Invalid JSON for schema-qualified type names requiring identifier quoting

[AurelienVernay]

Database version : Postgres 17.6
Plugin version : 2.6-2+b1
Using postgraphile V4 with "@graphile/subscriptions-lds" plugin at version "4.14.1".

Suppose we are working on a schema called "int". We define the following type & table :

CREATE TYPE int.int_pi_document_type AS ENUM(
    'DI',
    'PI',
    'OWNER'
);
CREATE TABLE int.int_pi_documents(
    id serial PRIMARY KEY,
    pi_id int,
    parent_id int,
    pi_document_type "int".int_pi_document_type,
    name text,
    relative_path text,
    mime_type text,
    size int,
    s3_bucket_name text,
    s3_object_name text,
    uploaded_by_user_id int,
    uploaded_at timestamptz
);

When executing a request that inserts rows into int.int_pi_directories, my postgraphile plugin will catch it & executes this request :

SELECT lsn, data FROM pg_catalog.pg_logical_slot_get_changes($1, $2, $3, 'add-tables', $4::text)

that request will produce an array of changes (inserts / updates) with a consistent quoting error on columntypes property :

"{\"change\":[{\"kind\":\"update\",\"schema\":\"int\",\"table\":\"int_pi_documents\",\"columnnames\":[\"id\",\"pi_id\",\"parent_id\",\"pi_document_type\",\"name\",\"relative_path\",\"mime_type\",\"size\",\"s3_bucket_name\",\"s3_object_name\",\"uploaded_by_user_id\",\"uploaded_at\"],\"columntypes\":[\"integer\",\"integer\",\"integer\",\"int\".int_pi_document_type,\"text\",\"text\",\"text\",\"integer\",\"text\",\"text\",\"integer\",\"timestamp with time zone\"],\"columnvalues\":[563,1848,682,\"PI\",\"file.pdf\",\"output/file.pdf\",\"application/pdf\",85703,\"test\",\"data\",43,\"2026-08-06 11:22:08.347596+00\"],\"oldkeys\":{\"keynames\":[\"id\"],\"keytypes\":[\"integer\"],\"keyvalues\":[563]}}]}"

(prettified):

{
  "change": [
    {
      "kind": "update",
      "schema": "int",
      "table": "int_pi_documents",
      "columnnames": [
        "id",
        "pi_id",
        "parent_id",
        "pi_document_type",
        "name",
        "relative_path",
        "mime_type",
        "size",
        "s3_bucket_name",
        "s3_object_name",
        "uploaded_by_user_id",
        "uploaded_at"
      ],
      "columntypes": [
        "integer",
        "integer",
        "integer",
        "int".int_pi_document_type, // here lies the problem
        "text",
        "text",
        "text",
        "integer",
        "text",
        "text",
        "integer",
        "timestamp with time zone"
      ],
      "columnvalues": [
        563,
        1848,
        682,
        "PI",
        "file.pdf",
        "output/file.pdf",
        "application/pdf",
        85703,
        "test",
        "data",
        43,
        "2026-08-06 11:22:08.347596+00"
      ],
      "oldkeys": {
        "keynames": ["id"],
        "keytypes": ["integer"],
        "keyvalues": [563]
      }
    }
  ]
}

That formatting issue causes my plugin to crash when trying to parse the output into an object using JSON.parse.
https://github.com/graphile/graphile-engine/blob/v4/packages/lds/src/pg-logical-decoding.ts#L180 & https://github.com/graphile/graphile-engine/blob/v4/packages/lds/src/pg-logical-decoding.ts#L79

My analysis of the postgres-wal2json plugin code led me to believe error is near this section :
https://github.com/eulerto/wal2json/blob/master/wal2json.c#L1235

But i'm not confident enough to propose a pull request of bugfix on my own (very little C experience).