C#•14mo ago

Deserializing binary data

I have the following structures:

Root:
    int32             nameLength
    byte[nameLength]    name
    int32            numItems
    Item[numItems]        Items

Item:
    int32             itemNameLength
    byte[itemNameLength]    itemName
    int32            VertexLength
    Vertex[ArrayLength]    VertexList

Vertex:
    float    X
    float     Y
    float     z

Root:
    int32             nameLength
    byte[nameLength]    name
    int32            numItems
    Item[numItems]        Items

Item:
    int32             itemNameLength
    byte[itemNameLength]    itemName
    int32            VertexLength
    Vertex[ArrayLength]    VertexList

Vertex:
    float    X
    float     Y
    float     z

I am curious, what is the best way to serialize this? Currently I simply read it into a stream and reads one attribute at the time. Coming from a c++ background this feels highly inefficent. When I try to google the problem I dont get a lot of new results, but I suspect there might be some changes to the language in the last 10 years which have improved tasks like this?

10 Replies

TheRanger•14mo ago

whats this, yaml?

Thinker•14mo ago

Looks like a custom format? You'd probably have to write your own parser for it, although should be pretty simple

Ole (ping please)OP•14mo ago

I just trying to explain the data layout. Its a custom format that there is no real documentation or name for. I have made a parser which basically reads ffield by field, but I am wondering if there is a "smarter" way of doing this. I have a lot of binary structures to serialize and deserialize so I am looking for improvments 🙂

Thinker•14mo ago

Other than just general performance improvements (eg. using spans), you can't get much better than a hand-crafted parser There are libraries like Pidgin which allow you to write parsers in a simpler way, but you're still making a parser.

Ole (ping please)OP•14mo ago

This is my current approach:

public class PluginPropertyValue
{
    public uint propertyId { get; set; }
    public byte rtpcAccum { get; set; }
    public float fValue { get; set; }

    public static PluginPropertyValue Create(ByteChunk chunk)
    {
        var instance = new PluginPropertyValue();

        instance.propertyId = chunk.ReadUInt32();
        instance.rtpcAccum = chunk.ReadByte();
        instance.fValue = chunk.ReadSingle();

        return instance;
    }
}

public class PluginPropertyValue
{
    public uint propertyId { get; set; }
    public byte rtpcAccum { get; set; }
    public float fValue { get; set; }

    public static PluginPropertyValue Create(ByteChunk chunk)
    {
        var instance = new PluginPropertyValue();

        instance.propertyId = chunk.ReadUInt32();
        instance.rtpcAccum = chunk.ReadByte();
        instance.fValue = chunk.ReadSingle();

        return instance;
    }
}

Where chunk is a helper around a memory stream

nukleer bomb•14mo ago

If you just need a binary (de)serializator, then you can use MemoryPack

nukleer bomb•14mo ago

https://github.com/Cysharp/MemoryPack

GitHub

GitHub - Cysharp/MemoryPack: Zero encoding extreme performance bina...

Zero encoding extreme performance binary serializer for C# and Unity. - GitHub - Cysharp/MemoryPack: Zero encoding extreme performance binary serializer for C# and Unity.

wasabi•14mo ago

A decent way to write these, that I usually end up doing, is to use a SequenceReader, and a series of record structs. I split it up like this. I have a ref struct named FooFormatReader. On that, I add some TryRead* methods, broken down into the basic operations of how the data is composed. FooFormatReader holds a reference of SequenceReader<byte>, and dispatches it's TryReadBar methods to that. Then I have a number of record structs, which represent the compound structures in the format. Those end up with TryRead(ref FooFormatReader reader, out FooRecord foo) that call the FooFormatReader, and initialize themselves.

wasabi•14mo ago

Here's an example: https://github.com/ikvmnet/ikvm/blob/main/src/IKVM.ByteCode/Parsing/ClassRecord.cs

GitHub

ikvm/src/IKVM.ByteCode/Parsing/ClassRecord.cs at main · ikvmnet/ikvm

A Java Virtual Machine and Bytecode-to-IL Converter for .NET - ikvmnet/ikvm

jcotton42•14mo ago

I've done the same for reading Minecraft's network protocol, though I made them extension methods on SequenceReader the use of ref for the reader (or ref this if it's an ext method) is really important

Gaming

Programming

Deserializing binary data

Did you find this page helpful?