Re: [SQLObject] Re: Re: Question about multiple deletes

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

Chris Gahan wrote:
>>>There's nothing in SQLObject right now that does this, but you could add
>>>your own method pretty easily...
>>>
>>>Just add a function to the SQLObject class called .multiDelete() or
>>>something, which works the same as .select().. (i.e. takes in the same
>>>parameters... obviously it would do something quite different. ;)
>>
>>Oops, guess I missed this message.
>>
>>It's a little more complicated than that, because a delete should really
>>go through the SQLObject instances.  Selects should be extended so you
>>can retrieve IDs instead of full objects, then there should be a way to
>>fetch an object only if it already exists in Python.  Then you'd iterate
>>over the select results and call destroySelf on any objects that
>>existed, and then do another SQL statement to delete any rows that
>>didn't have Python instances.
> 
> 
> Wow, yeah, I didn't think about object-to-database consistency. What a
> bitch. :)
> 
> Object databases are hard!

Actually, while there's a bunch of details, none of them are 
particularly difficult.  Everything's there, a few new APIs just have to 
be exposed (to get IDs, and to get objects from the cache)

> Hmmm.. couldn't you have a simpler method though, where the object was
> ignorant of its being deleted until it actally tried to sync itself with the
> database? I mean, what's the reason for using destroySelf to kill objects?
> Is it to keep the cache clean?

Actually, I think they might stay in the cache now (though obviously 
they shouldn't).  As a basic principle, destroySelf should be called so 
that objects can clean themselves up.  This allows for things like 
cascading deletes on databases that won't do it for you.  (But it also 
means we have to calculate the cascades even on databases that do 
support doing it themselves)

   Ian