v2.0.0 Release Plan

[Apricot-S]

• [Feature Proposal] Change instance methods to static methods #74
• Change the use of class variables to instance variables to resolve issues in free-threaded builds #84
• [Proposal] Improve type safety with potential Pydantic adoption #90
• Add Shanten.TENPAI_STATE = 0
• Remove deprecated attributes
• Drop support for Python 3.9, which will reach EOL in October 2025
• Refactor unit tests
  • Split multiple asserts into separate parameterized tests
  • Add missing test cases
• Refactor code to resolve existing mypy type errors
• Split the example into use cases
• Wiki for v2
  • Replace the link to the repository in the v1 Wiki with a link tagged with v1.4.0
• API documentation (Sphinx)

[Nihisil]

I'm thinking about whether we should add the pydantic dependency to the project to have typed inputs and outputs. It will hurt performance a bit, but it will make using the library much easier. I have not made a decision on this yet, as it involves a lot of work, but maybe I will find time for that improvement. What are your thoughts on this?

[Apricot-S]

Personally, I consider the fact that this library has zero external dependencies to be one of its characteristics.
For introducing types, I was thinking of using TypedDict.

On the other hand, if the library is used via an API, I agree that introducing type safety with pydantic could be very beneficial.

@Nihisil
Do you envision using pydantic models also for inputs and outputs,
or would its usage be limited to internal implementation only?

[Apricot-S]

Reference: Serialization & Validation Library Benchmarks
https://jcristharif.com/msgspec/benchmarks.html

[Nihisil]

Do you envision using pydantic models also for inputs and outputs

Yes, that was my plan, but I am still not sure since, as you mentioned, library without external dependencies can be a good characteristic for a lib