Bug 653

*** Bug 651 has been marked as a duplicate of this bug. ***

"My understanding of an min/max function is, that the plot shows the minimum and maximum value of the trace data which represent a single pixel."

That is also my understanding.

"For example, if a single pixel represents one or more periods of a sinus with amplitude +-1, the plot should show a line from -1 to +1. This is not the case in version 2014."

This is what the code looks like (simplified slightly, but with identical results):

        CalcPosition(index, ref X, ref Y);

        if (true/*ShouldDrawPoint(index, ref X, ref Y)*/)
        {
          if (!drawAll && drawAllStyle == Styles.DrawAllPointsStyle.MinMax)
          {
            if (X == oldX)
            {
              if (Y < iMin) iMin = Y;
              if (Y > iMax) iMax = Y;
              sameX = true;

              if (index == lastVisible)
              {
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMax : iMin);
              }
            }
            else
            {
              if (sameX)
              {
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMax : iMin);
              }

              DoDrawPoint(index, X, Y);

              sameX = false;
              iMin = int.MaxValue;
              iMax = int.MinValue;
            }
          } 

So first we calculate the pixel positions of the X and Y values. Then, if the X pixel position is the same as the X pixel position of the last point drawn, we cache the highest and lowest Y pixel positions. Then when the next X pixel position is different, we draw cached X and Y pixel positions and the next X and Y pixel positions.

Please bear in mind that the maximum and minimum Y pixel positions are only cached when the X pixel positions are identical. This does not necessarily mean that the cached Y pixel positions are going to represent the full amplitude of the Y value positions (+-1 in this case).

What I'm exposing here is how the code presently works, and how it fits in to what we both understand as a min/max function. 

I am not at all averse to creating a new DrawAllPointsStyle enum value that fits in with what you need, but in my opinion the min/max function presently works correctly according to what I understand as its design paradigm. Maybe you could help me design this new DrawAllPointsStyle, suggesting a name and which iMax and iMin values you would like to see cached when the X pixel values are identical.

Thank you.

I respond to the last EMail from Chris

Did you have a look at my three pictures of my last report? How do you interpret the shapes of the data, espcially
the asymetry beween upper and lower limit on the last picture, where the data are most horizontally compressed?
Did you try out my code, so that you can see the effect for yourself?

I'll try again to explain why this is really a MAJOR bug:
Take e.g. a sine wave with amplitude +-1. When you narrow down the display, so that the a single periond falls into 1 pixel, the display MUST SHOW a line from -1 to +1, The current version only goes to +1 but not to -1, see also comment at the end

> 
> Comment # 2 on bug 653 from christopher ireland 
> "My understanding of an min/max function is, that the plot shows the minimum
> and maximum value of the trace data which represent a single pixel."
> 
> That is also my understanding.
> 
> "For example, if a single pixel represents one or more periods of a sinus with
> amplitude +-1, the plot should show a line from -1 to +1. This is not the case
> in version 2014."
> 
 
> So first we calculate the pixel positions of the X and Y values. 
> Then, if the X
> pixel position is the same as the X pixel position of the last pointdrawn, we
> cache the highest and lowest Y pixel positions. Then when the next X pixel
> position is different, we draw cached X and Y pixel positions and the next X
> and Y pixel positions.

That seems plausible, however if the min and max fall on the same pixel (as is the case
when e.g a sine wave with many periods is compressed to very few pixels), the full amplitude
-1 and + 1 should be drawn for this pixel.

> 
> Please bear in mind that the maximum and minimum Y pixel positions are only
> cached when the X pixel positions are identical. This does not 
> necessarily mean
> that the cached Y pixel positions are going to represent the full amplitude of
> the Y value positions (+-1 in this case).

I do think so! If on one pixel the whole range between -1 and + 1 is represented,
there should be a (in this case vertical) line from -1 to +1 in the plot.
You apparently only do this for the max value as shown on my picture where data is completely
comprssed: there the upper limmit is OK, but on each pixel the min value is -1, so the
lines should go down there too!

best regards
Hans

Hans,

If I change this code:


              if (index == lastVisible)
              {
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMax : iMin);
              }
            }
            else
            {
              if (sameX)
              {
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMax : iMin);
              }

to this:

              if (index == lastVisible)
              {
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMax : iMin);
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMin : iMax);
              }
            }
            else
            {
              if (sameX)
              {
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMax : iMin);
                DoDrawPoint(index, oldX, GetVertAxis.Inverted ? iMin : iMax);
              }

then I get the plot in the attachment (modified.PNG). I think this could be fix you are looking for - instead of plotting only the (oldX, iMax) value, we are also plotting the (oldX, iMin) value.

Created attachment 146 [details]
Screenshot of compressed chart with modified code

The screenshot from comment #5 is exactly how we think that the plot should look like!
Do you consider to issue this fix soon?

Thank you
Hans

Hans,

The next maintenance release will contain this fix and will become publicly available before the end of this month. 

In the meantime, there is a possibility of deriving your own class from FastLine and overriding the DrawValue() method - this will mean re-implementing some private methods, code for which you could copy by opening TeeChart.dll in Reflector (or equivalent). The fix is nothing other than adding in the two lines I indicated.

Thank you.