required=False not working as expected for string fields

[kadeeraziz]

Description:

Expected behavior:
The request should be processed successfully, even if the string field is left empty, since required=False is specified for the field.

Actual behavior:
The request fails, and a validation error is returned, indicating that the string field is required, despite setting required=False.

Additional information:
I've tested this issue on django_jsonform version 2.17.1, using Python 3.8.10 and Django 3.2.18

Relevant code snippets:

schema = {
        "type": "dict",
        "keys": {
            "required_field": {
                   "type": "string", 
                   "required": False
            },
            "non_required_field": {
                  "type": "string"
             }
        },
    }

value = {"required_field": "Hello, world!"}

validator = JSONSchemaValidator(schema=schema)
validator(value)

This throws an exception, saying: These fields are missing from the data: non_required_field is required.

It's possible that the issue lies within the validation process or the way the JSON schema is handled.

Please let me know if there's any additional information required or if you have further questions.

[bhch]

Yes, this is kind of a "known" issue at this point.

As per the JSON Schema spec, all properties are optional by default. But I've chosen not to follow that rule and instead stick to how things work in Django.

E.g., even if a field is not required, such as — name = models.CharField(required=False) — Django will still insert an empty string in the database.

So, we interpret the required keyword as to whether this field (key) can be empty or not.

But we still require that the submitted data must contain all the keys which are declared in the schema.

So, the built-in JSONSchemaValidator will consider this valid: {"required_field": "", "non_required_field": ""}

[bhch]

I'm closing this issue because of lack of activity. And also because the existing behaviour will most likely not change.

If you have more arguments to make or questions to ask, feel free to reopen the issue.