lint --fix creates an unconstrained subschema in allOf

[jeremyfiel]

disregarding the invalid usage of allOf and additionalProperties: false,

if the root has type: object defined with allOf, the --fix command adds an unnecessary unconstrained subschema to the allOf. Can this be removed? Is there a reason we need to also add this? I don't believe the schema produces a different output if it were to be removed.

{
  "$schema": "http://json-schema.org/draft-04/schema#",
  "type": "object",
  "additionalProperties": false,
  "allOf": [
    { "$ref": "#/definitions/thing" },
    {
      "type": "object",
      "properties": { "name": { "type": "string" } }
    }
  ],
  "definitions": {
    "thing": {
      "type": "object",
      "properties": {
        "age": { "type": "integer" }
      }
    }
  }
}

---

--fix

{
  "$schema": "http://json-schema.org/draft-04/schema#",
  "type": "object",
  "additionalProperties": false,
  "properties": {
    "name": {
      "type": "string"
    }
  },
  "allOf": [
    {
      "$ref": "#/definitions/thing"
    },
    {
      "type": "object"
    }
  ],
  "definitions": {
    "thing": {
      "type": "object",
      "properties": {
        "age": {
          "type": "integer"
        }
      }
    }
  }
}

---

---

With the root definition removed, it produces the expected results.

{
  "$schema": "http://json-schema.org/draft-04/schema#",
  "type": "object",
  "properties": {
    "name": {
      "type": "string"
    }
  },
  "allOf": [
    {
      "$ref": "#/definitions/thing"
    }
  ],
  "definitions": {
    "thing": {
      "type": "object",
      "properties": {
        "age": {
          "type": "integer"
        }
      }
    }
  }
}

[jviotti]

Ah, good catch. So the problem here is elevating properties outside the allOf given additionalProperties. As the semantics do change. I'll send a fix in a bit

[jviotti]

We probably have a similar issues with prefixItems + items or items + additionalItems, and also patternProperties, and probably if + then + else too. Also maxContains + minContains. Maybe $id and $anchor and $dynamicAnchor too...

[jeremyfiel]

Some of these may just be bad authoring habits haha

[jeremyfiel]

Some examples of the others would be helpful to see.

[jviotti]

The whole thing is a bit complicated. What I'll do it simplify the rules to be mainly about $ref (as that's the main case we care about here), and then we can have more rules for other common cases, but in a more "safe" whitelist manner. Sending a PR in a bit

[jviotti]

I have the fix ready. I will come out on the next CLI version. Most probably on Monday! Thanks for reporting it! 🙏🏻

[jviotti]

Please see https://github.com/sourcemeta/jsonschema/releases/tag/v12.3.0. I'm closing this issue, but please re-open it if you still see similar issues! Thanks!