And the dataclass serves as documentation for the available attributes and type safety..

It's really the way to go, takes a minute or two to define and saves so much headaches in the long run.

Can dataclasses be serialized just as easily?

Yes via pickle.

Or for JSON, dataclasses_json

Be aware that pickle is unsafe (as documented near the top of https://docs.python.org/3/library/pickle.html), so prefer other serialization formats except between trusted processes

I can't remember the last time I _wanted_ to pickle something. Parquet and JSON pretty much cover it for me. I guess pickling made more sense before tools like Pydantic were good and popular?

If I want all the python-specific nuance on disk, I know I'm in a bad place!

I use pickle to serialize RPC requests between internal services. Nothing beats it in terms of speed and ease of use.

Yep! Both msgspec (https://jcristharif.com/msgspec/supported-types.html#datacla...) and orjson support encoding dataclasses to JSON natively.

In addition to the already-mentioned suggestions, you can also use marshmallow schemas to serialise complex objects, and there's even a package to autogen marshmallow schemas from dataclasses: https://pypi.org/project/marshmallow-dataclass/

Yes, there are many tools for converting dataclasses to dict, json or something else, mashumaro (https://github.com/Fatal1ty/mashumaro) is one of them.

100%. I don’t understand the obsession with bending dicts into something they are inherently bad at when actual alternatives in dataclasses and Pydantic models right there.