Fix __fields_set__ not using alias field names (#517) #518
Conversation
Codecov Report
@@ Coverage Diff @@
## master #518 +/- ##
=====================================
Coverage 100% 100%
=====================================
Files 14 14
Lines 2272 2271 -1
Branches 448 447 -1
=====================================
- Hits 2272 2271 -1 |
pydantic/main.py
Outdated
| else: | ||
| fields_set = data.keys() & self.__values__.keys() # type: ignore | ||
| object.__setattr__(self, '__fields_set__', fields_set) | ||
| object.__setattr__(self, '__fields_set__', self._process_fields(data)) |
There was a problem hiding this comment.
humm, I'm not sure this is the easiest or most efficient way of doing things. It might be easier to build __fields_set__ in validate_model.
I think you can change like ~610 to
else:
names_used.add(field.name if using_name else field.alias)Then add names_used along with the other results.
There was a problem hiding this comment.
validate_model currently doesn't mutate the model object at all, so if someone is using it then it may break unexpectedly. There's also no way to return the names_used. If you think it's OK to call object.__setattr__(model, '__fields_set__', names_used) from validate_model then I can change it to do that.
There was a problem hiding this comment.
I would change the signature of validate_model to return a tuple including fields_set.
No one should be using validate_model directly so this should be fine.
There was a problem hiding this comment.
It was a mistake to have to different return types for validate_model, best to change it to always return Tuple['DictStrAny', Set[str], Optional[ValidationError]]
There was a problem hiding this comment.
If that's too much work for this PR, I could implement a PR to make the return type always Tuple['DictStrAny', Optional[ValidationError]], then you could rebase off that?
There was a problem hiding this comment.
Managed to do it. Might need a bit more testing but it should work.
There was a problem hiding this comment.
Also if no one should be using validate_model model it should probably be _validate_model or at least not exported in __init__.
|
awesome, thank you very much. |
* upstream/master: (138 commits) add 'none-any.whl' to pypi upload (pydantic#564) uprev update benchmarks (pydantic#563) cython (pydantic#548) Fix issue with unspecified generic type (pydantic#554) Run dataclass' original __post_init__ before validation (pydantic#560) try to stop annoying warnings in azure pipeline (pydantic#549) azure pipeline failOnStderr: false Azure Pipelines - tests for windows (pydantic#538) Fix JSON Schema for list, tuple, and set, improving interoperability (pydantic#540) uprev. Colors (pydantic#516) Fix to schema generation for IPv{4,6}{Address,Interface,Network} (pydantic#532) Fix __fields_set__ not using alias field names (pydantic#517) (pydantic#518) Change return type hint for create_model (pydantic#526) Tuple ellipsis (pydantic#512) Fix to schema generation for IPvAny{Address,Interface,Network} (pydantic#498) (pydantic#510) uprev Scheduled monthly dependency update for May (pydantic#499) Implement const keyword in Schema. (pydantic#469) ...
fix #517
Change Summary
Basically, instead of just setting
__fields_set__frominput_data.keys()(which could contain aliases instead of the field names), it checks all the fields and adds the field's name if it or the alias is ininput_data. Not sure if this is the best way to do this.Related issue number
#517
Checklist
HISTORY.rsthas been updated#<number>@<whomever>