Bencode.cy

A fast bencode implementation based on Cython, supporting Python 3.9+.

The bencode functions are 15% to 230% faster than other compiled libraries, and the decode methods are 40% to 320% faster than other compiled libraries.

On average, they outperform pure Python versions by 250%.

Based on my tests, it is currently among the faster bencode libraries available.

It is written using Cython's pure Python syntax, so it can run directly without Cython compilation. Of course, performance is significantly better after compilation.

Installation

pip install bencode.cy

or, if you use uv (it might be better):

uv add bencode.cy

Usage Examples

from bencode import bencode, bdecode, compiled

assert True == compiled # Check if the code has been compiled with Cython

assert bencode("WWWWWW") == b'6:WWWWWW'
assert bencode(233) == b'i233e'

with open("debian-8.3.0-amd64-netinst.iso.torrent", "rb") as f:
    torrent = bdecode(f.read())
    print(torrent[b'announce'])

Speed Benchmark

bencode.cy.encode and bencode.cy.decode are the Cython-compiled versions of this library. bencode.cy.encodepy and bencode.cy.decodepy are the Python versions. They are the same code—one compiled, one not—which is quite interesting.

bencoder.pyx.encode and bencoder.pyx.decode are the encoding/decoding functions from the bencoder.pyx library. However, it currently only supports up to Python 3.11 and is slightly slower, though I learned a lot from its code.

fastbencode.encode and fastbencode.decode are from another excellent library, fastbencode, which is written in Rust.

better_bencode.encode and better_bencode.decode are a C version of the bencoder library from better_bencode, but I don't know why it is so slow. Maybe Python 3.11 on macOS wasn't compiled to C.

bencodepy.encode and bdecodepy.decode are part of a pure Python version of the bencoder library from BencodePy.

From the comparison, this library is currently among the fastest options. You can see more detailed results below.

deocde benchmar result

encode benchmar result

Test Environment: Python 3.11.13, MacBook Air M2, macOS 26.01, October 2025