Virtual Trees, TreeModel and TreeView

> Hi,
>
> in the Dolphin Smalltalk Companion Ted Brecht suggest that it's better to
> embed a hierarchy in the model itself and I couldn't agree more.
>
> However in the examples he gives, starting at page 213, he puts the
> hierarchy into the model and also additionally into a "Virtual Tree", in
> which every action to the model needs to be replicated, as only the
> "Virtual Tree" can be passed to a TreePresenter as the model.
>
> I don't know but to me that seems rather inelegant, and error prone.

> "Guenther Schmidt" <[hidden email]> wrote in message
> news:[hidden email]...
>
>>Hi,
>>
>>in the Dolphin Smalltalk Companion Ted Brecht suggest that it's better to
>>embed a hierarchy in the model itself and I couldn't agree more.
>>
>>However in the examples he gives, starting at page 213, he puts the
>>hierarchy into the model and also additionally into a "Virtual Tree", in
>>which every action to the model needs to be replicated, as only the
>>"Virtual Tree" can be passed to a TreePresenter as the model.
>>
>>I don't know but to me that seems rather inelegant, and error prone.
>
>
> Why? VirtualTreeModel is simply an Adaptor which can be used to create a
> TreeModel "view" onto an arbitrary hierarchical data structure.
> This allows for the use of hierarchies that wouldn't be useable directly
> with TreePresenter without building a parallel TreeModel containing the same
> hierarchy of nodes. Such hierarchies might need the adapter because they are
> pre-existing (e.g. the class hierarchy), or because they are implemented
> externally (e.g. the file system), or because you want to design a new
> hierarchy without the constraint of representing it as a TreeModel.
>
>
>>He merely indicates a third option, doing without the "Virtual Tree".
>>
>>Anybody care to indulge me how that could be done?
>
>
> 3 options:
> 1) Subclass TreeModel to implement your model.
> 2) Implement the <treeModel> protocol on your model so that it compatible
> with TreeModel without actually being a TreeModel.
> 3) Construct a parallel TreeModel to display your model when needed. This
> only really makes sense if the model is fairly static, or if a
> VirtualTreeModel would be too slow.
>
> Regards
>
> Blair
>
>

> Guenther Schmidt wrote:
>
>
>>However it doesn't automatically update the view, which might just be an
>>implementation error of mine.
>
>
> That's because there's nothing to tell the TreeView that the tree has changed.
>
> If you change any model then /something/ must tell the corresponding view(s)
> about the change.  If you are using a TreeView then it expects to see certain
> specific events triggered off its Model.  For instance #item:addedInParent:.
> If you don't generate those events (somehow) then it will not update as the
> tree changes.
>
> Now if you are using a VirtualTreeModel, then /you/ have to tell the virtual
> tree about any changes that you have made to the /real/ tree (the hierarchical
> arrangement of the objects in your real Model).  It will then respond by
> triggering the events that a TreeView is expecting to see.  The way
> you tell it about such changes is by calling methods like #add:asChildOf:.  If
> you look at the implementation of that method, you'll see that it just triggers
> an event.
>
> You may think it's odd that you have to call a method like #add:asChildOf: in
> order to tell it about changes, rather than -- say -- #notifyChild:addedTo:.
> I think it's odd myself.  However that is what its called, so that's what you
> have to use.  It may /look/ like error-prone duplication, but it's really just
> a change notification.
>
> BTW, I don't think the choice of selector is /wrong/ -- it means that the user
> of the tree model is insulated from knowing whether the tree model is actually
> virtual or non-virtual, or maybe some kind of hybrid of the two.  Still, it is
> definitely confusing when you first try to work out how Dolphin tree models
> work.
>
>     -- chris
>
> (BTW, looking at the tree example in Ted book, there's some stuff about
> #beFinalizable which I recommend you just ignore.  I think it must be an error
> that managed to get past the proof reading somehow.)
>
>
>