load many functionality doesn't seem to be of much help

I have a UserSchema with a pre_load(pass_many=False) decorator, suppose if the pre_load decorator raises ValidationError, all the objects in the collection I passed became invalid even though there are some valid objects available. IMO errors of one object shouldn't stop the other objects in the collection from being deserialized, validated, and post-processed.

The hack would be to run a loop on the collection and load the object one by one but it would be nice if the 'many' attribute can handle this. There are some trade-offs in loading the object one by one.

from marshmallow import Schema, ValidationError, fields, pre_load

class UserSchema(Schema):
    name = fields.Str()
    age = fields.Int()
    height = fields.Float()

    @pre_load(pass_many=False)
    def check_age(self, data, **kwargs):
        if data['age'] < 18:
            raise ValidationError('Minor user', 'age')
        
        return data

user_schema = UserSchema()

data = [
    {
        'name': 'dk',
        'age': 20,
        'height': '5.5'
    },
    {
        'name': 'dd',
        'age': 19,
        'height': '5.5'
    },
    {
        'name': 'df',
        'age': 10,
        'height': '5.5'
    }
]

try:
    user_schema.load(data, many=True)
except ValidationError as err:
    print(err.messages, err.valid_data)


for d in data:
    try:
        d = user_schema.load(d)
        print(d)
    except ValidationError as err:
        print(err.messages, err.valid_data)

Suggestions are welcome.

Start by reproducing the example with UserSchema.load(data, many=True) and compare it with loading each item separately. Trace the many=True deserialization and pre_load validation flow; done means invalid items report their errors while valid items are deserialized and post-processed, with the intended valid_data behavior covered by tests.