What got us in trouble before was the approach we took to infering the interval breaks. If you're willing to work through the logic related to 1d and 2d non-uniform coordinates, we can address this.

A little example of where I'm seeing this go wrong (I think). Here I'm plotting the terrain height variable across the Arctic region, the north pole is somewhere in the middle of the plot. As you can see, there is an odd streak across the plot when using the xarray pcolormesh.

``` Python import cartopy.crs as ccrs

projection class

class Rasm(ccrs.Projection):

def __init__(self):

    proj4_params = {'R': 6371200.0,
                    'lat_0': 90.0,
                    'lat_1': 90,
                    'lat_2': 90,
                    'lon_0': -114.0+360,
                    'proj': 'lcc',
                    'units': 'm',
                    'x_0': 9469302.950316086,
                    'y_0': 6201952.603370549}

    super(Rasm, self).__init__(proj4_params)

@property
def boundary(self):
    coords = ((self.x_limits[0], self.y_limits[0]),(self.x_limits[1], self.y_limits[0]),
              (self.x_limits[1], self.y_limits[1]),(self.x_limits[0], self.y_limits[1]),
              (self.x_limits[0], self.y_limits[0]))

    return ccrs.sgeom.Polygon(coords).exterior

@property
def threshold(self):
    return 100000.0

@property
def x_limits(self):
    return (0, 16423961.103252266)

@property
def y_limits(self):
    return (0, 12228062.194885937)

plt.figure() ax = plt.subplot(projection=Rasm()) elev.plot.pcolormesh('longitude', 'latitude', transform=ccrs.PlateCarree(), vmin=0, vmax=2500, cmap='terrain', add_colorbar=False) ax.set_title('Using xarray plotting')

plt.figure() ax = plt.subplot(projection=Rasm()) ax.pcolormesh(elev.longitude, elev.latitude, elev.to_masked_array(), transform=ccrs.PlateCarree(), cmap='terrain', vmin=0, vmax=2500) ax.set_title('Using pyplot and Cartopy') ```

@shoyer - I'm not sure that problem exists when plotting with carotpy. I'd have to check though.

Yes, my coordinates are quite unevenly spaced (Equal area projection over the north pole).

My suggestion is that we let cartopy handle locating the coordinates when applicable.