Inconsistent behavior when setting default

[AAriam]

Using the example in the documentation, extend Draft202012Validator to set default values:

import jsonschema

def extend_with_default(validator_class):

    validate_properties = validator_class.VALIDATORS["properties"]

    def set_defaults(validator, properties, instance, schema):
        for property, subschema in properties.items():
            if "default" in subschema:
                instance.setdefault(property, subschema["default"])

        for error in validate_properties(
            validator,
            properties,
            instance,
            schema,
        ):
            yield error

    return jsonschema.validators.extend(validator_class, {"properties": set_defaults})

validator = extend_with_default(jsonschema.Draft202012Validator)

Our data can be either a string, or an object with property x, which itself can be either a string, or an object with property x1. If x is not provided, the default should be an object, and the default of x1 should be True. The function below creates a schema for this scenario; it also takes an optional schema_url, which if provided, is added as $schema to the subschema where the data is an object:

def make_schema(schema_url: str | None = None) -> dict:
    schema = {
        "$schema": "https://json-schema.org/draft/2020-12/schema",
        "anyOf": [
            {"type": "string"},
            {
                "$schema": schema_url,
                "type": "object",
                "properties":{
                    "x": {
                        "default": {},
                        "anyOf": [
                            {
                                "type": "string"
                            },
                            {
                                "type": "object",
                                "properties": {
                                    "x1": {
                                        "default": True
                                    }
                                }
                            }
                        ],
                    }
                },
                "required": ["x"]  
            },
        ]
    }
    if not schema_url:
        schema["anyOf"][1].pop("$schema")
    return schema

Now if we create a schema without schema_url, everything works as expected. For example, if we pass an empty dictionary as our data, it successfully validates, x is set to an empty dict, and the default value of x1 is added to it:

data = {}
schema = make_schema()
validator(schema).validate(data)
print(data)

returns

{'x': {'x1': True}}

However, if schema_url is set to https://json-schema.org/draft/2020-12/schema, then we get a ValidationError:

data = {}
schema = make_schema("https://json-schema.org/draft/2020-12/schema")
validator(schema).validate(data)  # raises ValidationError

Interestingly, if schema_url is set to anything else (either other schema specification URLs or even invalid URLs), again everything works as expected. For example all these work:

for schema_url in ("https://json-schema.org/draft-07/schema", "NOT A URL"):
    data = {}
    schema = make_schema(schema_url)
    validator(schema).validate(data)
    print(data)

[Julian]

This sounds like the (not yet solved) #994 (which will likely be solved only once we have a new validation API) -- have a look there, and if it seems different let me know.

(PRs to improve the example code are also welcome.)