[Bug]: payloads preset inlines duplicate `$id` into `theCodeGenSchema`, breaking AJV `createValidator()`

[jonaslagoni]

Description

When the same shared JSON Schema (with its own $id) is referenced from two sibling subschemas inside a payload, the payloads preset inlines two literal copies of that schema — both keeping the same $id — into the parent payload's static theCodeGenSchema. AJV rejects the resulting schema at ajv.compile(...) time with reference "<id>" resolves to more than one schema, which makes the generated createValidator() throw. Any consumer that calls <Payload>.createValidator() (including the channels preset's subscription setup) crashes on first invocation.

Environment

• CLI Version: @the-codegen-project/cli@0.72.0 (current latest)
• Node Version: v22.14.0
• OS: Windows 11

Minimal Reproduction

input.asyncapi.json

{
  "asyncapi": "3.0.0",
  "info": { "title": "Duplicate $id repro", "version": "0.1.0" },
  "channels": {
    "StartRunCommand": {
      "address": "runs.command.start",
      "messages": { "startRun": { "$ref": "#/components/messages/startRun" } }
    }
  },
  "operations": {
    "startRun": {
      "action": "receive",
      "channel": { "$ref": "#/channels/StartRunCommand" },
      "messages": [{ "$ref": "#/channels/StartRunCommand/messages/startRun" }]
    }
  },
  "components": {
    "messages": {
      "startRun": {
        "name": "startRun",
        "payload": {
          "$schema": "http://json-schema.org/draft-07/schema#",
          "type": "object",
          "properties": {
            "generatorA": {
              "$schema": "http://json-schema.org/draft-07/schema#",
              "$id": "GeneratorA",
              "type": "object",
              "properties": {
                "documentContext": {
                  "$schema": "http://json-schema.org/draft-07/schema#",
                  "$id": "DocumentContext",
                  "type": "object",
                  "properties": { "specificationDocument": { "type": "string" } }
                }
              }
            },
            "generatorB": {
              "$schema": "http://json-schema.org/draft-07/schema#",
              "$id": "GeneratorB",
              "type": "object",
              "properties": {
                "documentContext": {
                  "$ref": "#/channels/StartRunCommand/messages/startRun/payload/properties/generatorA/properties/documentContext"
                }
              }
            }
          }
        }
      }
    }
  }
}

Note the asymmetry: generatorA.documentContext is the canonical inline subschema ($id: "DocumentContext"), and generatorB.documentContext is an internal JSON-pointer $ref to it. This is what an AsyncAPI bundler produces when two siblings both originally $ref-ed the same shared schema file.

codegen.ts

import { TheCodegenConfiguration } from '@the-codegen-project/cli'
const config: TheCodegenConfiguration = {
  inputType: 'asyncapi',
  inputPath: './input.asyncapi.json',
  language: 'typescript',
  generators: [
    {
      preset: 'payloads',
      outputPath: './out/payloads',
      serializationType: 'json',
      language: 'typescript',
    },
  ],
}
export default config

Command

codegen generate ./codegen.ts

Expected Output

Any one of the following would let AJV compile the generated theCodeGenSchema:

1. Keep one inline definition and use a JSON-pointer $ref for subsequent occurrences (matching the input shape).
2. Hoist the shared schema into a $defs entry on the root of theCodeGenSchema and $ref it from each site.
3. If full dereferencing is intentional, strip $id from the inlined duplicates so AJV only sees one anchor for that schema.

Example (option 1):

public static theCodeGenSchema = {
  // ...
  "properties": {
    "generatorA": { /* ... */ "documentContext": { "$id": "DocumentContext", /* ... */ } },
    "generatorB": { /* ... */ "documentContext": { "$ref": "#/properties/generatorA/properties/documentContext" } }
  },
  "$id": "startRun"
};

Actual Output

out/payloads/StartRun.ts (relevant excerpt, formatted):

public static theCodeGenSchema = {
  "type": "object",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "properties": {
    "generatorA": {
      "$schema": "http://json-schema.org/draft-07/schema#",
      "$id": "GeneratorA",
      "type": "object",
      "properties": {
        "documentContext": {
          "$schema": "http://json-schema.org/draft-07/schema#",
          "$id": "DocumentContext",         // copy 1
          "type": "object",
          "properties": { "specificationDocument": { "type": "string" } }
        }
      }
    },
    "generatorB": {
      "$schema": "http://json-schema.org/draft-07/schema#",
      "$id": "GeneratorB",
      "type": "object",
      "properties": {
        "documentContext": {
          "$schema": "http://json-schema.org/draft-07/schema#",
          "$id": "DocumentContext",         // copy 2 — same $id, different position
          "type": "object",
          "properties": { "specificationDocument": { "type": "string" } }
        }
      }
    }
  },
  "$id": "startRun"
};

public static createValidator(...): ValidateFunction {
  // ...
  const validate = ajvInstance.compile(this.theCodeGenSchema); // throws
  return validate;
}

Calling StartRun.createValidator() (or feeding the generated theCodeGenSchema directly into new Ajv().compile(...)) throws:

Error: reference "DocumentContext" resolves to more than one schema
    at ambiguos (.../ajv/lib/compile/resolve.ts:147:12)
    at Ajv.addRef (.../ajv/lib/compile/resolve.ts:115:38)
    ...
    at Ajv.compile (.../ajv/lib/core.ts:388:22)

Impact

Any service consuming payloads runtime validation on a payload whose schema tree contains a shared $id-bearing schema referenced from two siblings will crash at startup the first time <Payload>.createValidator() is called. The channels preset calls createValidator() eagerly inside subscription setup, so the failure surfaces immediately — not only when a message arrives.

The same shape also produces duplicate $id: "npm-release" etc. in real specs (with multiple shared schemas referenced from multiple generator subschemas); AJV simply reports the first conflict it hits.

Additional Context

• I searched the repo for DocumentContext, duplicate $id, resolves to more than one schema, ajv duplicate, theCodeGenSchema, inline schema duplicate, and payloads validate. No existing issue covers this.
• Tested with @the-codegen-project/cli@0.72.0 (latest). The behavior reproduces; nothing in v0.64.x→v0.72.0 release notes mentions related fixes.
• Verified the AJV failure independently by extracting the generated theCodeGenSchema literal and passing it to new Ajv().compile(...).