Feather Format

0.4.1 verified Fri May 01 auth: no python maintenance

Simple wrapper library to the Apache Arrow-based Feather File Format. Current version 0.4.1. Development is in maintenance mode; users are encouraged to use pyarrow directly.

pip install feather-format

Common errors

error ModuleNotFoundError: No module named 'feather' ↓

cause You installed `feather` (a different, unrelated package) instead of `feather-format`.

fix

Run pip uninstall feather then pip install feather-format or pip install pyarrow.

error AttributeError: module 'feather' has no attribute 'read_dataframe' ↓

cause Common name collision: `import feather` often imports the wrong `feather` package (an IRC bot library) instead of feather-format.

fix

Use explicit import: from feather import read_dataframe ensures you get the right module.

error feather.libfeather.FeatherError: Unsupported Feather version ↓

cause The file was written by a newer Feather specification (V2) that feather-format cannot read. feather-format only supports V1.

fix

Use pyarrow.feather.read_feather to read modern Feather files: pip install pyarrow then from pyarrow.feather import read_feather.

Warnings

deprecated Feather-format (library) is deprecated; use pyarrow.feather directly. Development moved to Apache Arrow. ↓

fix Replace `from feather import read_dataframe` with `from pyarrow.feather import read_feather` and adjust function names accordingly.

breaking Version 0.4.0 switched from a custom C++ implementation to pyarrow-based backend. Backward compatibility with files written by older versions is maintained but the API changed slightly. ↓

fix Ensure you use the new API (`from feather import read_dataframe`). Old `feather.read_dataframe` still works but is less explicit.

gotcha The feather-format library from PyPI is NOT the same as the 'feather' R package or the 'feather' specification. Use pyarrow for updated features. ↓

fix Use `pip install pyarrow` and import from `pyarrow.feather` for the latest Feather format support.

Imports

read_dataframe
wrong
```
import feather; feather.read_dataframe(...)
```
correct
```
from feather import read_dataframe
```
Legacy API was feather.read_dataframe; the recommended import is from feather import read_dataframe to avoid namespace confusion.

write_dataframe

wrong

import feather; feather.write_dataframe(...)

correct

from feather import write_dataframe

Same as read_dataframe.

Quickstart

Write and read a Feather file using feather-format.

import pandas as pd
from feather import read_dataframe, write_dataframe

df = pd.DataFrame({'x': [1,2,3], 'y': ['a','b','c']})
write_dataframe(df, 'test.feather')
df2 = read_dataframe('test.feather')
print(df2)