Taylor diagram loses y-axis with 2.12 version

[lee1043]

I am testing the Taylor Diagram with 2.12 version but it looks like it is losing y-axis as showing below.

# Create dummy 7 data
import MV2
corr = [.2, .5, .7, .85, .9, .95, .99]
std = [1.6, 1.7, 1.5, 1.2 , .8, .9, .98]
data = MV2.array(zip(std, corr))
data.id = "My Taylor Diagram Data"

import vcs
taylor = vcs.createtaylordiagram()
x=vcs.init()
x.plot(data,taylor)

It also gives warning as below, not sure it is related to lose y-axis.

/Users/lee1043/anaconda2/envs/uvcdat-2.12/lib/python2.7/site-packages/numpy/ma/core.py:6385: MaskedArrayFutureWarning: In the future the default for ma.maximum.reduce will be axis=0, not the current None, to match np.maximum.reduce. Explicitly pass 0 or None to silence this warning.
  return self.reduce(a)
/Users/lee1043/anaconda2/envs/uvcdat-2.12/lib/python2.7/site-packages/numpy/ma/core.py:6385: MaskedArrayFutureWarning: In the future the default for ma.minimum.reduce will be axis=0, not the current None, to match np.minimum.reduce. Explicitly pass 0 or None to silence this warning.
  return self.reduce(a)

[lee1043]

@doutriaux1 Sorry, accidentally clicked assigning part, but maybe it was right?

[doutriaux1]

@lee1043 are you using mesa or not? I generated the jupyter notebooks with mesa and I don't see this issue

[jypeter]

@lee1043 Charles has apparently fixed the Warning (CDAT/genutil#14). I initially also got this warning when using vcs (boxfill plot), but I attributed it to genutil.minmax since it was mentioning both minimum and maximum.

There could of course be some stray minimum and maximum calls not using minmax out there

[lee1043]

@doutriaux1 How I can see if I am using mesa?

$ conda list mesa
# packages in environment at /Users/lee1043/anaconda2/envs/uvcdat-2.12:
#

Not sure this is right way to check. Does mesa come with uvcdat-2.12 conda install?

[doutriaux1]

@lee1043 that's the right way. Ok it means you're not using mesa in this environment. Thanks I'll take a look. @danlipsa any idea?

[doutriaux1]

@lee1043 you can choose with or without mesa, simply add mesalib to the command line or install mesalib after the fact. I will update the instructions page to make this clearer.

[lee1043]

@doutriaux1 Thanks, but it seems like conda install mesalib does not work. Did I miss something?

$ conda install mesalib
Fetching package metadata .........

PackageNotFoundError: Package missing in current osx-64 channels: 
  - mesalib

Close matches found; did you mean one of these?

    mesalib: kealib

[lee1043]

@jypeter @doutriaux1 Thanks for inform. Does the warning correction included in the nightly version?

[zshaheen]

@lee1043 Here's the command to install mesalib:

conda install mesalib -c conda-forge -c uvcdat

[lee1043]

@zshaheen Thanks, it works. @doutriaux1 Having mesalib fixed the issue. Thanks. Curious why it was not included when I was installing uvcdat 2.12.

[doutriaux1]

@lee1043 that's how I made the notebook. I need to figure out why non mesa does not work. @danlipsa any idea?

[danlipsa]

@doutriaux1 Maybe it is clipped by the clip planes in fittoviewport (the clip plane and the axis have the same coordinates, one implementation chooses to clip the other chooses to let it pass)?