[IDV #WAX-506623]: Color bleeds across colored contours unacceptably

Brian,

I spent the day digging into the guts of the VisAD API. I made some progress,
but never really got to the bottom of the problem. I contacted the folks at SSEC
about the issue. That's where I am at. Hopefully, we'll hear back from them
shortly.

Best,

Unidata IDV Support

> Thanks!
> 
> Yes, if a contour is a line of constant value, then its color should be 
> constant.
> 
> That is pure logic, although the numerics of the approximations are 
> understandable.
> 
> Interesting how subtle visad contour segments are.
> 
> Brian Mapes
> This message was typed on a cell phone, please forgive its terseness and any 
> errors.
> 
> > On Jan 21, 2016, at 7:24 PM, Unidata IDV Support <address@hidden> wrote:
> >
> > Brian,
> >
> > This is what matplotlib does. See attached.
> >
> > Best,
> >
> > Unidata IDV Support
> >
> >> Animated gif with numerous, fat contours is clearer
> >>
> >> [cid:76F2C985-9C25-4213-B9FB-DB0E76605743@attlocal.net]
> >>
> >>
> >> On Jan 20, 2016, at 10:37 PM, Brian Mapes 
> >> <address@hidden<mailto:address@hidden>> wrote:
> >>
> >> Oh, I notice that smoothing was on in the contour display. When I set the 
> >> countours unsmoothed,there is still some purple (color mixing), but it is 
> >> all within pixel scale it seems.
> >>
> >>
> >> <PastedGraphic-6.png>
> >>
> >> <IDVcont.colormixing.gif>
> >>
> >>
> >>
> >>
> >> On Jan 20, 2016, at 7:31 PM, Unidata IDV Support 
> >> <address@hidden<mailto:address@hidden>> wrote:
> >>
> >> Brian,
> >>
> >> Yes, I don't think it is a problem with the color scale. This may be a 
> >> subtle
> >> contouring artifact where the contouring algorithm is drifting into areas 
> >> in
> >> shouldn't, if you see what I mean. This is something you would never notice
> >> except in these two-tone scenarios. (I would continue experimenting with 
> >> the
> >> contouring parameters rather than the color table at this point.)
> >>
> >> Can you explain to me how you are deriving DTdy? I am curious how other
> >> plotting packages handle these data.
> >>
> >> Best,
> >>
> >> Unidata IDV Support
> >>
> >> I wonder if it is a matter of having only 6 colors in the table?
> >>
> >> No, I made it more (still half red, half blue) using the + in the color 
> >> table editor, and changed the range, and the contours are still woozy 
> >> purplish .
> >> So it is not some color-interpolation thing I think.
> >>
> >>
> >>
> >> On Jan 20, 2016, at 6:44 PM, Unidata IDV Support 
> >> <address@hidden<mailto:address@hidden>> wrote:
> >>
> >> Brian,
> >>
> >> Thinking about this a bit more, I don't believe there is a bug or problem 
> >> here.
> >>
> >> The color table is being displayed in exactly the manner you wish. In
> >> particular, where you are seeing "blending" are areas where values in the
> >> contour data are actually going back and forth across zero and therefore 
> >> are mixed
> >> purple. It is not the colors that are blending, it is the data themselves 
> >> across
> >> the "zero boundary".
> >>
> >> Another way of looking at this is choosing a two tone black and white color
> >> table. The white will fade into the background. You will see grey as the 
> >> data
> >> approach either side of zero.
> >>
> >> Do you buy this theory? :-)
> >>
> >> Best,
> >>
> >> Unidata IDV Support
> >>
> >> Same issue when I import this new table.
> >> Contours are not uniform red or blue, but mixed purple.
> >> Thicker contour line width helps show it.
> >>
> >> [cid:3EC5425C-85EE-4B22-87E4-1F10BC1A1C6C]
> >>
> >>
> >>
> >>
> >>
> >> On Jan 20, 2016, at 3:16 PM, Unidata IDV Support 
> >> <address@hidden<mailto:address@hidden><mailto:address@hidden>> wrote:
> >>
> >> Brian,
> >>
> >> It is not that the colors are bleeding, it is your color table that 
> >> appears to
> >> be discreet, but apparently is not. Remove all other displays and you will 
> >> see
> >> what I mean. I am unfortunately having trouble loading your bundle though 
> >> I did
> >> get it to load once where I was able to come to that conclusion.
> >>
> >> Try the attached color table which I believe is actually discreet.
> >>
> >> Best,
> >>
> >> Unidata IDV Support
> >>
> >>
> >>
> >> See how some of the contours are purplish? The color bar is purely 
> >> blue/red .
> >>
> >> Is it a bug, or something we have to live with?
> >> Small .zidv bundle attached along with screenshot.
> >>
> >> Brian
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> Ticket Details
> >> ===================
> >> Ticket ID: WAX-506623
> >> Department: Support IDV
> >> Priority: Normal
> >> Status: Open
> >>
> >>
> >> =
> >>
> >>
> >>
> >> *********************************************
> >> Brian Mapes, Professor
> >> Department of Atmospheric Sciences
> >> Meteorology and Physical Oceanography Program
> >> RSMAS, University of Miami
> >> 4600 Rickenbacker Causeway
> >> Miami, FL 33149-1098
> >>
> >> phone: (305) 421-4275
> >> fax: (305) 421-4696
> >> email: address@hidden<mailto:address@hidden>
> >> Web: http://www.rsmas.miami.edu/users/bmapes/
> >> **********************************************
> >
> >
> > Ticket Details
> > ===================
> > Ticket ID: WAX-506623
> > Department: Support IDV
> > Priority: Normal
> > Status: Open
> > <temp4.png>
> > <temp5.png>
> > <matplotlib.txt>
> 
> 


Ticket Details
===================
Ticket ID: WAX-506623
Department: Support IDV
Priority: Normal
Status: Open