Possibility to change default Pydantic to SQLAlchemy mappings

[sambadi]

Hello folks!
First of all i want to apologise for my bad english )
So i want to present two features realised for my project and what i want to share with community:
1)To be allow to change default Python to SqlAlchemy types mapping
2) To be allow to store Pydantic model data as Json at database (with ability to use SQLAlchemy Mutable wrap)

Possible is closes issue #503, but not sure

[github-actions]

📝 Docs preview for commit 6b5f287 at: https://6b70ec7d.sqlmodel.pages.dev

[sambadi]

Looks like in new version of pydantic (2.6.1) added some breaking changes. All completed tests at MR pipeline was started with previous stable version of pydantic - 2.5.3

This commit at Pydantic.main totally breaks the eq behaviour

pydantic/pydantic@62ed0aa#diff-8708af8d26c26644976866f974e0b810b0940ede43ff487e457057d5a1f07999

See discussion about _eq at #761

[github-actions]

📝 Docs preview for commit fb58b86 at: https://db9f51b8.sqlmodel.pages.dev

[KrilleGH]

That looks interesting. I'm using Pydantic's NaiveDatetime, and since it isn't a subclass of datetime.datetime I have to pass sa_type=saDateTime (from sqlalchemy import DateTime as saDateTime) on every field. Otherwise get_sqlalchemy_type() wouldn't know how to handle this type in the database.

However I think this case should be handled internally by SQLModel and without requiring extra code since SQLModel already depends on Pydantic and could just as well support these types.

[github-actions]

This pull request has a merge conflict that needs to be resolved.

[YuriiMotov]

@sambadi, thank you for your interest and efforts!

So, the idea of this PR is that instead of having a bunch of IFs in get_sqlalchemy_type we can create a sa_types_map dict where we can store all mapping information. Then in get_sqlalchemy_type we would iterate through sa_types_map and choose the correct type if present.

This will allow users to re-map type for their apps in one place instead of providing sa_type for every field of that type.

I personally like the idea in general.
Let's wait for Sebastian to validate the idea

[KrilleGH]

@sambadi, thanks for implementing this! I considered doing it myself one day, but now I don't have to. ❤️
Could you please rebase the branch and resolve the conflicts so we can move on easier?

[sambadi]

@sambadi, thanks for implementing this! I considered doing it myself one day, but now I don't have to. ❤️ Could you please rebase the branch and resolve the conflicts so we can move on easier?

I'll try, but not sure that can find enough time =(