Added support to filter nested fields

[jtrain]

#42 replacement for this PR

@albertisfu here is an alternative approach where we do allow each serializer the ability to filter itself instead of orchestrating the filtering from the parent.

Does this approach work for your use-case? You can see that this cuts down on the logic spread throughout the fields method, and contains the hard work of finding the fields to filter / omit in some helper functions - I'm sure these could be simplified further - they are just gpt code. (for example I don't like while statements as a rule)

[albertisfu]

Thank you, @jtrain. I’ve tested this approach, and it works for our use case.

I only needed to apply a simple fix in the compute_level method, because the level computation was failing when serializing multiple instances, for example:

serializer = self.SchoolSerializer([school, school_1], many=True, context={"request": request})

Tweaking the logic to avoid incrementing the level for a ListSerializer resolves the issue.

def compute_level(serializer):
    level = 0
    current = serializer

    while getattr(current, "parent", None) is not None:
        parent = current.parent
        current = parent
        if not isinstance(parent, serializers.ListSerializer):
            level += 1
    return level

We should also update and extend the tests to cover this scenario.

I think it’s fine that you removed all the logic for determining which fields to defer—we can derive that from the request and handle it in our codebase, which makes this PR simpler.

I agree with this approach. If you’d like, I can refine the logic further and improve the tests (I might need to open a separate PR based on this one due to permissions).

Let me know what you think.

[jtrain]

@albertisfu great! yes please if you like you can use this PR as your base to make the improvements you've suggested.