Efficient way to clone a HashSet<T>?

user_98713 0
Viewed 14593

A few days ago, I answered an interesting question on SO about HashSet<T>. A possible solution involved cloning the hashset, and in my answer I suggested to do something like this:

HashSet<int> original = ...
HashSet<int> clone = new HashSet<int>(original);

Although this approach is quite straightforward, I suspect it's very inefficient: the constructor of the new HashSet<T> needs to separately add each item from the original hashset, and check if it isn't already present. This is clearly a waste of time: since the source collection is a ISet<T>, it is guaranteed not to contain duplicates. There should be a way to take advantage of that knowledge...

Ideally, HashSet<T> should implement ICloneable, but unfortunately it's not the case. I also checked with Reflector to see if the HashSet<T> constructor did something specific if the source collection was a hashset, but it doesn't. It could probably be done by using reflection on private fields, but that would be an ugly hack...

So, did someone come up with a clever solution to clone a hashset more efficiently ?

(Note that this question is purely theoretical, I don't need to do that in a real program)

clone hashset performance .net
6 Answers

I checked the .NET Framework source code for both version 4.5.2 and version 4.7.2. Version 4.7.2 does have optimization in the constructor to handle when the passed in collection is of type HashSet, using some internal cloning logic. You would need to also pass in the comparer into the constructor for this logic to work. Version 4.5.2 does NOT have this optimization it seems.

Example:

var clonedSet = new HashSet(set, set.Comparer);
user_1194180
Related