Preserve annotations from bases

[dmontagu]

Change Summary

Preserve annotations from superclasses when modified in subclasses, even if the annotation is not repeated. (This prevents the field from being inferred as a different type.)

Related issue number

Fixes #757

Checklist

@samuelcolvin I don't think this change affects any existing documentation, and it feels subtle enough to me that it probably doesn't merit new docs (I think people should have assumed things would work this way, like @kszucs did). But let me know if you'd prefer I add a note.

• Unit tests for the changes exist
• Tests pass on CI and coverage remains at 100%
• Documentation reflects the changes where applicable
• changes/<pull request or issue id>-<github username>.rst file added describing change
  (see changes/README.md for details)

[codecov]

Codecov Report

Merging #758 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master   #758   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files          15     15           
  Lines        2726   2726           
  Branches      537    536    -1     
=====================================
  Hits         2726   2726

[kszucs]

@dmontagu for my PR the following snippet has failed with validation error for the list field:

from pydantic import BaseModel
from typing import List


class Base(BaseModel):
        i: int = 1
        s: str = 'string'
        f: float = 1.1
        l: List[str] = []


class Child(Base):
    s = 'str'
    l = ['a']

Child(i=4, l=['a', 'b'])

I think this solution is more robust so I'm closing mine in favour of this PR.

[samuelcolvin]

what happens in the case that I change this to value: str = 'G'?

[kszucs]

It should update the type annotation to str, I guess a test case is required.

[samuelcolvin]

otherwise LGTM.

[kszucs]

@dmontagu @samuelcolvin many thanks for the fix!

[kszucs]

@dmontagu @samuelcolvin it doesn't work for nested types, example to reproduce:

    class A(BaseModel):
        seq: List[int] = [0, 1, 2]

    class B(A):
        seq = [4, 5]

    class C(A):
        seq = []

Inferred B.seq type=list default=[4, 5]
A.seq type=int default=[0, 1, 2]

[dmontagu]

@kszucs It looks like maybe the example you shared wasn't run against the current master branch (note: master is targeting 1.0, rather than 0.32.x, I think). On master, your example actually breaks:

from typing import List

from pydantic import BaseModel


class A(BaseModel):
    seq: List[int] = [0, 1, 2]


class B(A):
    seq = [4, 5]

# TypeError: The type of B.seq differs from the new default value; if you wish to change the type of this field, please use a type annotation

This is because we currently don't infer value types. If you reannotate the field as List[int] in the subclass, it just works.

If you find this annoying and would prefer it behave differently I think that could make sense -- in that case, could you create a new issue for discussion?

(At the very least, the error message should probably be a little clearer about how the inferred type differed; I think this specific example would be a confusing error message to deal with.)

[kszucs]

@dmontagu yep, I would expect it not to break. Sure I can create a new issue.