[vtk-developers] Charts regressions

[Bill Lorensen]

Thanks. You are correct I did not mean to send it to the arb.

And, I guess I forgot I had mentioned this before.

I'll look forward to the new code.

Bill

On Sun, Oct 17, 2010 at 6:24 PM, Marcus D. Hanwell
<marcus.hanwell at kitware.com> wrote:
> On Sun, Oct 17, 2010 at 3:24 PM, Bill Lorensen <bill.lorensen at gmail.com>
> wrote:
>>
>> Sorry, I did mean to send this to arb at vtk.org. gmail was overly agressive.
>
> Do you mean you didn't mean to send this to arb at ...? I was wondering why it
> was an ARB issue.
>>
>> Marcus,
>>
>> I noticed a while back on the wiki examples that the y axis label does
>> not fit on the window for vtk charts. I'm not sure when this
>> regression occurred since we don't update VTK every night.
>
> I remember you emailing me about this, and I gave you an overview of what I
> was doing at the time. Due to my travel schedule, and some other events, my
> topic branch to address this has been a little delayed. I had most of it
> staged last week, but was working on fixing a few more regressions before
> merging it. I think I got most of it finished yesterday.
>>
>> I looked at the VTKData git repository. On 8/23, many Charts baselines
>> were changed.
>>
>> Please compare this baseline:
>>
>> http://vtk.org/gitweb?p=VTKData.git;a=blob;f=Baseline/Charts/TestScatterPlot.png;hb=e293ebc8db06c21a574f42911b27b9b8c92c6a95
>> with this baseline
>>
>> http://vtk.org/gitweb?p=VTKData.git;a=blob;f=Baseline/Charts/TestScatterPlot.png;hb=fcdb83896ab1de8ec8efcb10bf5a079f2c7219f7
>>
>> The former looks correct to me. The Y-Axis lies completely in the
>> window, while the later looks wrong to me. The Y-Axis label is
>> truncated.
>
> The problem is that it was only correct for short strings, for longer ones
> they would still fall off the edge. The newer code uses font metrics to
> determine the chart layout and works for all font sizes and string lengths.
> The issue was that only the Qt label render strategy was respecting the
> rotation, but the freetype strategy was returning the wrong metrics and not
> rotating the font.
>>
>> This occurs consistently on all of the baselines that hav ea Y-Axis labe;.
>>
> It looks a lot better when using the Qt label render strategy, and as I said
> I was working on getting this into shape with the freetype rendering (which
> is default for most VTK applications. The StringToImage topic branch
> contains a lot of these changes, and there are a few more I will push
> shortly and integrate tomorrow.
> I didn't want to cause all of the baseline images to need multiple updates,
> but once I merge the changes in tomorrow many tests will require updates. I
> think the new baselines look far better, and I am hoping to land some more
> string rendering improvements in master over the next few weeks. I did not
> intend to leave the axis rendering like this for a release.
> I hope that clears things up, these changes would have likely made it in
> last week if it weren't for some other large changes also being integrated
> along with some other timing issues.
> Marcus