Re: [Changeset]: Re: cla() ?

[Michael Goffioul]

On Wed, Oct 8, 2008 at 8:26 PM, John W. Eaton <address@hidden> wrote:
> On  7-Oct-2008, Michael Goffioul wrote:
>
> | While we're talking about __go_delete__, does anybody
> | remember what was the idea behind removing an object
> | from its parent's children in __go_delete__, instead of
> | gh_manager::free? This makes life slighlty more complex
> | if you want to delete an object from C++, where you would
> | typically use gh_manager::free, but you then have to remove
> | the child manually from its parent's children.
> |
> | I don't think I wrote that part of the code, and I can't figure
> | out why it is written that way.
>
> It's likely that I wrote it, and I assumed that a call to the delete
> function should remove the handle from its parent's list of children.
> Is that not what should happen?  If not, then what should happen?  For
> example, what does Matlab do for
>
>  h = line (); get (gca (), 'children'), delete (h); get (gca (), 'children')
>
> ?  Shouldn't the children of the current axes be empty after H is
> deleted?  If that doesn't happen automatically in the delete function,
> then how does it happen?

It should definitely be removed from the children list, that's not the
problem here. My question is why is that operation (== removing
the handle from its parent's children) in __go_delete__ and not in
gh_manager::free.

I'm asking, because when you want to delete an object from C++
world, you typically call gh_manager:free. But this does not take
care of updating the parent's children and you might easily forget
about it.

Michael.