BUG: `.str.split()` fails with pyarrow strings

[mattharrison]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd
import io

data = '''name,x,y
matt b,10,20
fred c,20,1
jeff d,2,3
lisa e,5,7
adam f,23,1'''

df = pd.read_csv(io.StringIO(data), engine='pyarrow', dtype_backend='pyarrow')
df.name.str.split()

Issue Description

In Pandas 1, you could split with no parameters (like python does with 'matt h'.split(). This doesn't work with pyarrow types.

It gives a TypeError:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[47], line 12
      4 data = '''name,x,y
      5 matt a,10,20
      6 fred b,20,1
      7 jeff c,2,3
      8 lisa d,5,7
      9 adam e,23,1'''
     11 df = pd.read_csv(io.StringIO(data), engine='pyarrow', dtype_backend='pyarrow')
---> 12 df.name.str.split()

File ~/.envs/menv/lib/python3.10/site-packages/pandas/core/strings/accessor.py:136, in forbid_nonstring_types.<locals>._forbid_nonstring_types.<locals>.wrapper(self, *args, **kwargs)
    131     msg = (
    132         f"Cannot use .str.{func_name} with values of "
    133         f"inferred dtype '{self._inferred_dtype}'."
    134     )
    135     raise TypeError(msg)
--> 136 return func(self, *args, **kwargs)

File ~/.envs/menv/lib/python3.10/site-packages/pandas/core/strings/accessor.py:915, in StringMethods.split(self, pat, n, expand, regex)
    913 if is_re(pat):
    914     regex = True
--> 915 result = self._data.array._str_split(pat, n, expand, regex)
    916 return self._wrap_result(result, returns_string=expand, expand=expand)

File ~/.envs/menv/lib/python3.10/site-packages/pandas/core/arrays/arrow/array.py:2273, in ArrowExtensionArray._str_split(self, pat, n, expand, regex)
   2271 else:
   2272     split_func = pc.split_pattern
-> 2273 return type(self)(split_func(self._pa_array, pat, max_splits=n))

File ~/.envs/menv/lib/python3.10/site-packages/pyarrow/compute.py:258, in _make_generic_wrapper.<locals>.wrapper(memory_pool, options, *args, **kwargs)
    256 else:
    257     option_args = ()
--> 258 options = _handle_options(func_name, options_class, options,
    259                           option_args, kwargs)
    260 if args and isinstance(args[0], Expression):
    261     return Expression._call(func_name, list(args), options)

File ~/.envs/menv/lib/python3.10/site-packages/pyarrow/compute.py:221, in _handle_options(name, options_class, options, args, kwargs)
    216     if options is not None:
    217         raise TypeError(
    218             "Function {!r} called with both an 'options' argument "
    219             "and additional arguments"
    220             .format(name))
--> 221     return options_class(*args, **kwargs)
    223 if options is not None:
    224     if isinstance(options, dict):

File ~/.envs/menv/lib/python3.10/site-packages/pyarrow/_compute.pyx:1894, in pyarrow._compute.SplitPatternOptions.__init__()

File ~/.envs/menv/lib/python3.10/site-packages/pyarrow/_compute.pyx:1872, in pyarrow._compute._SplitPatternOptions._set_options()

File stringsource:15, in string.from_py.__pyx_convert_string_from_py_std__in_string()

TypeError: expected bytes, NoneType found

Expected Behavior

I expect Pandas 1 behavior.

Installed Versions

Details

INSTALLED VERSIONS

commit : e86ed37
python : 3.10.13.final.0
python-bits : 64
OS : Darwin
OS-release : 21.6.0
Version : Darwin Kernel Version 21.6.0: Wed Aug 10 14:28:23 PDT 2022; root:xnu-8020.141.5~2/RELEASE_ARM64_T6000
machine : arm64
processor : arm
byteorder : little
LC_ALL : en_US.UTF-8
LANG : None
LOCALE : en_US.UTF-8

pandas : 2.1.1
numpy : 1.23.5
pytz : 2023.3
dateutil : 2.8.2
setuptools : 67.6.1
pip : 23.2.1
Cython : 3.0.4
pytest : 7.2.0
hypothesis : 6.81.2
sphinx : None
blosc : None
feather : None
xlsxwriter : 3.1.2
lxml.etree : 4.9.2
html5lib : 1.1
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.8.0
pandas_datareader : None
bs4 : 4.11.1
bottleneck : None
dataframe-api-compat: None
fastparquet : None
fsspec : 2023.3.0
gcsfs : None
matplotlib : 3.6.2
numba : 0.56.4
numexpr : 2.8.4
odfpy : None
openpyxl : 3.0.10
pandas_gbq : None
pyarrow : 13.0.0
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.10.0
sqlalchemy : 2.0.21
tables : None
tabulate : 0.9.0
xarray : None
xlrd : 2.0.1
zstandard : None
tzdata : 2023.3
qtpy : None
pyqt5 : None

[phofl]

Just to clarify, you would expect this to split on " " if nothing is given?

[mattharrison]

For Python strings, calling .split() with sep=None:

When set to None (the default value), will split on any whitespace
character (including \n \r \t \f and spaces) and will discard
empty strings from the result.

[lopof]

Reproducible with only pyarrow:

import pyarrow as pa
test = pa.array(['foo t', 'bar', 'baz'], type=pa.string())
pa.compute.split_pattern(test, pattern=None)

Pyarrow needs a string as pattern https://arrow.apache.org/docs/python/generated/pyarrow.compute.split_pattern.html. How does pandas normally handles such cases? Especially if in the pandas documentation is stated:

String or regular expression to split on. If not specified, split on whitespace.

I'm new around here and happy to contribute if there is anything to do here!

[phofl]

I would lean towards patching the arrow behaviour in pandas and to specify a proper dtype

[lopof]

if pat == None:
    return ArrowExtensionArray(pa.array([x.as_py().split(pat, n) for x in self._pa_array]))

What do you think about something like that (here: https://github.com/pandas-dev/pandas/blob/main/pandas/core/arrays/arrow/array.py#L2352)?
What do you mean by "specify a proper dtype"?