Support for MongoDB $bsonType extensions

[ee987]

When working with mongodb the database can enforce document structure with jsonSchema according to [1].

However mongodb has some own datatypes [2], like ObjectId, Date, Bool, Int that are not supported natively by json schema. These can be encoded by using the extension "bsonType" property instead of "type".

It would be very useful if jsonschema could validate such objects also. When a document fails validation inside mongodb you don't get any explanation why [3], so you always have to validate it with another validator client-side on failed insertions. Right now there is no such validator for python so the best workaround is to remove all bsonTypes before using jsonschema or convert the document into nodejs and use ajv. Neither option is really smooth.

[1] https://docs.mongodb.com/manual/reference/operator/query/jsonSchema/#jsonschema-extension
[2] https://docs.mongodb.com/manual/reference/operator/query/type/#document-type-available-types
[3] https://jira.mongodb.org/browse/SERVER-20547

[Julian]

This is as far as I know perfectly well supported, and there are users doing so.

What issue are you having and have you read the documentation on type extension

[Julian]

Closing, but happy to help if there's something specific here you're running into.

[ee987]

I think i managed to solve it now. Here is the core of it in case anyone else is interested.
It would have to be extended to support all required bsonTypes.
Small caveat is that bsontype and type shares underlying type-checker implementation so it will also accept bsontypes in the types property, but such schema would be rejected by mongo before even loaded so i think that's acceptable.

import jsonschema
from jsonschema import Draft7Validator
from bson import ObjectId

def is_objectId(checker, instance):
    return isinstance(instance, ObjectId)

type_checker = Draft7Validator.TYPE_CHECKER.redefine("objectId", is_objectId)
CustomValidator = jsonschema.validators.extend(Draft7Validator, type_checker=type_checker)
CustomValidator.VALIDATORS["bsonType"] = Draft7Validator.VALIDATORS["type"]

[juan-dura]

I am trying to implement this solution. But I obtain this error

AttributeError: 'dict' object has no attribute 'iter_errors'

@ee987 can you help me?

    import jsonschema
    from jsonschema import Draft7Validator
    from bson import ObjectId
    
    def is_objectId(checker, instance):
        return isinstance(instance, ObjectId)
    
    type_checker = Draft7Validator.TYPE_CHECKER.redefine("objectId", is_objectId)
    CustomValidator = jsonschema.validators.extend(Draft7Validator, type_checker=type_checker)
    CustomValidator.VALIDATORS["bsonType"] = Draft7Validator.VALIDATORS["type"]
    
    schema =  {
      	"bsonType": "object",
        "properties":{
          "scan":{
            "bsonType": "object",
            "required": ["code"],
            "properties":{
              "code": {
                "bsontype" : "string"
                },
              "device": {
                "bsonType": "int"
                }
              
              }
            }
        }
      }
      
    dic = { 
      "scan" : {
        "code": "hello",
        "device": 34
        } 
      } 
    
    
    CustomValidator.validate(dic,schema)

[qhaas]

@juan-dura

In jsonschema 4.20.0, the jsonschema.validators.extend function returns a class: RETURNS: a new jsonschema.protocols.Validator class extending the one provided

You need to instantiate your new CustomValidator class with your schema, e.g.

my_validator = CustomValidator(schema)
my_validator.validate(dic)

MongoDB 4-7 uses json-schema draft4, not Draft 7, so you might want to use Draft4Validator instead of Draft7Validator

There is also another issue in that json-schema draft 4 uses integer instead of int. So, you will need to switch your TYPE_CHECKER.redefine to a TYPE_CHECKER.redefine_many and add it, along with any other bson specific types you might need, e.g.

bson_redefines = {
        "objectId": lambda checker, instance:  isinstance(instance, ObjectId),
        "int": lambda checker, instance:  isinstance(instance, int),
}
bson_checkers = Draft4Validator.TYPE_CHECKER.redefine_many(bson_redefines)
CustomValidator = jsonschema.validators.extend(Draft4Validator, type_checker=bson_checkers)