Binned coordinates (pd.Interval) not plottable on the y-axis.

[juseg]

Minimal Code Sample

import numpy as np
import xarray as xr
import matplotlib.pyplot as plt

# create fake data
data = xr.DataArray(np.random.rand(101)).rename('data')

# group by bins and compute sum
bins = np.linspace(0, 1, 11)
data = data.groupby_bins(data, bins).sum()

# plot group means
data.plot()                             # works
data.plot(x='data_bins')                # works
plt.plot((bins[:-1]+bins[1:])/2, data)  # works

# plot with axes transposed
data.plot(y='data_bins')                # TypeError
data.plot.line(y='data_bins')           # TypeError
plt.plot(data, (bins[:-1]+bins[1:])/2)  # workaround

Expected Output

I would like to plot binned data (produced with groupby_bins) with transposed axes, i.e. data on the x-axis, and bins on the y-axis.

Problem Description

Attempting to plot using an array of pd.Interval on the y-axis yields the following TypeError. The error message reports data should be numeric, dates or pd.Interval, which they are.

Traceback (most recent call last):
  File "./groupbybug.py", line 19, in <module>
    data.plot(y='data_bins')                # TypeError
  File "~/.local/lib/python3.7/site-packages/xarray/plot/plot.py", line 465, in __call__
    return plot(self._da, **kwargs)
  File "~/.local/lib/python3.7/site-packages/xarray/plot/plot.py", line 202, in plot
    return plotfunc(darray, **kwargs)
  File "~/.local/lib/python3.7/site-packages/xarray/plot/plot.py", line 323, in line
    _ensure_plottable(xplt_val, yplt_val)
  File "~/.local/lib/python3.7/site-packages/xarray/plot/utils.py", line 512, in _ensure_plottable
    "Plotting requires coordinates to be numeric "
TypeError: Plotting requires coordinates to be numeric or dates of type np.datetime64, datetime.datetime, cftime.datetime or pd.Interval.

Output of xr.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.7.4 (default, Oct 4 2019, 06:57:26)
[GCC 9.2.0]
python-bits: 64
OS: Linux
OS-release: 4.19.84-1-MANJARO
machine: x86_64
processor:
byteorder: little
LC_ALL: None
LANG: fr_CH.utf8
LOCALE: fr_CH.UTF-8
libhdf5: 1.10.5
libnetcdf: 4.7.2

xarray: 0.14.1
pandas: 0.25.2
numpy: 1.17.3
scipy: 1.3.1
netCDF4: 1.5.3
pydap: None
h5netcdf: None
h5py: None
Nio: None
zarr: None
cftime: 1.0.4.2
nc_time_axis: None
PseudoNetCDF: None
rasterio: 1.1.0
cfgrib: None
iris: None
bottleneck: None
dask: 2.5.2
distributed: 0+untagged.1.gd096901.dirty
matplotlib: 3.1.1
cartopy: 0.17.0
seaborn: None
numbagg: None
setuptools: 41.2.0
pip: 19.2.3
conda: None
pytest: 5.2.1
IPython: 7.9.0
sphinx: 2.2.1

[dcherian]

Thanks for the report @juseg. It looks like the logic here needs to be updated:

xarray/xarray/plot/plot.py

Lines 300 to 323 in ff6051d

	if _valid_other_type(xplt.values, [pd.Interval]):
	# Is it a step plot? (see matplotlib.Axes.step)
	if kwargs.get("linestyle", "").startswith("steps-"):
	xplt_val, yplt_val = _interval_to_double_bound_points(
	xplt.values, yplt.values
	)
	# Remove steps-* to be sure that matplotlib is not confused
	kwargs["linestyle"] = (
	kwargs["linestyle"]
	.replace("steps-pre", "")
	.replace("steps-post", "")
	.replace("steps-mid", "")
	)
	if kwargs["linestyle"] == "":
	del kwargs["linestyle"]
	else:
	xplt_val = _interval_to_mid_points(xplt.values)
	yplt_val = yplt.values
	xlabel += "_center"
	else:
	xplt_val = xplt.values
	yplt_val = yplt.values
	_ensure_plottable(xplt_val, yplt_val)

Note that the 2D plotting functions do this:

xarray/xarray/plot/plot.py

Lines 697 to 698 in ff6051d

	xplt, xlab_extra = _resolve_intervals_2dplot(xval, plotfunc.__name__)
	yplt, ylab_extra = _resolve_intervals_2dplot(yval, plotfunc.__name__)

This should be relatively easy fix if you're up for it.