Weird streaks with 2D ocean data about proplot HOT 5 CLOSED

Fixing this one was crazy. It turns out that this artifact is related to this particular dataset and occurs in matplotlib normally, but cartopy patches that error in some complex and mysterious way, but I was doing something that ruined cartopy's patch. Try these examples after updating (the s is just to make the dataset smaller):

import matplotlib.pyplot as plt
import proplot as plot
import climpred as cp
s = 2
# Load data
data = cp.tutorial.load_dataset('MPI-control-3D')
data = data.isel(time=0).tos
# Does not generate artifact
f, ax = plot.subplots(width=5, proj='pcarree', axcolorbar='r')
ax.pcolormesh(data.lon[::s,::s], data.lat[::s,::s], data[::s,::s], levels=20, colorbar='r')
# Basemap, generates artifact
# f, ax = plot.subplots(width=5, basemap=True, proj='pcarree', axcolorbar='r')
# ax.pcolormesh(data.lon[::s,::s], data.lat[::s,::s], data[::s,::s], levels=20, colorbar='r')
# On Cartesian axes/without projection, generates artifact
# f, ax = plot.subplots(width=5, axcolorbar='r')
# ax.pcolor(data.lon[::s,::s], data.lat[::s,::s], data[::s,::s], levels=20, colorbar='r')
# With pyplot, generates artifact
# f, ax = plt.subplots()
# ax.pcolor(data.lon[::s,::s], data.lat[::s,::s], data[::s,::s])

The cartopy patch evidently works by masking content in your data array. My BinNorm normalizer that I apply to all colormap objects was deleting/ignoring this mask (see the ma.getmaskarray lines in commit 222392e).

Also the latest ProPlot versions fix the weird font bug we came across the other day. It had to do with me adding a font .ttf file with a style unrecognized by matplotlib/that matplotlib interpreted as 'normal'.

I also added your polar projection zoom-ins to getting started, fixed a tight-layout bug that appeard with zoomed-in projections, fixed a bug related to the circular bounding box limits, fixed a bug where the facecolor attribute for map axes would get reset, added the 'nplaea', 'splaea', 'npaeqd', and 'spaeqd' projections to cartopy like in basemap, and various other bugfixes/cleanup tasks. It's all a blur.

Still just pushing to master at will.... will get into your workflow once I set up CI with some smaller tests.

from proplot.

Also thanks for posting this issue. It definitely needed to be fixed and would probably have appeared mysteriously for various other users with various other datasets. I was exhausted when I posted this.

from proplot.

@lukelbd
Awesome, thanks for all the updates here. So in summary, only use the cartopy projection format for these 2D maps? Looks like all the other options (including Basemap) generate those artifacts.

from proplot.

Yep that's right. My guess is it has to do with your data graticule crossing the edge of the map; cartopy takes pains to fix issues associated with stuff running over the map edge.

from proplot.

Having the Greenland-pole grid on the ocean always causes issues. With 2D lat/lon you can't really contour properly. I'm used to these problems, but they shouldn't occur on pcolormesh. Glad you were able to fix it!

from proplot.