set contains(obj) returns true but remove(obj) fails.

[bytor99999]

OK, this is related to my vertex question. But this looks like the brunt of my problem.

In my User object I have a method to remove the ItemUserSignedUpToBringToEvent from its Set collection. I have implemented ItemUserSignedUpToBringToEvent's equals method to compare nodeIds.

public boolean removeItemSignedUpFor(ItemUserSignedUpToBringToEvent itemSignedUp) {
        return (itemsSignedUpFor.contains(itemSignedUp) &&
                itemsSignedUpFor.remove(itemSignedUp));
    }

contains call returns true, but the call to remove fails. Why???

Here is my equals and hashcode for the ItemUserSignedUpToBringToEvent

    @Override
    public boolean equals(Object o) {
        if (this == o) return true;
        if (o == null || getClass() != o.getClass()) return false;

        ItemUserSignedUpToBringToEvent signedUpItem = (ItemUserSignedUpToBringToEvent) o;
        if (nodeId == null) return super.equals(o);
        return nodeId.equals(signedUpItem.nodeId);

    }

    @Override
    public int hashCode() {
        return nodeId != null ? nodeId.hashCode() : super.hashCode();
    }

In stepping through code, the Spring Data Neo4J class ManagedFieldAccessorSet remove method is returning false.

@Override
    public boolean remove(Object o) {
        if (delegate.remove(o)) {
            update();
            return true;
        }
        return false;
    }

Thanks

Mark

[bytor99999]

I hate to say this but I had to do a workaround. Basically had to rewrite the remove() method of my Set instead. So basically not use the Java algorithm Big O notation for removal in a HashSet, which is backed by a HashMap.

Instead my workaround is brute force or
I should say brute recreate Set by copying the elements of the first set that I want to keep into the new Set. But it does the trick for now, I am done spending 2 weeks on this one problem. If you remember a couple months ago I posted about this problem but never got resolved.

Ugly code. (I guess you could call this a Joel Spolsky Duct Tape programmer solution) I apologize in advanced.

public boolean removeItemSignedUpFor(ItemUserSignedUpToBringToEvent itemSignedUp) {
        boolean results = false;
        HashSet<ItemUserSignedUpToBringToEvent> newSet = new HashSet<ItemUserSignedUpToBringToEvent>();
        if (this.itemsSignedUpFor.contains(itemSignedUp)) {
            for (ItemUserSignedUpToBringToEvent inUser : this.itemsSignedUpFor) {
                if (!inUser.equals(itemSignedUp)) {
                    newSet.add(inUser);
                } else {
                    results = true;
                }
            }
        }
        this.itemsSignedUpFor = newSet;
        return results;
    }

Mark

[bytor99999]

Michael, I am going to raise a Jira issue for this, although I have no idea what is causing this problem. I have done the same for a few other collections and that same code worked perfectly. And I don't think I could make a test that duplicates it either.

Thanks

Mark

[vmarcinko]

is there JIRA issue raised for this?
I also ran into this weird situation - I have parent containing children entitities, and childrenSet.contains(child) == true
but
childrenSet.remove(child) == false
?1

[bytor99999]

OK, just created the Jira

https://jira.springsource.org/browse/DATAGRAPH-267

Are you using simple mapping or the advanced. I am using Simple at the moment, and wondering if that might have an affect.

But remember, I am using my hack/cludge as a workaround for now until 2.1 is at least RC1 or 2.

Mark
Mark

[MichaelHunger]

Any chance for a test-case-project? I have a test for remove in SDN which works fine. Seems to be dependent on something else like equals/hashCode?

I don't think super.hashCode and super.equals work very well in this case. As objects can get an id after they were put into the set and then their equals/hashCode might change.

[lassewesth]

Here is my equals and hashcode for the ItemUserSignedUpToBringToEvent

    @Override
    public boolean equals(Object o) {
        if (this == o) return true;
        if (o == null || getClass() != o.getClass()) return false;

        ItemUserSignedUpToBringToEvent signedUpItem = (ItemUserSignedUpToBringToEvent) o;
        if (nodeId == null) return super.equals(o);
        return nodeId.equals(signedUpItem.nodeId);

    }

    @Override
    public int hashCode() {
        return nodeId != null ? nodeId.hashCode() : super.hashCode();
    }

In stepping through code, the Spring Data Neo4J class ManagedFieldAccessorSet remove method is returning false.

@Override
    public boolean remove(Object o) {
        if (delegate.remove(o)) {
            update();
            return true;
        }
        return false;
    }

Mark,

The delegate for ManagedFieldAccessorSet should be of type HashSet, and as such, it will look at equals + hashcode for identity. That snippet of code in ManagedFieldAccessorSet is so simple we can ignore it I think.

So my guess is, there is something subtle going on with your equals and hashcode implementations.

• try a hashcode implementation of 'return 42;' - not for production use, just to eliminate that as a source of problems
• for equals, try 'return o instanceof ItemUserSignedUpToBringToEvent && id.equals( (((ItemUserSignedUpToBringToEvent) o).getId()) );' - does that make your test pass?
• I guess it can be reduced to writing tests for inserting and removing from a HashSet, try writing a unit test like that, to gain confidence your equals and hashcode work independently of SDN

I just implemented a test to replicate, and actually managed to get equals+hashcode wrong in the first instance, which reproduced your problem Anyway, I got my test passing, so please try these simple remedies, and if it doesn't work I'll have another look.

Regards,

Lasse

[bytor99999]

Thanks guys. Lasee, I will test your suggestions. And see if I can isolate my stuff for a good sample.

Mark

[bytor99999]

Although, what is odd, is that in many other collections it works. And that all my entity/domain objects have basically the exact same equals/hashcode. I just copied and pasted the code in all my objects, then just had to change the types in the code to match the domain object it was in.

Mark

[vmarcinko]

Just want to point out that I also created yesterday the issue tied to this (maybe not the best issue title):
https://jira.springsource.org/browse/DATAGRAPH-265

But I finally think that crux of problem is described in:
http://forum.springsource.org/showth...o-lazy-loading