cbrrr v1.0

[DavidBuchanan314]

Here's all the things I want to complete before it's ready for a v1.0 release:

• Implement parsing into "atjson" format ($link, $bytes)
• Implement base32 encode/decode in C (to optimise perf in atjson mode)
• Clean up any (important) TODOs in the source
• Rename "parse" to "decode"
• Clean up the CID class, audit correctness
• Robust testing, within a proper testing framework Edit: I'm marking this as done, but there's lots of room for improvement
• Set up a fuzz harness
• Automate testing on GHA
• Figure out some standard benchmarks, show off perf vs dag_cbor, libipld - done at https://github.com/DavidBuchanan314/dag-cbor-benchmark although some more atproto-specific tests would be good
• Documentation - the API is simple enough that a good README should be enough
• Address Don't use EOFError to signal truncated input #5
• Get it on pypi

Deferred to 1.x:

• More test coverage
• GHA testing on macos and windows, and aarch64 if I can figure that out.
• GHA cross-platform builds and pypi upload
• Test edge-cases that might arise on 32-bit platforms (and fix any related bugs) 32-bit [non-]support #3
• Optional hard resource limits - recursion depth, number/size of objects allocated Resource Exhaustion DoS Vectors #2
• Parsing into an existing bytearray object, and parsing directly into a new bytes object without copy
• Implement the CID type entirely in C

[MarshalX]

pre-built wheels for all platforms and python versions (uploaded to pypi, automated via GHA)?

[DavidBuchanan314]

pre-built wheels for all platforms and python versions (uploaded to pypi, automated via GHA)?

I'd love to have that but it's not something I know how to do currently, is it maybe something you could help with setting up? I'll put it down as a 1.x goal but that doesn't mean it can't happen sooner (my main aim with 1.0 is to have it ready to use in my PDS rewrite)