Requiring one of two properties to exist but not both

Describe the bug Attempting to define a constraint where only one of two properties should exist but not both.

To Reproduce

Example schema:

{
    "$schema": "http://json-schema.org/draft-04/schema#",
    "type": "object",
    "properties": {
        "ReqHeader": {
            "type": "object",
            "properties": {
                "AThing": {
                    "type": "object",
                    "properties": {
                        "A_prop1": {
                            "type": "string"
                        },
                        "A_prop2": {
                            "type": "string"
                        }
                    }
                },
                "BThing": {
                    "type": "object",
                    "properties": {
                        "B_prop1": {
                            "type": "string"
                        },
                        "B_prop2": {
                            "type": "string"
                        }
                    }
                }
            },
            "oneOf": [
                {
                  "required": [
                    "AThing"
                  ]
                },
                {
                  "required": [
                    "BThing"
                  ]
                }
              ]
        }
    },
    "required": [
        "ReqHeader"
    ]
}

Used commandline:

datamodel-codegen --input test.json --input-file-type jsonschema --output test.py --target-python-version 3.6

Resultant output:

# generated by datamodel-codegen:
#   filename:  test.json
#   timestamp: 2021-08-11T15:37:19+00:00

from typing import Any, Union

from pydantic import BaseModel


class Model(BaseModel):
    ReqHeader: Union[Any, Any]

Expected behavior A clear and concise description of what you expected to happen.

• Have all of the objects created in the output module
• A constraint to enforce existence of AThing or BThing but not both.

Example input scenario expectations: Scenario: AThing Only (OK)

{
    "ReqHeader": {
        "AThing": {"A_prop1": "val", "A_prop2": "val"}
    }
} 

Scenario: BThing Only (OK)

{
    "ReqHeader": {
        "BThing": {"B_prop1": "val", "B_prop2": "val"}
    }
} 

Scenario: BThing and AThing (Not OK)

{
    "ReqHeader": {
        "AThing": {"A_prop1": "val", "A_prop2": "val"},
        "BThing": {"B_prop1": "val", "B_prop2": "val"}
    }
}

Version:

• OS: CentOS Linux release 7.6.1810 (Core)
• Python version: 3.6.8
• datamodel-code-generator version: 0.11.9

[tseader]

Works great - many thanks to you @koxudaxi for your help. StackOverflow unfortunately provided a different path that I tried, which functionally meets json-schema standards and pans out as far as the OK/NOT-OK scenarios above, but didn’t reflect well in Pydantic.

To summarize - the idea is to create encapsulation objects [ReqHeaderA and ReqHeaderB] for each choice so that the property name in the encapsulation object(s) [AThing and BThing in my scenario] can be appropriately exposed/used.

[hypogealgaol]

I have a similar use case, except that I want it to be one of two properties, either, but not none. I am getting the same result with the Union[Any, Any] but unsure if it is because of how the JSON schema is defined.

for example, I want to be able to have two string properties A and B be validated, where the only validation failure is when neither are present.

A json schema with a field definition after properties where A and B are defined, such as

"anyOf": [
               {
                    "required": [
                        "A"
                    ]
                },
                {
                    "required": [
                        "B"
                    ]
                }
            ],

generates a model where both A and B are just Optional[str] and not one where the non-presence of both fails validation, despite that occurring when the json validation runs.