ray.init() fails with AttributeError: module 'pandas.core' has no attribute 'arrays'

[robertnishihara]

In the current master, ray.init() fails as follows (both on Mac and Linux) (both with pandas 0.22.0 as well as 0.20.3). It succeeds with 0.23.4.

This is most likely due to #3765 or #3760.

In [1]: import ray
ray---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-1-d310bc7e5061> in <module>()
----> 1 import ray

~/ray_scratch/ray/python/ray/__init__.py in <module>()
     26 
     27 try:
---> 28     import pyarrow  # noqa: F401
     29 except ImportError as e:
     30     if ((hasattr(e, "msg") and isinstance(e.msg, str)

~/ray_scratch/ray/python/ray/pyarrow_files/pyarrow/__init__.py in <module>()
    167 localfs = LocalFileSystem.get_instance()
    168 
--> 169 from pyarrow.serialization import (default_serialization_context,
    170                                    register_default_serialization_handlers,
    171                                    register_torch_serialization_handlers)

~/ray_scratch/ray/python/ray/pyarrow_files/pyarrow/serialization.py in <module>()
    303 
    304 
--> 305 register_default_serialization_handlers(_default_serialization_context)

~/ray_scratch/ray/python/ray/pyarrow_files/pyarrow/serialization.py in register_default_serialization_handlers(serialization_context)
    294         custom_deserializer=_deserialize_pyarrow_table)
    295 
--> 296     _register_custom_pandas_handlers(serialization_context)
    297 
    298 

~/ray_scratch/ray/python/ray/pyarrow_files/pyarrow/serialization.py in _register_custom_pandas_handlers(context)
    175         custom_deserializer=_load_pickle_from_buffer)
    176 
--> 177     if hasattr(pd.core.arrays, 'interval'):
    178         context.register_type(
    179             pd.core.arrays.interval.IntervalArray,

AttributeError: module 'pandas.core' has no attribute 'arrays'

[robertnishihara]

Should be fixed by #3786.