import arviz throws an error in dependency "xarray"

[aramakus]

After installing arviz on Ubuntu 18.04 using either pip or conda an error is thrown in an attempt to import arviz:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-4-980441a934be> in <module>
----> 1 import arviz
      2 
      3 #az.plot_trace(mcmc.get_samples(), filter_vars="like")

~/anaconda3/envs/jax/lib/python3.8/site-packages/arviz/__init__.py in <module>
     26 
     27 from .rcparams import rcParams, rc_context
---> 28 from .data import *
     29 from .stats import *
     30 from .plots import *

~/anaconda3/envs/jax/lib/python3.8/site-packages/arviz/data/__init__.py in <module>
      1 """Code for loading and manipulating data structures."""
----> 2 from .inference_data import InferenceData, concat
      3 from .io_netcdf import from_netcdf, to_netcdf
      4 from .datasets import load_arviz_data, list_datasets, clear_data_home
      5 from .base import numpy_to_data_array, dict_to_dataset, CoordSpec, DimSpec

~/anaconda3/envs/jax/lib/python3.8/site-packages/arviz/data/inference_data.py in <module>
     13 from xarray.core.options import OPTIONS
     14 
---> 15 from ..utils import _subset_list, HtmlTemplate
     16 from ..rcparams import rcParams
     17 

~/anaconda3/envs/jax/lib/python3.8/site-packages/arviz/utils.py in <module>
    668 
    669 
--> 670 class HtmlTemplate:
    671     """Contain html templates for InferenceData repr."""
    672 

~/anaconda3/envs/jax/lib/python3.8/site-packages/arviz/utils.py in HtmlTemplate()
    694             """
    695     specific_style = ".xr-wrap{width:700px!important;}"
--> 696     css_template = f"<style> {xr.core.formatting_html.CSS_STYLE}{specific_style} </style>"

AttributeError: module 'xarray.core.formatting_html' has no attribute 'CSS_STYLE'

xarray is installed with arviz as a dependency, so my system did not have any other version of the package.

[aramakus]

UPDATE : latest version, available on Github works fine.

[michaelosthege]

linking the related PR #1389 that fixed this on master (for now)