When to use a SortedList<TKey, TValue> over a SortedDictionary<TKey, TValue>?

Question

This may appear to be a duplicate of this question, which asks "What’s the difference between SortedList and SortedDictionary?" Unfortunately, the answers do nothing more than quote the MSDN documentation (which clearly states that there are performance and memory use differences between the two) but don't actually answer the question.

In fact (and so this question doesn't get the same answers), according to MSDN:

The SortedList<TKey, TValue> generic class is a binary search tree with O(log n) retrieval, where n is the number of elements in the dictionary. In this, it is similar to the SortedDictionary<TKey, TValue> generic class. The two classes have similar object models, and both have O(log n) retrieval. Where the two classes differ is in memory use and speed of insertion and removal:

• SortedList<TKey, TValue> uses less memory than SortedDictionary<TKey, TValue>.

• SortedDictionary<TKey, TValue> has faster insertion and removal operations for unsorted data, O(log n) as opposed to O(n) for SortedList<TKey, TValue>.

• If the list is populated all at once from sorted data, SortedList<TKey, TValue> is faster than SortedDictionary<TKey, TValue>.

So, clearly this would indicated that SortedList<TKey, TValue> is the better choice unless you need faster insert and remove operations for unsorted data.

The question still remains, given the information above what are the practical (real-world, business case, etc.) reasons for using a SortedDictionary<TKey, TValue>? Based on the performance information, it would imply that there really is no need to have SortedDictionary<TKey, TValue> at all.

Answer 1

That's all there is to it. Retrieval of keys is comparable, but addition is much faster with Dictionaries.

I try to use SortedList as much as possible because it allows me to iterate over the keys and value collections. This is not possible with SortedDictionary as far as I know.

I'm not sure about this, but as far as I know Dictionaries store data in Tree structures, whereas List store data in linear arrays. That explains why insertion and removal is much faster with dictionaries, since less memory has to be shifted around. It also explains why you can iterate over SortedLists but not SortedDictionary.

Answer 2

I'm not sure how accurate the MSDN documentation is on SortedList and SortedDictionary. It seems to be saying both are implemented using a binary search tree. But if the SortedList uses a binary search tree, why would it be much slower on additions than SortedDictionary?

Anyway, here are some performance test results.

Each test operates on a SortedList / SortedDictionary containing 10,000 int32 keys. Each test is repeated 1.000 times (Release build, Start without Debugging).

The first group of tests add keys in sequence from 0 to 9,999. The second group of tests add random shuffled keys between 0 to 9,999 (every number is added exactly once).

***** Tests.PerformanceTests.SortedTest

SortedDictionary Add sorted: 4411 ms
SortedDictionary Get sorted: 2374 ms


SortedList Add sorted: 1422 ms
SortedList Get sorted: 1843 ms

***** Tests.PerformanceTests.UnsortedTest

SortedDictionary Add unsorted: 4640 ms
SortedDictionary Get unsorted: 2903 ms


SortedList Add unsorted: 36559 ms
SortedList Get unsorted: 2243 ms

As with any profiling, the important thing is the relative performance, not the actual numbers.

As you can see, on sorted data the sorted list is faster than the SortedDictionary. On unsorted data the SortedList is slightly quicker on retrieval, but about 9 times slower on adding.

If both are using binary trees internally, it is quite surprising that the Add operation on unsorted data is so mcuh slower for SortedList. It is possible that sorted list may also be adding items to a sorted linear data structure at the same time, which would slow it down.

However, you would expect the memory usage of a SortedList to be equal or greater than or at least equal to a SortedDictionary. But this contradicts what the MSDN documentation says.