darrenkopp / murmurhash-net Goto Github PK

With the same seed, the two algorithms do not generate the same hashes given the same input. Is this expected?

// Example
var murmur86 = MurmurHash.Create128(0, true, AlgorithmPreference.X86);
var murmur64 = MurmurHash.Create128(0, true, AlgorithmPreference.X64);
byte[] data = Guid.NewGuid().ToByteArray();

// Assert fails
Assert.Equal(murmur86.ComputeHash(data), murmur64.ComputeHash(data));

I ran into this on accident because the default argument, AlgorithmPreference.Auto picks based on bitness of the application. I was generating hashes from two different applications against the same data store. One of those applications happened to be 32bit and the other 64.

Is your feature request related to a problem? Please describe.

In dependency track this package is reporting as not having a licence type specified

Describe the solution you'd like

The PackageLicenseExpression property is set in the csproj so thay the NUGET package has the information. This can then be included in the SBOM so that analysis can occur in the appropriate tools

Describe alternatives you've considered

Ignore them, bundle licence file into package & reference the file.

Additional context

N/a

i initialize MurMurHash with a fixed seed (131).
Different assemblies calling ComputeHash with a given array of bytes produce different hashes .
I double checked and i swear i'm sober :)

How is that possible?

Hi @darrenkopp

Could you publish a nuget for net core version?

thanks in advance

There should only be a single csproj file which when compiled produces dll's for each framework and all also produces a nuget package containing the appropriate dll's.

Hello.

I just saw the usage of this library in Seq's tutorial: https://blog.datalust.co/serilog-tutorial/ (check for Event type enrichment).

When I created the enrichment listed there I notice the Murmur32 class is a IDisposable and Seq's example does't dispose it.

Should I create this instance every time I need to hash something and then dispose it or I should have a global static instance?

Thanks.

Testing with the value "test", I receive the int value: 332557242 (see here: https://dotnetfiddle.net/p72M2c)

When compared to other implementations in other languages, this is an incorrect value (should be 3127628307).

Currently this library cannot be used in .Net Core apps. It would be great if this library can target .Net Standard.

When the length of the input data is not a multiple of 16, TransformBlock() does not behave correctly. In the following example, the same hash is generated for different byte sequences:

using System.Security.Cryptography;
using System.Text;
using Murmur;

{
    using HashAlgorithm hashAlgorithm = MurmurHash.Create128(preference: AlgorithmPreference.X64);
    TransformSingleByte(hashAlgorithm, 1);
    TransformSingleByte(hashAlgorithm, 1);
    TransformSingleByte(hashAlgorithm, 1);
    TransformSingleByte(hashAlgorithm, 1);
    hashAlgorithm.TransformFinalBlock(Array.Empty<byte>(), 0, 0);
    PrintBytes(hashAlgorithm.Hash!);
}

{
    using HashAlgorithm hashAlgorithm = MurmurHash.Create128(preference: AlgorithmPreference.X64);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 0);
    hashAlgorithm.TransformFinalBlock(Array.Empty<byte>(), 0, 0);
    PrintBytes(hashAlgorithm.Hash!);
}

{
    using HashAlgorithm hashAlgorithm = MurmurHash.Create128(preference: AlgorithmPreference.X64);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 1);
    TransformSingleByte(hashAlgorithm, 1);
    hashAlgorithm.TransformFinalBlock(Array.Empty<byte>(), 0, 0);
    PrintBytes(hashAlgorithm.Hash!);
}

{
    using HashAlgorithm hashAlgorithm = MurmurHash.Create128(preference: AlgorithmPreference.X64);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 1);
    TransformSingleByte(hashAlgorithm, 0);
    TransformSingleByte(hashAlgorithm, 1);
    hashAlgorithm.TransformFinalBlock(Array.Empty<byte>(), 0, 0);
    PrintBytes(hashAlgorithm.Hash!);
}

static void TransformSingleByte(
    HashAlgorithm hashAlgorithm,
    byte @byte) =>
    hashAlgorithm.TransformBlock(new[] {@byte}, 0, 1, null, 0);

static void PrintBytes(IEnumerable<byte> bytes)
{
    var stringBuilder = new StringBuilder();
    foreach (var @byte in bytes)
    {
        stringBuilder.AppendFormat("{0:X2}", @byte);
    }
    Console.WriteLine(stringBuilder.ToString());
}

Output:

BC764CD8DDF7A0CFF126F51C16239658
BC764CD8DDF7A0CFF126F51C16239658
BC764CD8DDF7A0CFF126F51C16239658
BC764CD8DDF7A0CFF126F51C16239658

Summary

I wish for the nuget packages to have a more informative readme.

Details

The nuget package should be using the same readme as the repo to make it as easy as possible for a user to get started with the package.

Does it make sense to add some benchmarks? That would be great to compare performance with another implementations.

The generic HashAlgorithm.TryComputeHash(ReadOnlySpan<byte> source, Span<byte> destination, out int bytesWritten) implementation of this function gives an error on the FinalBlock call.

It would be nice to overload this method