Pandas 1.3.0 DataFrame repr fails due to IndexError

[Sam Davis]

Pandas 1.3.0 passes an Ellipsis object to __getitem__ when the user prints it. This causes an IndexError to be thrown as __getitem__ does not, and is not specified to, support Ellipsis indexing.

Note that this only happens when the array is "large" so it must be interacting with printoptions somewhere.

It displays okay on Pandas 1.2.5 (error not thrown).

Example:

import fletcher as fr
import numpy as np
import pandas as pd
import pyarrow as pa


array = pa.array(np.arange(1000))

# ok
print(array)

# ok
fl_array = fr.FletcherContinuousArray(array)
print(fl_array)

# ok
ser = pd.Series(fl_array)
print(ser)

# fails, IndexError
df = pd.DataFrame({"foo": fl_array})
print(df)

Version information:

$ python --version
Python 3.8.0

$ for lib in fletcher numpy pandas pyarrow; do python -c "import $lib; print('$lib', $lib.__version__)"; done
fletcher 0.7.2
numpy 1.20.3
pandas 1.3.0
pyarrow 4.0.1

[Sam Davis]

Went to report on pandas GitHub also and discovered the issue reported there: pandas-dev/pandas#42430

[Uwe L. Korn]

This project has been archived as development has ceased around 2021.
With the support of Apache Arrow-backed extension arrays in pandas, the major goal of this project has been fulfilled.