cattrs VS squall

I'd suggest you look at my cattrs (https://catt.rs) library as a good serde lookalike in Python (sum type support present and getting better), and to use attrs instead of dataclasses in general.

Pydantic is by far not the only library of its kind, with prominent members of the same class being attrs, cattrs or even plain dataclasses for some use cases.

That being said, data serialization is a very common thing to do, so naturally there are tons of libraries that automate it for you. Personally, using dataclasses and cattrs is my goto way for doing such things.

I'm fond of attrs and cattrs myself, attrs make creating data classes a snap, writing all of the stupid code python requires to have a dataclass. Note the new built in dataclass is actually a limited copy of attrs. https://www.attrs.org/en/stable/ and https://github.com/python-attrs/cattrs

This month, I've released version 0.17, and it's now blazing fast; there is in fact no more comparison with Pydantic, which more than 5x slower (up to 30x in serialization). It's also faster than alternatives like mashumaro or cattrs. (See the quick benchmark result in documentation, and the code)

```

Cattrs has some problems with generics [1] [2]. Dacite and marshmallow-dataclasses don't support generics well either, with some issues around Union types.

They do work well for simple python types but what I'd like to see is guarantee that the serialisation operation is completely reversible and if not raise warning/exception.

[1] https://github.com/Tinche/cattrs/issues/149

By the way, while I was looking at the GitHub projects that are using apischema, I've discover https://github.com/mtag-dev/squall, a fork of FastAPI but using apischema among other differences.