Fix for Pydantic 2.x

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !

[fantix]

Background:

• Pydantic v2 has a limitation (PydanticUndefinedAnnotation: forward references in imported dataclass for TypeAdapter pydantic/pydantic#7111) that TypeAdapter needs to look back in the stack for typing context (e.g. in globals). Therefore, FastAPI cannot currently use non-Pydantic types with forward type references properly.
• Pydantic v2 changed the internal API and deprecated __get_validators__() in favor of the new schema design.

Solution:

Skip constructing dataclass schema under Pydantic v2 through __get_pydantic_core_schema__(), which:

1. Avoids looking into the model types at all, unblocking FastAPI with Pydantic v2;
2. Skips Pydantic validation;
3. Is preferred over the __get_validators__(), so the latter can stay for v1 compatibility.

Fixes#466

[fantix]

When returning edgedb-python query results directly as a FastAPI response as demonstrated in the tutorial, the objects are inferred as dataclass by Pydantic, and serialized directly through the dict from dataclass. Both edgedb-python and Pydantic in this case are native Python extensions.

[jsimonlane]

Hey @fantix -- what's the status of getting this CR reviewed? Thanks!

[jsimonlane]

Bumping this one more time, any update on getting this CR approved? Guessing the above comment got lost in the Holiday scuffle. @1st1@fantix ?

Thanks!

[fantix]

Sorry yeah, I'm hoping to get a review and put out a new release in a day or two

[jsimonlane]

Thanks @fantix !