pandas-dev/pandas

BUG: series created from pyarrow array with date32 type in 1.5rc0 can be created but not displayed

a-reich opened this issue · 8 comments

a-reich commented

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, pyarrow as pa, numpy as np, datetime as dt
arrow_dt = pa.array([dt.date.fromisoformat('2020-01-01')], type=pa.date32())
ser = pd.Series(arrow_dt, dtype=pd.ArrowDtype(arrow_dt.type))
ser

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\core\series.py", line 1594, in __repr__
    return self.to_string(**repr_params)
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\core\series.py", line 1687, in to_string
    result = formatter.to_string()
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 397, in to_string
    fmt_values = self._get_formatted_values()
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 381, in _get_formatted_values
    return format_array(
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 1328, in format_array
    return fmt_obj.get_result()
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 1359, in get_result
    fmt_values = self._format_strings()
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 1833, in _format_strings
    fmt_values = [formatter(x) for x in values]
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 1833, in <listcomp>
    fmt_values = [formatter(x) for x in values]
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 1802, in <lambda>
    return lambda x: _format_datetime64_dateonly(
  File "C:\Users\asafs\Miniconda3\envs\test\lib\site-packages\pandas\io\formats\format.py", line 1792, in _format_datetime64_dateonly
    return x._date_repr
AttributeError: 'datetime.date' object has no attribute '_date_repr'

Issue Description

I saw in the "what's new" for the new release the functionality for creating general Arrow-backed objects and wanted to try it out. So I created a simple test Arrow Array with date32 type and tried to make a series from it as described, but the series cannot be displayed without erroring. The ser value itself can be created and its array values accessed still.

This issue didn't occur with eg pa.int32 but I did not comprehensively test across Arrow data types.

And while I'm here, thank you to the team for all your work on the awesome Arrow integration/extension features!

Expected Behavior

ser can be displayed in an interactive REPL as usual and repr(ser) does not error.

Installed Versions

INSTALLED VERSIONS

commit : 224458e
python : 3.10.1.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19044
machine : AMD64
processor : Intel64 Family 6 Model 78 Stepping 3, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_United States.1252

pandas : 1.5.0rc0
numpy : 1.23.2
pytz : 2022.2.1
dateutil : 2.8.2
setuptools : 60.2.0
pip : 21.3.1
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 7.31.0
pandas_datareader: None
bs4 : None
bottleneck : None
brotli :
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : None
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 8.0.0
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : None
snappy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
zstandard : None
tzdata : None

mroeschke commented

Thanks for the report! I was able to replicate this issue as well.

mroeschke commented

One point of note: the Series constructor wasn't necessarily intended to accept a pa.array and a pd.ArrowDtype, but I suppose this should be supported as well.

a-reich commented

the Series constructor wasn't necessarily intended to accept a pa.array and a pd.ArrowDtype

Gotcha - what’s the right way to make a pd object from arrow array then? Or is only the “non-arrow data -> pd object construction-> pandas manages arrow conversion” path supported?

mroeschke commented

arrays.ArrowExtensionArray was also added to the public API and would probably be the most explicit way to create a pandas object from arrow array: https://pandas.pydata.org/pandas-docs/version/1.5/reference/api/pandas.arrays.ArrowExtensionArray.html

I'll make sure there's some documentation around using ArrowExtensionArray for this use case.

In [4]: pd.Series(pd.arrays.ArrowExtensionArray(pa.array([1, 2, 3])))
Out[4]:
0   1
1   2
2   3
dtype: int64[pyarrow]
mroeschke commented

Noting that this was fixed by https://github.com/pandas-dev/pandas/pull/48489/files and should be fixed when 1.5 is released. Just needs a unit test

a-reich commented

Is the test needed as simple as just making an appropriate series and calling repr()? Basically, is this issue something a first-time contributor could do?

mroeschke commented

Yeah @a-reich essentially this test needs to be added which hopefully should be a good first time contribution: https://github.com/pandas-dev/pandas/pull/48258/files#diff-290226273f5cd28655f0fa2a4c154cd9909edc6a1f5abee07fec8ab374a9133eR1755

a-reich commented

take