Plotting areacello data fails

[chaosphere2112]

The offending file is:
/cmip5_css02/scratch/cmip5/output1/NCAR/CCSM4/1pctCO2/fx/ocean/fx/r0i0p0/v20130312/areacello/areacello_fx_CCSM4_1pctCO2_r0i0p0.nc

/work/cmip5/fx/fx/areacello/cmip5.CCSM4.1pctCO2.r0i0p0.fx.ocn.fx.areacello.ver-v20130312.latestX.xml
Traceback (most recent call last):
  File "make_evsp.py", line 171, in <module>
    c.plot(areacello)
  File "/usr/local/uvcdat/2016-02-04/lib/python2.7/site-packages/vcs/Canvas.py", line 2424, in plot
    a = self.__plot(arglist, keyargs)
  File "/usr/local/uvcdat/2016-02-04/lib/python2.7/site-packages/vcs/Canvas.py", line 3747, in __plot
    returned_kargs = self.backend.plot(*arglist, **keyargs)
  File "/usr/local/uvcdat/2016-02-04/lib/python2.7/site-packages/vcs/VTKPlots.py", line 533, in plot
    vtk_backend_grid, vtk_backend_geo))
  File "/usr/local/uvcdat/2016-02-04/lib/python2.7/site-packages/vcs/vcsvtk/pipeline2d.py", line 290, in plot
    self._createMaskedDataMapper()
  File "/usr/local/uvcdat/2016-02-04/lib/python2.7/site-packages/vcs/vcsvtk/pipeline2d.py", line 335, in _createMaskedDataMapper
    deep=False)
  File "/usr/local/uvcdat/2016-02-04/lib/python2.7/site-packages/vcs/vcs2vtk.py", line 117, in putMaskOnVTKGrid
    vtkmask.SetNumberOfTuples(attributes2.GetPedigreeIds().GetNumberOfTuples())
AttributeError: 'NoneType' object has no attribute 'GetNumberOfTuples'

The equivalent plot in ferret looks like:

Migrated from: CDAT/cdat#1833

[durack1]

@chaosphere2112 @doutriaux1 here too

[durack1]

@danlipsa this issue is still a problem with 2.10-rc2 the latest master. @doutriaux1 is this solve-able for v2.10.0?
Ferret:

CDAT:

[doutriaux1]

@durack1 I don't think we addressed this at all in this version.

[durack1]

@doutriaux1 @williams13 there are a bunch of open issues that really would be nice to get sorted before the v2.10 release, I know you're pushing hard on getting things locked down before the end of this month, but I think a better product is more important than a faster (buggy) product.. Having said that I suppose v3.0 is around the corner too

[durack1]

@doutriaux1 @danlipsa some more examples:
Ferret:

and CDAT:

[doutriaux1]

@durack1 if you want the ferret version plot data.filled() it will lose dim information and will plot a straight numpy array just like ferret

[durack1]

@doutriaux1 my point of the examples above was that ferret appears to plot things correctly, so land (for areacello) is missing or zero, and the ocean points only have values.. For vcs it doesn't do that..

[durack1]

@danlipsa @aashish24 @doutriaux1 @dnadeau4 I have a heap of these data in a couple of tar.bz2 archives, it would be great to centralize these so that you all have access to the data plotted above that is causing the problems

-rw-r--r--  1 durack1 climate  19M Feb 16  2012 120216_ipcc5_tas_Amon.tar.bz2
-rw-r--r--  1 durack1 climate 377M Feb 28  2012 120218_ipcc5_so_Omon.tar.bz2

Ping me if you want these dropped on a webserver to add to a test suite.. I've also got some of the CMIP6 test grids to play with, and have requested more

[durack1]

@doutriaux1 @danlipsa I really think this should be up on the high priority fix list for the next release, I have just tried to plot the vanilla variable tos from a suite of models, and the first four I plotted looked like this:




That is to say, everyone of them was plotted incorrectly

And for reference:
anaconda2/envs/uvcdatNightlynox/conda-meta/
vtk-cdat-nox-7.1.0.2017.3.23.73fb094d17ccdd594a6abf9ff8732149c88dc2e6-py27_0.json
vcs-nox-2017.4.4.c22e1d924a0145ef526b56a8ca123484adfd660a-0.json
vcsaddons-nox-2.8-py27_2.json