Fix #4192 bug with BaseModel.construct and aliased Fields

[kylebamos]

Change Summary

Update BaseModel.construct to consider field aliases when constructing.

Related issue number

fix #4192

Checklist

• 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>.md file added describing change
  (see changes/README.md for details)
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

[kylebamos]

If construct were to behave consistently with respect to allow_population_by_field_name_config, this would raise a ValueError, which is how I had the behavior in my initial commit, but that would be a breaking change, so I decided against it.

[kylebamos]

please review

[hramezani]

Thanks @kylebamos for your patch 👍

Here is my understanding.
By having Foo1 as a model with allow_population_by_field_name=True and Foo2 as a model with allow_population_by_field_name=False:

from pydantic import BaseConfig, BaseModel, Field


class Foo1(BaseModel):
    class Config(BaseConfig):
        allow_population_by_field_name = True

    bar_: int = Field(..., alias="bar")


class Foo2(BaseModel):
    class Config(BaseConfig):
        allow_population_by_field_name = False

    bar_: int = Field(..., alias="bar")


expected_value = 2

Here is the current state:

• Passing alias in standard instantiation:

# 1 - OK
assert Foo1(bar=expected_value).bar_ == expected_value

# 2 - OK
assert Foo2(bar=expected_value).bar_ == expected_value

• Passing alias in construct:

# 3 - AttributeError: 'Foo' object has no attribute 'bar_' -- Has to be fixed
assert Foo1.construct(bar=expected_value).bar_ == expected_value

# 4 - AttributeError: 'Foo' object has no attribute 'bar_' -- Ok
assert Foo2.construct(bar=expected_value).bar_ == expected_value

• Passing field name in standard instantiation:

# 5 - OK
assert Foo1(bar_=expected_value).bar_ == expected_value

# 6 - ValidationError: 1 validation error for Foo2 bar -- OK
assert Foo2(bar_=expected_value).bar_ == expected_value

• Passing field name in construct:

# 7 - OK
assert Foo1.construct(bar_=expected_value).bar_ == expected_value
# 8 - OK
assert Foo2.construct(bar_=expected_value).bar_ == expected_value

And your patch will fix the Case 3 which is OK.

Regarding Case 8, I think it should raise an error like Case 6.
I think it's not related to your patch and it may break our existing implementation.
What do you think @samuelcolvin and @PrettyWood ?

[hramezani]

Please update

[kylebamos]

Regarding case 8 you highlighted above, I addressed that in my comment as well because I initially had it fixed to accommodate that scenario, but it does break other tests.

https://github.com/samuelcolvin/pydantic/pull/4191/files#r909024428

It's as simple as changing line 592 in main.py to be elif cls.__config__.allow_population_by_field_name and name in values:

and adding else: errors.append(ErrorWrapper(MissingError(), loc=field.alias)) at the end of the if...elif chain in that same spot

[kylebamos]

please review

[samuelcolvin]

@kylebamos I have to say I find your avatar very confusing, I keep thinking there are lots of negative comments and reviews.

I'm marginally pro this change, my concern is whether it could be a breaking change to anyone. @PrettyWood do you have any thoughts?

Also conflicts needs solving, please update.

[kylebamos]

Conflict resolved. Avatar fixed to your liking, I hope.

[kylebamos]

Please review

[samuelcolvin]

Avatar fixed to your liking, I hope.

Much better 👍 I'm sure I'm not the only one who will prefer it 😏.

[PrettyWood]

LGTM! Good for v1.10!

[samuelcolvin]

@kylebamos thanks so much for this.