ZODB: don't like self._p_changed

[Dylan Reinhardt]

There's no way changing Python built-ins is preferable.

What you *should* do is initialize *all* your list/dict subobjects using 
PersistentList and PersistentMapping.   Now, every time you make a change 
to self.something, you know it gets persisted.  Easy enough to keep track 
of, yes?

More the point, leaving the built-ins alone means that using them as local 
variables won't trigger persistence.  It shouldn't, after all.  ZODB bloats 
very quickly if you over-use persistence.

If it's the syntax you don't like, another thing you can do is define:

def save(self):
     self._p_changed = 1

and then call self.save() at the end of any methods that change 
subobjects.  Just a bit of sugar, perhaps, but I find it far more readable 
than making an assignment to a private inherited attribute.

HTH,

Dylan



At 04:58 AM 1/9/2003, you wrote:
>Max M schrieb:
>>Thomas Guettler wrote:
>>
>>>Max M schrieb:
>>
>>>>Probably not. How should an object know if a mutable subobject has been 
>>>>changed?
>>>
>>>
>>>The method "append" could set the flag.
>>
>>Then you would need to use specialized objects instead of the built in 
>>objects. That would shurely be more cumbersome than setting the 
>>persistence flag?
>
>Yes, there a classes like PersistentList and PersistentMapping, but
>this means I need to change my code.
>
>Example:
>
>my_list=PersistentList()
>my_list.append(...)
>
>my_list=some_other_list[1:] # now my_list is no more a persistent list
>
>This means, I always need to keep in mind if a list is a peristent list
>or not. That's like setting _p_changed=1.
>
>1. Solution:
>
>One solution could be to change the meaning of [] and {}.
>Dreaming:
>
>import list_persistance
>foo=[] # foo is now a persistent list.
>
>I don't know if this is possible.
>
>
>2. Solution:
>Change the functions of list and mapping datatypes to set _p_changed if 
>the content was changed. This might decrease performance for nonpersistent 
>application. But since there changes need to be implemented in the python 
>core, they would be in C.
>
>thomas
>
>--
>http://mail.python.org/mailman/listinfo/python-list