Optimising binary serialization for multi-dimensional generic arrays

Question

I have a class that I need to binary serialize. The class contains one field as below:

private T[,] m_data;

These multi-dimensional arrays can be fairly large (hundreds of thousands of elements) and of any primitive type. When I tried standard .net serialization on an object the file written to disk was large and I think .net is storing a lot of repeated data about element types and possibly not as efficiently as could be done.

I have looked around for custom serializers but have not seen any that deal with multi-dimensional generic arrays. I have also experimented with built-in .net compression on a byte array of the memory stream following serializing with some success, but not as quick / compressed as I had hoped.

My question is, should I try and write a custom serializer to optimally serialize this array for the appropriate type (this seems a little daunting), or should I use standard .net serialization and add compression?

Any advice on the best approach would be most appreciated, or links to resources showing how to tackle serialization of a multi-dimensional generic array - as mentioned existing examples I have found do not support such structures.

Answer 1

Here's what I came up with. The code below makes an int[1000][10000] and writes it out using the BinaryFormatter to 2 files - one zipped and one not.

The zipped file is 1.19 MB (1,255,339 bytes) Unzipped is 38.2 MB (40,150,034 bytes)

        int width = 1000;
        int height = 10000;
        List<int[]> list = new List<int[]>();
        for (int i = 0; i < height; i++)
        {
            list.Add(Enumerable.Range(0, width).ToArray());
        }
        int[][] bazillionInts = list.ToArray();
        using (FileStream fsZ = new FileStream("c:\\temp_zipped.txt", FileMode.Create))
        using (FileStream fs = new FileStream("c:\\temp_notZipped.txt", FileMode.Create))
        using (GZipStream gz = new GZipStream(fsZ, CompressionMode.Compress))
        {
            BinaryFormatter f = new BinaryFormatter();
            f.Serialize(gz, bazillionInts);
            f.Serialize(fs, bazillionInts);
        }

I can't think of a better/easy way to do this. The zipped version is pretty damn tight.

I'd go with the BinaryFormatter + GZipStream. Making something custom would not be fun at all.

---

[edit by MG] I hope you won't be offended by an edit, but the uniform repeated Range(0,width) is skewing things vastly; change to:

        int width = 1000;
        int height = 10000;
        Random rand = new Random(123456);
        int[,] bazillionInts = new int[width, height];
        for(int i = 0 ; i < width;i++)
            for (int j = 0; j < height; j++)
            {
                bazillionInts[i, j] = rand.Next(50000);
            }

And try it; you'll see temp_notZipped.txt at 40MB, temp_zipped.txt at 62MB. Not so appealing...