Creation of Prepare Data Config

[camillebrianceau]

• Creation of the transforms folder and files
• Creation of the PrepareDataConfig and so the PrepareDataImageConfig, PrepareDataPatchConfig, PrepareDataSliceConfig and PrepareDataRoiConfig
• Creation of the prepare_data_param folder with the CLI adapted to the Config classes
• Creation of new Enum: DTIMeasure, DTISpace, ExtractionMethod, SliceDirection, SliceMode, Template, Pattern
• Adaptation of the code and the tests to these changes.

Future work: creation of unittests, better adaptation to Config in the data.py file,

[NicolasGensollen]

Something that could be useful is to define custom constructors for these classes:

class PrepareDataPatchConfig:
    caps_directory: Path
    preprocessing_cls: Preprocessing
    use_uncropped_image: ...

    @classmethod
    def from_preprocessing(cls, caps_directory: Path, preprocessing: dict):
        return cls(
            caps_directory,
            Preprocessing(preprocessing_dict["preprocessing"]),
            preprocessing_dict["use_uncropped_image"],
            ....
        )

This way, the code would be very simple:

self.config = PrepareDataPatchConfig.from_preprocessing(caps_directory, preprocessing_dict)

[camillebrianceau]

But these lines are just temporary...

[NicolasGensollen]

You might be able to use inheritance between these classes (if it makes sense) to factorize some code.

[thibaultdvx]

LGTM, except a mistake i told you about ConfigDict. Plus some minor remarks

[thibaultdvx]

Sorry @camillebrianceau, I told that you that we should use ConfigDict instead of Config (which is deprecated), but it turns out that they are not use in the same way. Some examples:

class ClassTest(BaseModel):
    a: str = "a"
    class Config:
        validate_assignment=True

test = ClassTest()
test.a = 0

will raise a ValidationError as expected, but the Config class usage is deprecated.

class ClassTest(BaseModel):
    a: str = "a"
    class ConfigDict:
        validate_assignment=True

test = ClassTest()
test.a = 0

will not raise any error unfortunately.

from pydantic import BaseModel, ConfigDict
class ClassTest(BaseModel):
    a: str = "a"
    model_config = ConfigDict(validate_assignment=True)

test = ClassTest()
test.a = 0

will raise a ValidationError as expected.

To sum up, we should use the last option.