I had to implement some data structures for my computational geometry class. Deciding whether to implement the data structures myself or using the build-in classes turned out to be a hard decision, as the runtime complexity information is located at the method itself, if present at all. So I went ahead to consolidate all the information in one table, then looked at the source code in Reflector and verified them. Below is my result.
| Internal Implement- ation |
Add/insert | Add beyond capacity | Queue/Push | Dequeue/ Pop/Peek |
Remove/ RemoveAt |
Item[i]/Find(i) | GetEnumerator | MoveNext | |
| List | Array | O(1) to add, O(n) to insert | O(n) | - | - | O(n) | O(1) | O(1) | O(1) |
| LinkedList | Doubly linked list | O(1), before/after given node | O(1) | O(1) | O(1) | O(1), before/after given node | O(n) | O(1) | O(1) |
| Stack | Array | O(1) | O(n) | O(1) | O(1) | - | - | O(1) | O(1) |
| Queue | Array | O(1) | O(n) | O(1) | O(1) | - | - | O(1) | O(1) |
| Dictionary | Hashtable with links to another array index for collision | O(1), O(n) if collision | O(n) | - | - | O(1), O(n) if collision | O(1), O(n) if collision | O(1) | O(1) |
| HashSet | Hashtable with links to another array index for collision | O(1), O(n) if collision | O(n) | - | - | O(1), O(n) if collision | O(1), O(n) if collision | O(1) | O(1) |
| SortedDictionary | Red-black tree | O(log n) | O(log n) | - | - | O(log n) | O(log n) | O(log n) | O(1) |
| SortedList | Array | O(n) | O(n) | - | - | O(n) | O(1) | O(1) | O(1) |
| SortedSet | Red-black tree | O(log n) | O(log n) | - | - | O(log n) | O(log n) | O(log n) | O(1) |
Note:
| Dictionary | Add, remove and item[i] has expected O(1) running time |
| HashSet | Add, remove and item[i] has expected O(1) running time |
