily. I wouldn't try to copy it at all. I would run away. I *think* what may be happening here is that most of the vocabularies in plone.app.vocabularies have code like this: context = getattr(context, 'context', context) This hack is there to support the IAdding view use case where the context of the add form is another view and the first "real" content object from which you can acquire stuff is self.context.context. We're moving away from that in favour of not acquiring tools and not using views-on-views, which are evil anyway. Hence, the 'context' property in StoreSettings, which I presume is used as the context for a form, is being cajoled into being the Plone site root. But that's using a hack to trick a hack into doing different hackery. This is getting far off topic for the ZODB list, but if you can explain what problem you're actually trying to solve, I'm sure we can come up with a suggestion that is less of an architecture spacewalk. When you start having to ask the ZODB list, maybe it's time to think whether you could find a simpler solution. :-) > I don't think I am doing it incorrectly. The other reason of using > this pattern is that the behavior factory does not need to have a > write or (an extra) object load if the behavior is never set on the > context object. The default object returned by factory acts similar > than the persistent object and I can set the behavior defaults using > zope.schema - I don't need to do "if not set" logic anywhere else in > my code. There is no reason you can't get this "don't-write-unless-necessary" behaviour in a much more sane way. In fact, it ought to be the default. To help you, I'd need a more comprehensive understanding of what you're doing and why, though. Martin -- Author of `Professional Plone Development`, a book for developers who want to work with Plone. See http://martinaspeli.net/plone-book ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
Re: [ZODB-Dev] Python properties on Persistent objects
> > This isn't right: a z3c.form form is just a view like any other. It is > looked up on the context and the request. In Dexterity (which I assume > you're using?), that's going to be the content object. Sorry, I think I mixed with zope.formlib. In any case I hope to find the answer to the orignal question. The original pattern which I copied is is here: http://code.google.com/p/getpaid/source/browse/Products.PloneGetPaid/trunk/Products/PloneGetPaid/preferences.py > > I think you need to take a step back and look at how your form is set > up, not dive into deep magic of persistence. I don't think I am doing it incorrectly. The other reason of using this pattern is that the behavior factory does not need to have a write or (an extra) object load if the behavior is never set on the context object. The default object returned by factory acts similar than the persistent object and I can set the behavior defaults using zope.schema - I don't need to do "if not set" logic anywhere else in my code. =) Cheers, -Mikko ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
Re: [ZODB-Dev] Python properties on Persistent objects
Laurence Rowe wrote: > 2009/12/17 Mikko Ohtamaa: >> Hi, >> >> I need to have little clarification should properties work on >> Persistent objects. I am running ZODB 3.8.4 on Plone 3.3. >> >> I am using plone.behavior and adapters to retrofit objects with a new >> behavior (HeaderBehavior object). This object is also editable through >> z3c.form interface. z3c.form requires a context variable on the object >> e.g. to look up dynamic vocabularies. To avoid having this >> object.context attribute to be peristent (as it's known every time by >> the factory method of the adapter which creates/look-ups >> HeaderBehavior) I tried to spoof context variable using properties and >> internal volatile variable. This was a trick I learnt somewhere >> (getpaid.core?) > > This sounds like you are passing context somewhere where view is expected. Yeah, to me as well it sounds like you're barking up the wrong tree here. You shouldn't need any kind of faking like this. "z3c.form requires a context variable on the object e.g. to look up dynamic vocabularies" This isn't right: a z3c.form form is just a view like any other. It is looked up on the context and the request. In Dexterity (which I assume you're using?), that's going to be the content object. I think you need to take a step back and look at how your form is set up, not dive into deep magic of persistence. Martin -- Author of `Professional Plone Development`, a book for developers who want to work with Plone. See http://martinaspeli.net/plone-book ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
Re: [ZODB-Dev] Python properties on Persistent objects
2009/12/17 Mikko Ohtamaa : > Hi, > > I need to have little clarification should properties work on > Persistent objects. I am running ZODB 3.8.4 on Plone 3.3. > > I am using plone.behavior and adapters to retrofit objects with a new > behavior (HeaderBehavior object). This object is also editable through > z3c.form interface. z3c.form requires a context variable on the object > e.g. to look up dynamic vocabularies. To avoid having this > object.context attribute to be peristent (as it's known every time by > the factory method of the adapter which creates/look-ups > HeaderBehavior) I tried to spoof context variable using properties and > internal volatile variable. This was a trick I learnt somewhere > (getpaid.core?) This sounds like you are passing context somewhere where view is expected. Laurence ___ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev
[ZODB-Dev] Python properties on Persistent objects
Hi, I need to have little clarification should properties work on Persistent objects. I am running ZODB 3.8.4 on Plone 3.3. I am using plone.behavior and adapters to retrofit objects with a new behavior (HeaderBehavior object). This object is also editable through z3c.form interface. z3c.form requires a context variable on the object e.g. to look up dynamic vocabularies. To avoid having this object.context attribute to be peristent (as it's known every time by the factory method of the adapter which creates/look-ups HeaderBehavior) I tried to spoof context variable using properties and internal volatile variable. This was a trick I learnt somewhere (getpaid.core?) It didn't work. Looks like Persistent class is not aware of properties and interprets property set as a transaction start. So even though I thought I was having a volatile context, my Undo log kept having new write entries on every page read for pages having HeaderBehavior objects in their annotations. Data.fs grew steadily with small writes and there was a whole bunch of ConflictErrors. I think I found a workaround. By overriding __setattr__ and not letting through properties and volatile attribute set to Persistent __setattr__ you can avoid the problem. Is my rationale correct? Should Persistent behave well with properties? I found only one discussion in Google, circa 2004. Sample code below: from persistent import Persistent class VolatileContext(Persistent): """ Mix-in class to provide non-persistent context attribute to persistent classes. Some subsystems (e.g. z3c.forms) expect objects to have a context reference to parent/site/whatever. However, storing this back-reference persistenly is not needed, as the factory method will always know the context. This helper class creates a context property which is volatile (never persistent), but can be still set on the object after creation or after database load. """ zope.interface.implements(IVolatileContext) def _set_context(self, context): self._v_context = context def _get_context(self): return self._v_context def _set_factory(self, factory): self._v_factory = factory def _get_factory(self): return self._v_factory # http://docs.python.org/library/functions.html#property context = property(_get_context, _set_context) factory = property(_get_factory, _set_factory) def save(self): """ """ self.factory.makePersistent(self) def __setattr__(self, name, value): if name not in ("context", "factory", "_v_context", "_v_factory"): Persistent.__setattr__(self, name, value) print "leaking " + name + " " + str(value) else: print "property set " + name + " " + str(value) object.__setattr__(self, name, value) class AnnotationPersistentFactory(object): """ A factory pattern to manufacture persistent objects stored within the parent object annotations. Until the first write, the default (non-persistent) object is return. This prevents possible situations where database read could cause write. The first write must call AnnotationPersistentFactory.makePersistent(object). Alternative, you can call AnnotationPersistentFactory.makePersistent(object) when entering the editing interface for the first time. After the first write, the saved persistent object is return. """ def __init__(self, persistent_class, key): """ @param persistent_class: Class reference / factory method which will create new objects. Created classes must conform VolatileContext interface @param key: ASCII string, Key name used with IAnnotations """ self.persistent_class = persistent_class self.key = key self._assertProperlySetUp() def _assertProperlySetUp(self): """ Check that the framework is properly set up """ assert callable(self.persistent_class), "Factory is missing" assert hasattr(self.persistent_class, "context"), "The persistent object must support volatile context interface" assert self.key is not None, "You must give the annotations key" def makePersistent(self, object): """ Write created persistent object to the database. This will store the object on the annotations of its context. """ assert isinstance(object, self.persistent_class), "Object %s was not type of %s" % (str(object), str(self.persistent_class)) annotations = IAnnotations(object.context) annotations[self.key] = object def __call__(self, context): """ Called by Zope framework when doing a factory call. Usually this class is refered as and this method creates a new, read-only, persistent object. """ annotations = IAnnotations(context) if not self.key in annotations: # Construct a new (default) i