The purpose of this code is to implement something like Dictionary<T, U> where T is List<>
I've seen hashcode calculation based implementations,
but they were too slow for me I did something wrong when benchmarking
and I decided to use tree-based approach, which's explained below.
.NET CLI: dotnet add package DictionaryList.Core --version 1.0.0
var dict = new DictionaryList<int, int>();
var list1 = new List<int> { 1, 2, 3, 4, 5 };
var list2 = new List<int> { 1, 2, 3, 4, 5, 5, 6 };
dict.Add(list1, 5);
dict.Add(list2, 10);
if (dict.TryGet(list1, out var value1))
Console.WriteLine(value1);
dict.TryGet(list2, out var value2);
Console.WriteLine(value2);
var dict = new DictionaryList<int, int>();
var list1 = new List<int> { 1, 2 };
var list2 = new List<int> { 1, 2 };
dict.Add(list1, 5);
dict.Add(list2, 5); // It's fine since Key and Value are the same
dict.Add(list2, 123); // Throws an ArgumentException since Value is different for the same key
DictionaryList's constructor accepts a boolean allow_keys_with_nulls that indicates whether you want to have NULLs inside keys (list).
By default this value is set to false, so it doesn't allow NULLs inside keys.
So basically if you want to work with NULLs like this: new List<User> { new User(), null, new User() }, then you should use either:
var dict = new DictionaryList<Test, int>(true);
or
var dict = new DictionaryList<Test, int>();
dict.AllowNULLsInKeys = true;
If you don't allow NULLs in keys and insert one with NULL then there'll be an exception thrown.
It was possible to do it that way that it'd work fine for both use cases without some boolean but there was some performance penalty that I believe doesn't have to be paid by majority of use cases / users.
Basically it represents all those lists as a tree and stores pairs of list's elements and NULL or if that was the last element, then the value
For example, for those lists, the tree will look like:
Key: { 1, 2, 10 }; Value: 30
Key: { 1, 2, 3 }; Value: 50
Key: { 1, 2, 3, 3 }; Value: 70
Key: { 1, 2, 3, 3, 5 }; Value: 80
Key: { 1, 2, 3, 1 }; Value: 100
Benchmarks