Neo4j Aura and uuid issue

132 views Asked by At

Running this GraphQL mutation

  "Convenience mutation to destructively update DataPoints for a specific SurveyResponse and question. Accepts answers and/or value. This mutation deletes existing relationships that represent answer data. Returns a DataPoint."
  AnswerQuestion(
    "The id of the question"
    id: ID!
    surveyResponse: ID!
    answers: [ID] = []
    value: String = null
  ): DataPoint
    @hasScope(scopes: ["DataPoint: Update"])
    @cypher(
      statement: """
      MATCH (sr:SurveyResponse) WHERE sr.id = $surveyResponse
      MATCH (q:Question) WHERE q.id = $id
      OPTIONAL MATCH (sr)--(oldDp:DataPoint)--(q)
      DETACH DELETE oldDp
      MERGE (sr)-[:HAS_DATA_POINT]->(dp:DataPoint {id: coalesce(dp.id, apoc.create.uuid())})-[:FOR_QUESTION]->(q)
      WITH dp, q
      FOREACH (o IN CASE WHEN $value IS NOT NULL THEN [true] ELSE [] END | MERGE (dp)-[:ANSWERS {value: $value, propToWrite: q.propToWrite}]->(q))
      WITH dp, q
      OPTIONAL MATCH (a:Answer) WHERE a.id IN $answers
      FOREACH (o IN CASE WHEN a IS NOT NULL THEN [a] ELSE [] END | MERGE (dp)-[:SELECTED_ANSWER]->(a))
      RETURN dp
      """
    )

results in this type of error message on aura (4.2.0. enterprise edition - no matter whether this is run against leader or follower nodes):

Failed to invoke procedure apoc.cypher.doIt: Caused by: IndexEntryConflictException{propertyValues=( String("always-the-same-id-that-has-nothing-to-do-with-the-queried-data") ), addedNodeId=-1, existingNodeId=n}``` where "always-the-same-id-that-has-nothing-to-do-with-the-queried-data" and "n" are unimportant values.

The same mutation works fine on localhost (4.2.7 community edition).

When I query on all nodes for the id ("always-the-same-id-that-has-nothing-to-do-with-the-queried-data"), I see one DataPoint and another node which are not connected. Deleting the DataPoint and just running the reduced query below in neo directly yields the same error message.

      MATCH (sr:SurveyResponse) WHERE sr.id = $surveyResponse
      MATCH (q:Question) WHERE q.id = $id
      OPTIONAL MATCH (sr)--(oldDp:DataPoint)--(q)
      DETACH DELETE oldDp
      MERGE (sr)-[:HAS_DATA_POINT]->(dp:DataPoint {id: coalesce(dp.id, apoc.create.uuid())})-[:FOR_QUESTION]->(q) RETURN dp

This let me to believe that the uuid creation on aura is flawed, but running RETURN apoc.create.uuid() multiple times creates different strings. Sneaking in a WITH sr, q between the DETACH DELETE and MERGE statements has no effect.

1

There are 1 answers

0
manonthemat On

Thanks for being my rubber ducks.

This is the diff that made it work:

+      WITH oldDp, oldDp.id AS oldId, sr, q
       DETACH DELETE oldDp
-      MERGE (sr)-[:HAS_DATA_POINT]->(dp:DataPoint {id: coalesce(oldDp.id, apoc.create.uuid())})-[:FOR_QUESTION]->(q)
+      MERGE (sr)-[:HAS_DATA_POINT]->(dp:DataPoint {id: coalesce(oldId, apoc.create.uuid())})-[:FOR_QUESTION]->(q)

I'm sure there's a ton of optimizations to be done, but this is good enough for now and it works. Still no idea why aura just behaves so different at times.