On Wed, Apr 26, 2006 at 09:49:01AM +0200, Abdelrazak Younes wrote:
> Andre Poenitz a écrit :
> >On Mon, Apr 17, 2006 at 07:54:04PM +0200, Abdelrazak Younes wrote:
> >>+void TocModel::populate(toc::Toc const & toc_list)
> >>+{
> >>+ clear();
> >>+
> >>+ if (toc_list.empty())
> >>+ return;
> >>+
> >>+ int current_row;
> >>+ QModelIndex top_level_item;
> >>+
> >>+ toc::Toc::const_iterator iter = toc_list.begin();
> >>+ toc::Toc::const_iterator end = toc_list.end();
> >>+
> >>+ insertColumns(0, 1);
> >
> >Indentation.
>
> ?? This line is at the same indent level as the previous one.
Not in my vi. The previous lines uses a tab for indentation
(as they should) and the 'insertColums' line uses four spaces.
Believe it or not, that _is_ a difference.
> >>+void TocModel::populate(toc::Toc::const_iterator & iter,
> >>+ toc::Toc::const_iterator
> >>const & end,
> >>+ QModelIndex const & parent)
> >>+{
> >>+ int curdepth = iter->depth + 1;
> >>+ int current_row;
> >>+ QModelIndex child_item;
> >>+
> >>+ insertColumns(0, 1, parent);
> >
> >And here.
>
> Ditto.
Ditto.
> >>+ current_row = rowCount(parent);
> >>+ insertRows(current_row, 1, parent);
> >>+ child_item = QStandardItemModel::index(current_row, 0,
> >>parent);
> >>+ //setData(child_item, toqstr(iter->str));
> >>+ setData(child_item, toqstr(iter->str), Qt::DisplayRole);
> >>+ item_map_.insert(make_pair(child_item, *iter));
> >>+ index_map_.insert(make_pair(
> >>+ iter->str.substr(iter->str.find(' ') + 1),
> >>child_item));
> >
> >I'd find
> >
> >+ index_map_[iter->str.substr(iter->str.find(' ') + 1)] =
> >child_item;
> >
> >more readable.
>
> Me too but I was worried that a section with the same text could get
> replaced. The last version in svn do not take str anyway... but still
> make use of make_pair.
The two versions are semantically equivalent. In fact, operator[] is
just syntactic sugar. But it helps...
> I've change that in my local version, I'll commit when I have
> something more substancial to add. Or feel free to change it if you
> want.
>
> >Apart from that I never completely understood what invalidates a
> >QModelIndex. All I know that it is very prudent not to rely on its
> >stability, so maybe that's not the best approach.
>
> What is not the best approach?
Storing QModelIndex over 'a long period of time'.
Qt dcumentation is pretty vague in this area, and I had crashs which
went away after chosing something more robust.
If you are sure that your way is safe, fine. I just wanted to warn...
Andre'
