Build schema fields from plain annotated class.

[justanr]

• Generates fields from class level annotations
• Ignores ClassVar annotations
• Handles Optional/Union[T, None] by generating allow_none in the field options
• Allows registration/overriding of type to field definitions (Registration is provided at local and global level)
• Field constructors are encouraged to be more than just the field type
• Allows setting configuration for fields in a meta
• Will walk parents to find other configurations and combine into a single one
• Will respect fields that exist in the MRO already --- should there be a force option to override the field?
• Can auto register the schema itself when a target class is provided and registers as fields.Nested

import typing as t
from marshmallow_annotations import AnnotationSchema

class Frob:
    id: t.Optional[int]
    name: str
  
class Baz:
    frobs: t.List[Frob]

class FrobSchema(AnnotationSchema):
    class Meta:
        target = Frob
        register_as_scheme = True

class BazSchema(AnnotationSchema):
    class Meta:
        target = Baz
        register_as_schema = True

[coveralls]

Pull Request Test Coverage Report for Build 46

• 187 of 187 (100.0%) changed or added relevant lines in 6 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 100.0%

---

Totals
Change from base Build 6:	0.0%
Covered Lines:	187
Relevant Lines:	187

---

💛 - Coveralls

[justanr]

Few more things:

• List[Frob] -> Nested(FrobSchema, many=True) gonna punt on this for now, could be too hard for a timely merge
• Check forward declarations are handled properly forward declarations are fucky right now. :/
• Write docs

Gonna punt on the default value for now

edit: Of the three bullet points above, docs was the hardest to deal with. shows how much I know. lol

[justanr]

If converter grows an is_scheme method, that might solve the List(Nested(...)) issue.

See marshmallow-code/marshmallow#779 for a possible hiccup though.

[justanr]

Also, should this back port 3.6.5's get type hints fix?

edit: no, put a note in the docs recommending against forward declarations to target type in <3.6.5 without back porting the fix.

[justanr]

update readme, find a way to get mypy to ignore that file for proper depreciation (otherwise full nuclear will be needed), one more pass over docs is needed as well.