Error when passing custom parameter in Nested schema

[qurram-zaheer]

Hello, I have 2 schemas set up something like this:

from marshmallow import Schema, fields, post_dump
from xyz.abc import CustomISODateTime


class SchemaA(Schema):
    ....
    field = fields.Dict(
        required=False,
        load_default={},
    ) # This dict will look like {..., 'string': 'abc'} 
    

    def __init__(self, *args, **kwargs):
        self.replace_field_with_string = (
            kwargs.pop('replace_field_with_string', False))
        super().__init__(*args, **kwargs)

    @post_dump
    def convert_field_to_string(self, data, **kwargs):
        if (self.replace_field_with_string and 'field' in data
            and isinstance(data['field'], dict)):
            data['field'] = data['field'].get('string', '')
        return data


class SchemaB(Schema):
    data = fields.List(
        fields.Nested(lambda: SchemaA(replace_field_with_string=True)),
        required=True,
        dump_only=True,
    )

Is there something incorrect in this code? Because when I do something like this:

data = some_function_that_returns_dict() # Made sure this dict was in the right format
schema = SchemaB()
x = schema.dump(data)

I get dictionary update sequence element #0 has length 1; 2 is required error

[deckar01]

I'm unable to reproduce that error.

from marshmallow import Schema, fields, post_dump


class SchemaA(Schema):
    field = fields.Dict(
        required=False,
        load_default={},
    )

    def __init__(self, *args, **kwargs):
        self.replace_field_with_string = (
            kwargs.pop('replace_field_with_string', False))
        super().__init__(*args, **kwargs)

    @post_dump
    def convert_field_to_string(self, data, **kwargs):
        if (self.replace_field_with_string and 'field' in data
            and isinstance(data['field'], dict)):
            data['field'] = data['field'].get('string', '')
        return data


class SchemaB(Schema):
    data = fields.List(
        fields.Nested(lambda: SchemaA(replace_field_with_string=True)),
        required=True,
        dump_only=True,
    )

data = {'data': [{'field': {'string': 'bar'}}]}
schema = SchemaB()
x = schema.dump(data)
print(x)
# {'data': [{'field': 'bar'}]}

[qurram-zaheer]

Hm that is interesting. I apologize if I am not explaining this right, but I tried tracing this issue in the debugger and the first thing I noticed was that in schema._serialize (line 521), when it went through the nesting and the attr_name was "field", the field_obj was <fields.Dict>. I thought then it would call Field.serialize and try to serialize a string as a dict, which caused the issue. I'm also seeing in my debugger that self.mapping_type is a dict.