Bug in query variables whe used with @cascade

[vikram-ma]

I have a schema, where,

• one person can own multiple accounts (account1, account2)
• each account is is subscribed exactly one service (account1 -> service1, account2 -> service2)

For a user id say “1”, I want to get his “owns” predicate, which has subscribed to “service1”
Below query with @cascade works fine, and I get exactly one uid

{
  my_activities(func: eq(Id, 1)) @cascade {
    owns {
      uid
      subscribedTo @filter(eq(name, "service1"))
    }
  }
}

But when I use the same query in upsert, behavior is little bit strange. Even with cascade, I see multiple uid being returned

upsert{
  query{
    my_activities(func: eq(Id, 1)) @cascade {
      owns {
        subscribedTo @filter(eq(name, "Service1"))
	instance as uid
      }
  }
  mutation @if(eq(len(instance), 2) ) {
    # this block executes!!!
  }
  mutation @if(eq(len(instance), 1) ) {
     # this block DOESNOT execute. I was expecting this block to be executed.
  }
}

Note: See the discussion here https://discuss.dgraph.io/t/bug-in-cascade-directive/5896

[sleto-it]

Hi @vikram-ma,

Many thanks for opening this issue. We are on it and will share an update or ask more questions as soon as possible

Thanks,

[mangalaman93]

I am able to reproduce this issue as follows:

schema:

<Id>: int @index(int) .
<name>: string .
<owns>: [uid] .
<subscribedTo>: [uid] .

data:

{
  set {
    _:user <Id> "1" .
    _:user <owns> _:entity1 .
    _:entity1 <subscribedTo> _:service1 .
    _:service1 <name> "service1" .

    _:user <owns> _:entity2 .
    _:entity <subscribedTo> _:service2 .
    _:service2 <name> "service2" .
  }
}

query:

{
  my_activities(func: eq(Id, 1)) @cascade {
    owns {
      uid
      subscribedTo @filter(eq(name, "service1"))
    }
  }
}

upsert:

upsert{
  query{
    my_activities(func: eq(Id, 1)) @cascade {
      owns {
        subscribedTo @filter(eq(name, "Service1"))
        instance as uid
      }
    }
  }

  mutation @if(eq(len(instance), 2)) {
    set {
      _:data <wrong> "3" .
    }
  }

  mutation @if(eq(len(instance), 1)) {
    set {
      _:a <right> "3" .
    }
  }
}

Response that we get for final upsert:

{
  "data": {
    "code": "Success",
    "message": "Done",
    "queries": {
      "my_activities": []
    },
    "uids": {
      "data": "0x7"
    }
  },
  "extensions": {...},
    "txn": {
      "start_ts": 12,
      "commit_ts": 13,
      "preds": [
        "1-wrong"
      ]
    }
  }
}

There are two problems with this response:

1. "my_activities": [] query response is empty.
2. Incorrect mutation was executed.

[mangalaman93]

This is actually problem with our query processing logic, following query returns:

{
  var(func: eq(Id, 1)) @cascade {
    owns {
      instance as uid
      subscribedTo @filter(eq(name, "service1"))
    }
  }
    
  
  q(func: uid(instance)){
    uid
  }
}

following response:

{
  "data": {
    "q": [
      {
        "uid": "0x2"
      },
      {
        "uid": "0x4"
      }
    ]
  }
}

[mangalaman93]

And, this is a duplicate of #4024.