[Cypher] REMOVE inside MERGE ON MATCH throws InvalidSyntaxException

[ikwattro]

Hi,

First of all Happy New Year !!!!

Use case : Remove a label to a node inside a MERGE ON MATCH:

Based on the default console.neo4j.org graph :

MERGE (n:Crew {name:"Morpheus"}) 
ON MATCH REMOVE n:Crew


Error: Invalid input 'R': expected whitespace, comment or SET (line 1, column 44 (offset: 43))
"MERGE (n:Crew {name:"Morpheus"}) ON  MATCH REMOVE n:Crew"

[thobe]

You are right that it probably would be more friendly to let the parser accept more things in ON MATCH and ON CREATE, and then raise a semantic error instead.

The sample query is obviously not valid, even if we were to allow more types of mutations in ON MATCH and ON CREATE, since it tries to remove one of the parameters used to match the node, thus violating the invariant of MERGE.

[ikwattro]

@thobe Thanks for your reply.

To be honest, the obvious part was not really there in our team, this is the first time we hear that you can not remove parameters used to match the node. Simply because this query doesn't fail :

Query:
MERGE (n:Crew { name:"Morpheus" })
ON MATCH SET n.name = NULL


Query took 93 ms and returned no rows. 
Updated the graph - set 1 property 

[thobe]

Yes, that one should fail as well, but doesn't.

[ikwattro]

Thanks, good to know then.

So it turns out to be more a feature request at the end, my initial application use case is the following :

MERGE (n:PullRequest {zid: {zid} })
ON MATCH REMOVE n:OpenedPullRequest, n:ClosedPullRequest

[thobe]

That looks a lot more sensible, and I can understand the desire to want to do that.

You could of course just REMOVE those labels in both the case of the node being matched and in the case of it being created, the effect would be the same: that the node does not have those labels, regardless of whether it had them before or not.

MERGE (n:PullRequest {zid: {zid} })
REMOVE n:OpenedPullRequest, n:ClosedPullRequest

[ikwattro]

Sure yes.

[mvoila]

I fall in the same use-case.
Is it implemented now?