Local references about python-fastjsonschema HOT 7 CLOSED

Does the library supports local references? Like if I have common.json and then schema1.json and schema2.json, where I would like to use something defined in common.json. All files in the same directory. I noticed that it tried to download the referenced schema but I would prefer to use the local file - just giving relative path as "$ref": "common.json#/definitions/xyz".

Have you find a way to solve this problem ?

from python-fastjsonschema.

@EMAN6 Have a look here: https://github.com/robofit/arcor2/blob/6b57dc4a7431fab77a72d5c8547157a8607cea94/helpers.py#L60

from python-fastjsonschema.

hey @horejsek is this something we could contribute to fastjsonschema? In case you're interested in supporting it.

from python-fastjsonschema.

@heitorlessa Hi, it is already supported. You can pass handlers parameter where you can define how to retrieve your local references.

from python-fastjsonschema.

Personnally, I have a hard time using the handlers parameter. It is never ran and I get a urllib.error.URLError: <urlopen error [Errno 2] No such file or directory: '/soil_sample.schema.json'>. It makes me think my handler did not have the chance to change the ref to the absolute path because it was never ran.

from python-fastjsonschema.

@WilliamHarvey97 could you provide smallest runnable code to showcase the problem? Ideally with opening new issue.

from python-fastjsonschema.

I've made an example to demonstrate the incorrect behavior I'm seeing with local references. It might be that I'm using handlers incorrectly or my schema is invalid, but I don't think so.

I tested this with (Python 3.8.10) :
jsonschema==4.21.1
fastjsonschema==2.19.1

I've inlined the schemas into the python file to make this easier to debug, but it wouldnt have worked correctly if it was in an external file either.

import fastjsonschema
from jsonschema import Draft7Validator, RefResolver
from referencing import Registry, Resource
from referencing.jsonschema import DRAFT7


def remotes_handler(uri):
    print(f"remotes_handler: {uri}")
    raise Exception("remotes_handler: {uri}")
    # req = urlopen(uri)
    # encoding = req.info().get_content_charset() or "utf-8"
    # return json.loads(
    #    req.read().decode(encoding),
    # )


handlers = {"": remotes_handler}

shared_schema = {
    "$schema": "http://json-schema.org/draft-07/schema#",
    "$id": "shared_schema.template.json",
    "type": "object",
    "definitions": {"metadata": {"type": "object"}},
}

schema = {
    "$schema": "http://json-schema.org/draft-07/schema#",
    "$id": "id_path/example.template.json",
    "type": "object",
    "additionalProperties": {
        "type": "object",
        "properties": {
            "metadata": {"$ref": "shared_schema.template.json#/definitions/metadata"},
        },
    },
}

json_data = {"tacos": {"metadata": {"version": "1.0.0"}}}

# resolver = RefResolver(base_uri="file:///", referrer=handlers)

schema_resource = Resource.from_contents(shared_schema)
registry = Registry().with_resource(
    uri="id_path/shared_schema.template.json",
    resource=schema_resource,
)

validator = Draft7Validator(schema, registry=registry)
errors = sorted(validator.iter_errors(json_data), key=lambda e: e.path)
print(errors)

print(f"is_valid: {validator.is_valid(json_data)}")
validator = fastjsonschema.compile(
    schema,
    handlers=handlers,
)

# this doesnt get reached, because resolver is incorrect.
validator(json_data)

What seem to be happening is that inside RefResolver's resolving, its incorrectly concatenating paths with the $id's path:
https://github.com/horejsek/python-fastjsonschema/blob/master/fastjsonschema/ref_resolver.py#L126

When doing so it also gets into a recursive look of resolving references until my handler gets called, at which point the url sent to it is:
'id_path/id_path/id_path/id_path/id_path/shared_schema.template.json'

I'm not sure exactly the right general solution, but one idea is if this behavior is generally wanted, then for my case I could pass in my own RefResolver, then I could correct for the behavior on my side without strange hacks).

Since I also wanted to make sure my schemas arent somehow invalid, I validated them using python jsonschema which is also in this example.

from python-fastjsonschema.