A MessagePack implementation in pure AutoHotkey v2.

MessagePack is an efficient binary serialization format. It lets you exchange data among multiple languages like JSON. But it's faster and smaller. Small integers are encoded into a single byte, and typical short strings require only one extra byte in addition to the strings themselves.

The MsgPack class implements static functions which allow you to decode from and encode to Buffers and Files.

decoded := MsgPack.Decode(myFile)

Values are decoded as primitives or as native Map or Array objects.

AutoHotkey's type system does not map perfectly onto that of the MessagePack specification. There are a few notable differences listed here.

AutoHotkey has no explicit boolean type. As such, the values true and false will be encoded as the integers 1 and 0. To explicitly encode boolean values, use the MsgPack.BTrue and MsgPack.BFalse classes.

Boolean values are decoded as integers by default. To decode them as MsgPack.BTrue and MsgPack.BFalse objects, set MsgPack.NativeBools to false.

Unset variables are encoded as Nil. This is primarilly to allow the encoder to work with sparse arrays. You can also use the MsgPack.Nil object to deliberately encode Nil values.

By default, Nil values are decoded as empty Strings. To decode them as MsgPack.Nil objects, set MsgPack.NativeNils to false.

All AutoHotkey numbers are 64-bit. However, the MessagePack specification makes a distinction between 32-bit and 64-bit floating-point numbers ("Floats" vs "Doubles"). Both of these are always decoded as 64-bit floats.

By default, all floats (as determined by IsFloat) are encoded as 64-bit Doubles. However, if the MsgPack.CompactFloats value is truthy, the encoder will use a lossy heuristic to guess whether or not a float can be safelty encoded in 32 bits. This may result in lost precision and should only be used where that precision doesn't matter and where you are encoding a lot of floats. In most cases, you can spare the 4 bytes.

Ext format family stores a tuple of an integer and a byte array. These values are identified with a length prefix and a type integer. Negative values are reserved for spec-specified types.

You can make an object encodable and decodable by implementing an instance MsgPackEncode and a static MsgPackDecode method:

class EncodableObject {
    MsgPackEncode(writer) => Any

    static MsgPackDecode(writer, length) => EncodableObject
}

Before decoding data with extension types, register type's class with MsgPack. This is not required to merely encode data:

MsgPack.ExtensionTypes[1] := EncodableClass

Encoder and decoder methods take BinaryWriter and BinaryReader objects with which they an read binary data. The underlying data source is unknown to the object. Encode is responsible for writing both the prefix, type, and data array, while decode begins with the prefix and type already consumed.

BEWriter.WriteInt32(writer, 123465) ; or WriteInt16, WriteDouble, etc

See the MsgPackTimestamp type for an example.