Pydantic v2 Support #49
Conversation
- replace use of __fields__ with model_fields - change variabled model_orm to just orm to prevent class with pydantic protected name space - Change looking at type_ to annotation - remove const =True from field arguments
- use a context manager to change working directory to work with pydantic pathlib type better - move SQL tests away from config class - migraste orm_mode to from_attributes
- Added more =None to prevent validation errors on optional fields - Changed raises and warn test to validate/test malformed integers instead of strings - This is due to a change in pydanticv2 no longer being able to convert strings to integers - Changed some pytest to use tmp_path fixture instead of tmpdir due to pydanticv2 compatibility with pathlib
- more setting None as default values in optional[str] fields
- is_required will now only correctly set on Optional and Union fields if a default argument of None is specified - To fix this, added an if statement in text to look for the tests with Optional or Union and change the class with a default arg for those tests - Functionality has changed with having a nested literal inside an Optional type - updated the to sql type function to check if the type is still Literal when checking Optional or Union, if it is then perform the Literal parsing on the nested type
- remove final usage of .dict() for .model_dump()
|
Hi @Miksus, Just to let you know I've now marked this PR ready for review. Breaking changesThis PR contains breaking changes for anybody that consumes this, so you may have to version this appropriately.
The rest are just migration pieces. |
|
@Jypear, thanks for your work on this, I've switched to your fork for my projects. To avoid the |
Creating this as a work in progress/draft PR to get the ball rolling on the pydanticv2 migrations.
There's potentially a breaking change (changing the model_orm argument for SQLRepo to just orm). Using "model_" now clashes with a pydantic protected Namespace.
We might be able to get round this using a Field alias, but it would require some testing to see if it would still cause the clash.
Tests are still failing at this point but Memory Repo and SQLRepo are running in a basic test.
Will continue to work on this over the weekend hopefully.