skip to navigation
skip to content

feather-format 0.3.1

Python interface to the Apache Arrow-based Feather File Format

## Python interface to the Apache Arrow-based Feather File Format

Feather efficiently stores pandas DataFrame objects on disk.

## Installing

pip install feather-format

From [conda-forge][1]:

conda install feather-format -c conda-forge

#### Mac notes

Anaconda uses a default 10.5 deployment target which does not have C++11
properly available. This can be fixed by setting:


Deployments targets as early as 10.7 can be used if the compiler supports
C++11 and the correct mode is selected. For example using the following:

export CFLAGS="${CXXFLAGS} -stdlib=libc++ -std=c++11"
export CXXFLAGS="${CXXFLAGS} -stdlib=libc++ -std=c++11"

This may be necessary in some other OS X environments.

## Build

Building Feather requires a C++11 compiler. We've simplified the PyPI packaging
to include libfeather (the C++ core library) to be built statically as part of
the Python extension build, but this may change in the future.

### Static builds for easier packaging

At the moment, the libfeather sources are being built and linked with the
Cython extension, rather than building the `libfeather` shared library and
linking to that.

While we continue to do this, building from source requires you to symlink (or
copy) the C++ sources. See:

# Symlink the C++ library for the static build
ln -s ../cpp/src src
python build

# To install it locally
python install

# Source distribution
python sdist

To change this and instead link to an installed ``, look in
`` and make the following change:


## Limitations

Some features of pandas are not supported in Feather:

* Non-string column names
* Row indexes
* Object-type columns with non-homogeneous data

File Type Py Version Uploaded on Size
feather-format-0.3.1.tar.gz (md5) Source 2016-10-29 933KB