Description
Since upgrading to version 1.0.0 of the library, I've noticed that the update mutations that we translate to Cypher are not including our custom 'ID' type field in the SET statement. For various reasons, we need to use a separately declared 'ID' field that isn't _id.
The effect of this is that the ID field is dropped from the resulting object in the Neo4j database.
Here's a simplified example based on our current usecase:
Schema:
type Person {
userId : ID!
username: String
}
Input GraphQL:
mutation {
updatePerson(username: $username, userId: $userId) { username userId }
}
Output Cypher:
MATCH (updateParson:Person { userId: $userId }) SET updatePerson = { username: $username } WITH updatePerson RETURN updatePerson { .username, .userId } AS updatePerson
Relevant Java translation code:
final GraphQLSchema graphQLSchema = SchemaBuilder.buildSchema(schema);
translator = new Translator(graphQLSchema);
final List<Cypher> cypher = translator.translate(query, variables);
Pre-update Person:
{
userId: "<SOME USER ID>",
username: "John.Smith"
}
Post-update Person:
{
username: "Jeremy.Smith"
}
This used to work correctly in 1.0.0-M03 (see relevant code here), where the userId would correctly be passed through to the resulting Cypher.
In version 1.0.0, the Cypher is built here. It doesn't explicitly exclude the userId field, but I have a feeling that it's related to the fact that it's an ID type field.
This issue could possibly be related to issue 83.