On Thursday 01 August 2002 1:31 pm, Lars Gullik Bjønnes wrote:
> | Think of all those times when a BufferView * (or even a BufferView &) is
> | passed to an inset. It makes no sense at all to pass a shared_ptr in
> | these cases, but if the inset wants to cache this BufferView * then it
> | can do so as
>
> This is where we disagree... it think it makes perfect sense to
> _always_ pass a shared_ptr.
Well let's agree to differ on practical grounds then. I am /not/ going to
change 10000 inset classes s/BufferView */boost::shared_ptr<BufferView> &/
plus all associated changes when perhaps one or two things might possibly
want the shared_ptr not the thing it points to.
Maybe if André's great unification project comes off then this would be
easier since it looks like there'll be a single LyX inset that can hold any
other inset, à la mathed.
> And if the caller needs to cache the BufferView it should do that with
> a weak_ptr.
Yes.
Anyway you've already shown me how to do this minimally.
boost::shared_ptr<BufferView> const & LyXView::view() const
{ return bufferview_; }
boost::weak_ptr<BufferView> cached_view;
cached_view = bv->owner()->view();
Angus
