Question: is possible the values in `left` and `right` are change when inserting a new element? about hashmap HOT 7 CLOSED

You are right: right won't be GC-ed. I was wrong about it.

As you mentioned, the issue is that left or right has a chance it is already removed from the list when this method is called.

from hashmap.

Thanks for your bug report, indeed the old implementation had some issues, the current dev branch has a singly linked list which should not have those issues anymore. Please review and create a new ticket if you see any other issue, thanks!

from hashmap.

There is only a bug_grow_insert branch. Is it the dev branch you mentioned above?

from hashmap.

Since main branch is still the default branch, closing this issue without fixing it is inappropriate for users.

Please keep an issue of an unsolved problem open, to let users know there is a risk in using this lib.

from hashmap.

Main is the default/dev branch, tags are releases. Since the code now uses a singly linked list, this issue is not present or valid anymore and gets closed.

from hashmap.

I'm afraid the problem is still there as long as it uses a pointer address to identify an unchanged node:

The right has a chance it has been changed, but the address is the same.

Please keep this issue open until it's been fixed.

func (l *List[Key, Value]) insertAt(element, left, right *ListElement[Key, Value]) bool {
	if left == nil {
		left = l.head
	}

	element.next.Store(right)

	if !left.next.CompareAndSwap(right, element) {
		return false // item was modified concurrently
	}

	l.count.Add(1)
	return true
}

from hashmap.

Thanks for reviewing it again. I will reopen the issue once I can clearly see/understand the problem.

The only issue I can see here is that if the function insertAt returns false due to a concurrent change of the next pointer of the left node, the inserting should be retried but currently the callers do not do that. Is this what you are referring to?
Edit: HashMap.insertElement already does retry in case of concurrent interfering changes.

Other than that I can see no issue of using a pointer to identify a node. If there is still a reference/pointer to the node object, the GC should not garbage collect the object and reuse it.

from hashmap.