Add an option in Config to choose which fields to use in __repr_args__

[Lotram]

I may have missed something, but it seems there is no way to exclude a field from an object's representation

Feature Request

I think it would be nice to have an option in Config to choose which fields to use in __repr_args__.

Another solution would be to include an repr boolean in the Field class, exactly like in dataclasses. I don't find it as convenient as the other solution, but maybe it makes more sense, since it's a per-field option.

Why

I have multiple big models ( more than 10 fields) and I'd like to reduce the size of their representation (either when using str() or repr()
So I could override the BaseModel to do something like this:

from typing import List, Optional

from pydantic import BaseModel


class Model(BaseModel):
   def __repr_args__(self):
       if self.Config._fields_in_repr is None:
           return super().__repr_args__()

       return (
           (key, value)
           for key, value in super().__repr_args__()
           if key in self.Config._fields_in_repr
       )

   class Config:
       _fields_in_repr: Optional[List[str]] = None


class User(Model):
   id: int
   login: str
   bio: str
   # lot of other fields

   class Config:
       _fields_in_repr = ["id", "login"]

user = User(id=1, login="Lotram", bio="This could be a rather long description, I don't want it in the the repr")
repr(user) #  "User(id=1, login='Lotram')"

However, IMHO we almost always want to exclude some fields from the object's repr, and I feel like this is generic enough to justify a new option in the BaseConfig class.

[eruvanos]

Hi,
I do have a similar request.

In my case I use the representation to render a python file.
It would be perfect, if I would be able to use the same filter/configuration like in the .dict() converter.
I agree that this kind of configuration would fit into the model config.

An alternative could be to have a default reprlib.Repr implementation shipped (Python module reprlib).

        include: Union['AbstractSetIntStr', 'MappingIntStrAny'] = None,
        exclude: Union['AbstractSetIntStr', 'MappingIntStrAny'] = None,
        by_alias: bool = False,
        skip_defaults: bool = None,
        exclude_unset: bool = False,
        exclude_defaults: bool = False,
        exclude_none: bool = False,

[jab]

class User(Model):
    ...
    class Config:
        _fields_in_repr = ["id", "login"]

Rather than having to list all the desired fields below their definition, in a custom Config, what if pydantic.Field allowed passing repr=False (similar to dataclasses.field() and attr.ib())? That way you could opt out of repr participation for a field right at the field definition site:

class User(Model):
    password_hash: bytes = pydantic.Field(..., repr=False)
    ...

[Lotram]

Yep that's actually the other solution I suggested in the issue :) . I think I said I find it less convenient because you have to call the Field class instead of using only annotations, but that's not a real problem, so that's a good solution too.

[fnep]

For the repr=False I currently use this helper:

class ReducedRepresentation:
    def __repr_args__(self: BaseModel) -> "ReprArgs":
        return [
            (key, value)
            for key, value in self.__dict__.items()
            if self.__fields__[key].field_info.extra.get("repr", True)
        ]

… like this …

class Instance(ReducedRepresentation, BaseModel):
    node: str
    url: str = Field(repr=False)

But would also prefer to have this in the library.

[jeanmonet]

I think it would be nice to have an option in Config to choose which fields to use in __repr_args__.

This would be very useful. Additional parameterization of repr inside the Config class would be desirable - would allow to include additional fields, such as properties that depend on multiple attributes.