dacite VS dataclassy

This is a good approach, but FYI there are already very good libraries that can be used to create objects from JSON/dicts in a similar way to your example, like dacite or pydantic. A lot of boilerplate can be avoided with these.

You'll need to be able to serialize ProjectState objects into some file format that you can save to disk, like json. When working with dataclasses, dacite can also be very helpful to convert to/from dictionaries. (There's other options available, of course.)

You also can't read any nested dataclassed from a dictionary. dacite! will help with this, but witihout it you can't just initialize it with NestedDataclass(**some_dict) or somethings like this. In the way it works for Plain dataclass.

There are also tools such as dacite and desert, which take a different approach to dealing with serialized data

dataclassy doesn't seem to have field and the docs seem to say to just not write it. But I get the same timestamp for all tests. I guessed/tried name: str = (lambda: 'Study ' + util.timestamp())() but same result. How do I port this?