Perspective reading LZ4 compressed RecordBatches.

[sinistersnare]

Now you can drop a file created by

import pandas as pd
import pyarrow.feather as feather
data = [['a', 10], ['b', 15], ['c', 14]]
df = pd.DataFrame(data, columns=['Name', 'Age'])
# This defaults to compressing with LZ4.
feather.write_feather(df, 'lz4_compressed_data.feather')

[texodus]

Thanks for the PR! A few additional issues to the comments below:

• Doesn't compile on Windows.
• Needs tests.

Closing this in favor of #2339 which addresses these issues and adds write support, benchmarks and documentation.

[texodus]

We should IMO move the arrow_single_threaded_reader.cpp files here also from cpp/perspective/CMakeLists.txt

[texodus]

We don't need this file - there is no corresponding call to find_package, also this library is explicitly built statically.