I just love this idea:
But when you have an input such as
[42,1,2,3,4,5,6] you just get
 instead of
[1,2,3,4,5,6]. You can get a better
result if you use divide and conquer. And that is something Stalin
would do, don’t you think? So I implemented a recursive solution in
Java. It tries all possible results, which is not a single pass and therefore
defeats the purpose of having a fast
O(n) algorithm. And my implementation isn’t even in-place. A new data structure has to be created. That’s why I also added an
implementation using a linked list, which is what Mathew describes.
The code is on pastebin: https://pastebin.com/rEmvZSiA
Here’s an implementation of a multiset in Java 8. It uses a
Map<T, Integer> to count how many times an object was inserted (multiplicity). It is not thread-safe, but there’s
synchronizedMultiset(...) and it always knows the current size in
O(1). With the method
asMap() you can get a view of the multiset as a
Map, that will allow modification of the underlying multiset.
The code is on GitHub: github.com/claudemartin/multiset
This started as a simple implementation and is not rather complex. The iterator allows to remove elements during iteration. I tried to write my own
Spliterator, but it wasn’t faster than the default implementaion. Many methods are optimized and I added some set operations for multisets (union, intersect, minus). With
merge(...) you can merge two multisets with any operation.