Re: [lmi] switch census view to use wxDataViewCtrl

[Greg Chicares]

On 2011-08-12 14:00Z, Vadim Zeitlin wrote:
> On Thu, 11 Aug 2011 12:54:17 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> (1) To ascertain whether the selection is nonempty, I did:
> GC> 
> GC>     wxDataViewItemArray selection;
> GC>     unsigned int n_sel_items = list_window_->GetSelections(selection);
> GC>     e.Enable(0 < n_sel_items);
> GC> 
> GC> Is there a better way? It seems wasteful to populate a wxDataViewItemArray
> GC> that isn't used.
> 
>  Yes, indeed, and there are additional inefficiencies in both the MSW and
> GTK versions too (MSW one passes the array of selections by value and GTK
> one calls extra GTK functions for each selected item). I already noticed
> the absence of HasSelections() or GetSelectedItemsCount() in wxDVC and
> thought about adding it but didn't find time to do this before. If it can
> be useful to LMI it would be a sufficient motivation for me to finally do
> it...

Yes, it would be useful to lmi, even if only because it increases clarity.

[...GetSelection().IsOK() detects whether *exactly* one cell is selected...]
> GC> 
> GC> 
> http://docs.wxwidgets.org/trunk/classwx_data_view_ctrl.html#71cd6bc33658d944c65062f789a82f19
> GC> | virtual wxDataViewItem wxDataViewCtrl::GetSelection() const [virtual]
> GC> | Returns first selected item or an invalid item if none is selected.
> GC> 
> GC> If two cells are selected, shouldn't that return the first one, which
> GC> should pass the IsOK() test?
> 
>  The problem is that the real code behaviour is not the documented one in
> the generic version (used by wxMSW), even though it does behave like this
> in wxGTK.

Then lmi wouldn't behave as intended under wxGTK. We don't use a GTK build
in production at the moment, but sometimes things change for the better.

>       http://thread.gmane.org/gmane.comp.lib.wxwidgets.devel/132014
[...]
> I'll post here again when it's done or if we decide not to do it -- in
> which case the code above would need to be changed as then I'd change the
> generic version to behave as documented.

Great. Thanks.