Neo4j Merge and Atomic Transaction

问题

I am running the below MERGE query against my Neo4j server from a client application in 10 parallel threads, the newFoo and id parameters are the same on all 10 runs:

MERGE (foo:Foo { id: {id} })
ON MATCH
SET foo = {newFoo}

After running this, I run the below query to expect 1 but I instead get 10:

match (f:Foo)
return count(f)

I thought that MERGE runs in an atomic transaction but apparently not. Am I doing something wrong here?

Update

Below is the code that I used to reproduce the issue:

public static async Task RunInParallel()
{
    var client = new GraphClient(new Uri("http://localhost:7474/db/data"), "neo4j", "1234567890")
    {
        JsonContractResolver = new CamelCasePropertyNamesContractResolver()
    };

    client.Connect();

    var foo = new Foo
    {
        Id = "1",
        Name = "Foo",
        Slug = "foo-bar-foo"
    };

    List<Task> tasks = new List<Task>();
    for (int i = 0; i < 10; i++)
    {
        var task = client.Cypher
            .Merge("(foo:Foo { id: {id} })")
            .OnMatch()
            .Set("foo = {newFoo}")
            .WithParams(new
            {
                Id = foo.Id,
                NewFoo = foo
            })
            .ExecuteWithoutResultsAsync();

        tasks.Add(task);
    }

    await Task.WhenAll(tasks.ToArray());
}

回答1:

MERGE (by itself) does not guarantee uniqueness. When using MERGE (on a unique property) you should always create a uniqueness constraint for the specified property:

CREATE CONSTRAINT ON (f:Foo) ASSERT f.id IS UNIQUE

This will ensure you aren't creating any duplicates.

Edit

MERGE without a uniqueness constraint is not thread safe. Adding a uniqueness constraint ensures an index lock is held before writing, making the operation thread safe.

来源：https://stackoverflow.com/questions/34302176/neo4j-merge-and-atomic-transaction