xoryaml 0.2.0

yaml serializer and deserializer backed by yaml_rust2

xoryaml

Why?

Yet Another YAML library for Python ;)

I wanted a memory-safe and reasonably fast YAML library that could handle datetime types. It takes inspiration (as well as a bit of the test suite) from orjson.

Demo

@dataclasses.dataclass
class YAYAMLModule:
    name: str
    released: datetime.datetime
    tuples: tuple[str, str]
    sets: set[str]
    dicts: dict[int, str]

module = YAYAMLModule(
    "xoryaml",
    released=datetime.datetime.now(datetime.timezone.utc),
    tuples=("supported", "of course"),
    sets={"why", "not?"},
    dicts={1: "No,", 2: "fíjate."},
)
print(xoryaml.dumps(module))
# ---
# name: xoryaml
# released: "2025-03-08T17:01:14.569027Z"
# tuples:
#   - supported
#   - of course
# sets:
#   - not?
#   - why
# dicts:
#   1: "No,"
#   2: fíjate.
print(xoryaml.loads(xoryaml.dumps(module)))
# {
#     "name": "xoryaml",
#     "released": "2025-03-08T17:01:14.569027Z",
#     "tuples": ["supported", "of course"],
#     "sets": ["why", "not?"],
#     "dicts": {1: "No,", 2: "fíjate."},
# }

In addition to datetimes, dataclasses, sets, and tuples are automatically handled. You don't need to pass any settings for these to work. On the other hand, you don't get any other settings either ;)

If you want to restore the parsed data into python data structures, I highly recommend pydantic.

validator = pydantic.TypeAdapter(YAYAMLModule).validator.validate_python
print(validator(xoryaml.loads(xoryaml.dumps(module))))
# YAYAMLModule(
#     name="xoryaml",
#     released=datetime.datetime(2025, 3, 08, 17, 1, 14, 569027, tzinfo=TzInfo(UTC)),
#     tuples=("supported", "of course"),
#     sets={"not?", "why"},
#     dicts={1: "No,", 2: "fíjate."},
# )

Thanks

This project would not be possible without:

• pyo3: Rust-python interop
• yaml-rust2: Pure Rust YAML impl
• orjson: Parts of test suite
• ryaml: Parts of test suite
• speedate (pydantic): Date formatting

Dev install

• Install uv
• Install maturin: uv tool install maturin
• Make a venv uv venv
• Activate the venv source .venv/bin/activate
• Build and install uv pip install -e '.[bench]'
• Optional: build release maturin develop --release
• Run tests pytest

In order to keep the interface to native code as simple as possible, I have implemented part of the interface in Python. Not sure if this is a bad idea, but it seems to work fine? Let me know if otherwise.

Run tests on all supported versions of python locally

• Install tox: uv tool install tox --with tox-uv
• Run: tox