bencode2

A fast and correct bencode serialize/deserialize library

introduction

Why yet another bencode package in python?

because I need a bencode library:

1. Correct

It should fully validate its inputs, both encoded bencode bytes, or python object to be encoded.

And it should not decode bencode bytes to str by default.

Bencode doesn't have a utf-8 str type, only bytes, so many decoder try to decode bytes to str and fallback to bytes, this package won't, it parse bencode bytes value as python bytes.

It may be attempting to parse all dictionary keys as string, but for BitTorrent v2 torrent, the keys in pieces root dictionary is still sha256 hash instead of ascii/utf-8 string.

If you prefer string as dictionary keys, write a dedicated function to convert parsing result.

Also be careful! Even file name or torrent name may not be valid utf-8 string.

2. Fast enough

this package is written with c++ in CPython.

3. still cross implement

This package sill have a pure python wheel bencode2-${version}-py3-none-any.whl wheel on pypi.

Which means you can still use it in non-cpython python with same behavior.

install

pip install bencode2

basic usage

import bencode2

assert bencode2.bdecode(b"d4:spaml1:a1:bee") == {b"spam": [b"a", b"b"]}

assert bencode2.bencode({'hello': 'world'}) == b'd5:hello5:worlde'

Decoding

bencode type	python type
integer	int
string	bytes
array	list
dictionary	dict

bencode have 4 native types, integer, string, array and dictionary.

This package will decode integer to int, array to list and dictionary to dict.

Because bencode string is not defined as utf-8 string, and will contain raw bytes bencode2 will decode bencode string to python bytes.

Encoding

python type	bencode type
bool	integer 0/1
int, enum.IntEnum	integer
str, enum.StrEnum	string
bytes, bytearray,memoryview	string
list, tuple, NamedTuple	array
dict, OrderedDict	dictionary
types.MaapingProxy	dictionary
dataclasses	dictionary

free threading

bencode2 have a free threading wheel on pypi, build with GIL disabled.

When encoding or decoding, it will not acquire GIL and may call non-thread-safy c-api, which mean it's the caller's responsibility to ensure thread safety.

When calling bencode, it's safe to encode same object in multiple threading, but it's not safe to encoding a object and change it in another thread at same time.

Also, when decoding, bytes objects are immutable so it's safe to be used in multiple threading, but memoryview and bytearray maybe not, please make sure underlay data doesn't change when decoding.

Development

This project use meson for building.

For testing pure python library, make sure all so/pyd files in src/bencode2 are removed, then run PYTHONPATH=src pytest --assert-pkg-compiled=false.

For testing native extension, meson-python doesn't provide same function with python setup.py build_ext --inplace.

So you will need to run command like this:

meson setup build -Dbuildtype=release -Dpython.allow_limited_api=false
meson compile -C build
ninja -C build copy

ninja will need to build so/pyd with meson and copy it to src/bencode2,

then run tests with PYTHONPATH=src pytest --assert-pkg-compiled=true.