Generating hashes of ReadOnlySpan<char> that are equivalent to string.GetHashCode()

[PatVax]

I am writing a Specialized Dictionary class for string values where the lookup is done by a char sequence. This could be any string, ReadOnlySpan<char>, ReadOnlyMemory<char>, ReadOnlySequence<char>. The goal is to retrieve a string that is equivalent to the input without allocations or create it when it doesn't exist. I know that we have a ReadOnlySpan<char> overload of string.GetHashCode but none for ReadOnlySequence<char>.

To make it work with ReadOnlySequence<char> there are following possibilities I could think of so far:

• Check if the sequence is made of a single segment and if so use the FirstSpan property with the GetHashCode overload or if it is made of multiple segments copy the data into a stack allocated buffer and pass this span into GetHashCode. This is almost optimal. Only in the case of multiple segments we need a seemingly unnecessary data copy.
• Use the HashCode struct to produce hashes in every case. In this case I am not sure if the resulting hashes would be as good as the ones produced in string.GetHashCode with the Marvin algorithm. It would look something like this:

int Hash(ReadOnlySequence<char> seq)
{
    HashCode hash = new();
    foreach (var memory in seq)
    foreach (var c in memory)
        hash.Add(c);
    return hash.ToHashCode();
}

int Hash(ReadOnlySpan<char> span)
{
    HashCode hash = new();
    foreach (var c in span)
        hash.Add(c);
    return hash.ToHashCode();
}

int Hash(string s) => Hash(s.AsSpan());
int Hash(ReadOnlyMemory<char> memory) => Hash(memory.Span);

• Roll out my own hashing algorithm like Murmur2 as compared in this stackoverflow answer and use it for all cases.

So what I am looking for is a .NET built in way to produce string.GetHashCode equivalent hash from ReadOnlySequence<char> without making a copy of the sequence when it is made of multiple segments.

[EgorBo]

Consider using XxHash3, we might even use it as a default String.GetHashCode algorithm in future (#85206)

[PatVax]

I just found XxHash in System.IO.Hashing namespace. Is this the right thing to use?

[bartonjs]

If you want the same hashes as are produced by System.String; I believe they get exposed via CompareInfo.GetHashCode(ReadOnlySpan<char>, CompareOptions)

[PatVax]

If you want the same hashes as are produced by System.String; I believe they get exposed via CompareInfo.GetHashCode(ReadOnlySpan<char>, CompareOptions)

Yes I got that far. I think the static version of string.GetHashCode is essentially the same as CompareInfo.GetHashCode(ReadOnlySpan<char>, CompareOptions). This is fine. I was looking for an overload with CompareInfo.GetHashCode(ReadOnlySequence<char>, CompareOptions) my guess is that it simply doesn't exist. For know I am going to use XxHash3 from System.IO.Hashing for anything that could be considered a string. If sometime in the future ReadOnlySequence<char> overloads are built in I will gladly switch back.