Parquet: Support Encoding::BYTE_STREAM_SPLIT

[simonvandel]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
I would like to evaluate whether using the BYTE_STREAM_SPLIT encoding helps a Float64 column compress better. But it seems like it is not supported yet:

arrow-rs/parquet/src/encodings/encoding/mod.rs

Line 90 in 93484a1

e => return Err(nyi_err!("Encoding {} is not supported", e)),

Describe the solution you'd like
An implementation of the encoding. Even a naive, non-optimized version would resolve this issue. The implementation can be improved iteratively.

Describe alternatives you've considered
PyArrow seems to support it, but I would really like to stay within the Rust world.

Additional context

• Parquet format description here: https://github.com/apache/parquet-format/blob/master/Encodings.md#byte-stream-split-byte_stream_split--9
• The scalar impl in the C++ library is here: https://github.com/apache/arrow/blob/0bf777a5952be012e41f5b1ad443d4fec38e6f5a/cpp/src/arrow/util/byte_stream_split.h#L579-L602 . They also have SIMD variations, which will be more involved to port.

[simonvandel]

I'll give it a go myself, if possible.

[Weijun-H]

Hello @simonvandel, I was wondering if you're currently working on this task. If not, I would be happy to take on the project.

[simonvandel]

Hi @Weijun-H I have a working version that I can clean up a bit. I'll update you here