[vtk-developers] Parallel coordinates chart looking nice

[Marcus D. Hanwell]

Hi Eric,

On Tue, Apr 6, 2010 at 11:33 AM, Eric E. Monson <emonson at cs.duke.edu> wrote:

> Hey Marcus,
>
> Thanks for all the updates. Yes, it looks pretty with the transparency -- I
> like it! (Might look nice with some transparency in the selected lines,
> too.) The annotation link seems to be firing just fine, so that's great.
>

That is great to hear as I had not had chance to test it out yet.

>
> I'll have some comments about the ParaView version after a while, but I
> came across an interaction bug with the standard
> vtkChartParallelCoordinates.
>
> (As a note, add view->GetInteractor()->Initialize(); before the Start()
> call at the end of your Cxx test so the test will run properly in
> Interactive mode, at least on my machine.)
>

You can simply call ./bin/ChartsCxxTests TestPCPlot -I and it will cause the
interactor to get initialized (invoked in if(retVal ==
vtkRegressionTester::DO_INTERACTOR)). I think that this is the standard way
of optionally invoking an interactor in a test.

>
> To see the problem, select a range on one of the axes. Then, without
> clicking to clear the selection, drag another selection on the same axis. If
> the second red rectangle overlaps part of the first one, the overlapped
> selection will be retained, but any new lines you should have selected will
> not be highlighted. If you draw the second box not overlapping the first at
> all, no new lines will be selected. It's sort of like it's treating it as a
> separate axis selection and doing an AND on it.
>

This was on my list of things to fix - just committed something now that
takes care of it. I think it behaves much more nicely now.

>
> Then, just as a reminder, there are still the bugs where the red rectangles
> don't stay with the axes when the window is resized, and if you make the
> window too short the axis value ranges (except the first one) start to
> balloon up in value.
>

The red rectangles are now more tightly coupled to the axis positions, which
simplified some of the logic too. That is also committed. The issue with the
axis value range ballooning I will need to look into. I can reproduce it
here, and the pixel range is used in the calculation of the axis tick
locations. I may need to add a conditional into that calculation to prevent
this from happening when the axis is very space constrained.

>
> Let me know if it would work out better for you if I put these things on
> the bug tracker as I find them, and also if you can't reproduce them.
>

Right now outlining issues by email is totally fine by me as I am actively
working on it, and fleshing out the functionality. For things that might
take a little longer bug reports would be better. Still need to check
everything is in there that Berk needs, and also add some GUI to ParaView to
control chart parameters.

Thanks,

Marcus

On Apr 6, 2010, at 10:37 AM, Marcus D. Hanwell wrote:
>
> Hi Eric,
>
> Just a quick note - I have updated the logic in the charts and the
> annotation link's signal should be correctly fired now. I have also added
> transparency to the line drawing at Jeff's suggestion and this looks much
> better. I should add a couple of pens where the user can control the look of
> the lines through the public API.
>
> Marcus
>
> On Mon, Apr 5, 2010 at 5:02 PM, Marcus D. Hanwell <
> marcus.hanwell at kitware.com> wrote:
>
>> Hi Eric,
>>
>> Just to let you know I have bidirectional selections working in ParaView,
>> the one wrinkle is that the vtkAnnotationLink is not invoking its changed
>> event. You could listen to the chart's selection changed event, and I will
>> talk to Jeff about integrating it better with the vtkAnnotationLink API.
>>
>> I have tested it in ParaView with a line chart, spread sheet, 3D view
>> (line probe) and the parallel coordinate view. It is working quite well, but
>> there are still a few bugs I need to track down. It requires more selection
>> modes too, as well as better interaction with axes etc.
>>
>> Marcus
>>
>> On Mon, Apr 5, 2010 at 12:39 PM, Eric E. Monson <emonson at cs.duke.edu>wrote:
>>
>>> Hey Marcus,
>>>
>>> That sounds great. If you could give me a heads-up when you commit the
>>> bi-directional selections I'd really appreciate it.
>>>
>>> Talk to you later,
>>> -Eric
>>>
>>>
>>> On Apr 5, 2010, at 11:53 AM, Marcus D. Hanwell wrote:
>>>
>>> Hi Eric,
>>>
>>> On Mon, Apr 5, 2010 at 11:45 AM, Eric E. Monson <emonson at cs.duke.edu>wrote:
>>>
>>>> Hey Marcus,
>>>>
>>>> It looks like you're making some good progress on this, and I see that
>>>> it shows up as a view in ParaView now, too!
>>>>
>>>> I don't want to be bugging you about it, but it looks like so far the
>>>> AnnotationLink stuff is built in so that outside links will be reflected in
>>>> the parallel coordinates chart, but selections made within the parallel
>>>> coordinates chart do not update the internal annotation link selection list
>>>> (and do not fire AnnotationChangedEvent). Is this something that you plan to
>>>> add soon? (Or, am I misreading the code as it stands?)
>>>>
>>>
>>> Don't worry about bugging me, you are reading the code correctly. I have
>>> some local changes where I am working on getting the selections back out and
>>> into ParaView. I have not got the last part working (into ParaView), but
>>> hope to have something I can commit very soon.
>>>
>>>>
>>>> Again, I don't want to be bugging you, but I am curious whether I should
>>>> spend time trying to plug this into my personal copy since I need it for my
>>>> own use, or if you'll be addressing it soon in the main repository?
>>>>
>>>
>>> I am working on this right now. I will let you know once I have committed
>>> something that should work, it would be great to have feedback on this final
>>> part once it is in. Then I will have some rendering issues to work out, but
>>> selection should work in and out.
>>>
>>> Thanks,
>>>
>>> Marcus
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20100406/7f3d0096/attachment.html>