ModelSchema: allow model_fields = []

[hiaselhans]

trafficstars

This PR tests model_fields and model_exclude against None, so the following two ModelSchemes are valid:

class SampleSchema3(ModelSchema):
    class Config:
        model = User
        model_fields  = []


class SampleSchema4(ModelSchema):
    class Config:
        model = User
        model_exclude  = [] # equivalent to model_fields = "__all__"

[SmileyChris]

What's the reasoning behind this versus what already works?

Especially model_fields = [] doesn't naturally infer that you mean all fields. It reads as the opposite!

[hiaselhans]

its more idiomatic and it allows for a modelschema without any model_fields.

I think your deduction is wrong, please read the test included:

model_fields = [] => no fields model_exclude = [] => all fields

I think thats exactly what one would expect and its much easier to remember than __all__

[SmileyChris]

You're right, I just glanced at the code and description here and thought both options being an empty list were supposed to represent the same behaviour. My apologies for the overly quick skim read and reply -- thanks for following up with your reasoning.

I'm now personally just ambivalent on the change, probably because I'm used to seeing __all__ historically and I don't see that much benefit in introducing a second way to represent something we already can do. I can see your reasoning though. We'll see if anyone else has an opinion :)

[hiaselhans]

No worries, and sorry for the short answer :)

After all it might turn out to be a really just a matter of preference.

In my opinion, if i can exclude one field, I should also be able to exclude zero fields. If i can include one, i should also be able to include 0.

[vitalik]

sorry, closing this - as of django ninja v1 - we are following same guides as django model forms - so now recomended way is

class Foo(ModelSchema):
    class Meta:
          model = Foo
          fields = ['some'] # NOT !! model_fields