Re: Adding history/versioning to a Nested Set model (is it possible?)

On Oct 30, 5:37 pm, Ed Lucas <edward.lu..._at_gmail.com> wrote:
> On 30 Oct, 16:04, Tegiri Nenashi <tegirinena..._at_gmail.com> wrote:
>
>
>
>
>
> > On Oct 28, 9:55 am, Ed Lucas <edward.lu..._at_gmail.com> wrote:
>
> > > Hi All,
>
> > > I'm trying to implement a tree structure, hence I'd like the Nested
> > > Set model, but I need to be able to record the history of the tree
> > > structure *somehow* so that I can access previous versions.
>
> > > I've searched up and down online but all I can find are transactions
> > > +rollbacks.
>
> > > As far as I can see...
>
> > > The simplest option would be to add a version number to the model, but
> > > then every member of the set would have to be updated each time a
> > > change is made - not feasible because of how fast that would make the
> > > table would grow, and how slow that would make it to update?
>
> > > I imagine I could log each action and then replay all the log actions
> > > in a temporary table when I needed them, but that would get pretty
> > > slow after a few updates.
>
> > > The previous Nested Set could be serialised to an blob and stored.
> > > That's currently looking like the best option, but only by process of
> > > elimination...
>
> > > I've probably missed something obvious that's staring me in the face
> > > here! - any suggestions?
>
> > > Thanks for taking the time to read this far, best regards,
> > > Ed
>
> > When you make even small change to a tree (e.g. insert a node), you
> > have to recalculate half of the node encodings on average. This
> > feature alone invalidates nested sets approach for temporal
> > hierarchies. Google nested intervals.
>
> Aha! Thank you very much to you and Vladimir!
>
> Now I just need to sit in a darkened room with a lot of coffee while I
> get my head round it all...- Hide quoted text -
>
> - Show quoted text -
Received on Sat Oct 31 2009 - 19:03:47 CET