Message Structure

Each message uses the following structure:

- [7 bits] Message Type
- [1 bit ] Compression Flag
- [0/4 bytes] Decompressed Size
- [Remaining bytes] Message

Messages are deliberately simple, to avoid unnecessary complexity.

Message Type

This is denoted by TMessageType within the library, and is user defined.
Each message type can be assigned 1 handler.

Type: Signed byte
Values: 0x0 - 0x7F
Mask: 0b0111_1111

Compression Flag

Set in leftmost bit of message type.
1 if the message contains compression, else 0.

Type: Single bit
Values: 0x0 - 0x1
Mask: 0b1000_0000

Decompressed Size

Expected size after decompression.
Used for compressors to allow for memory pre-allocation.

Type: Signed 32-bit integer.
Optional: Used only if struct defines a compressor.

Values:
- 0x0 - 0x7FFFFFFF: Expected size of decompressed message.

Uses Little Endian.

The resulting data is always compressed, regardless of whether it ended up being larger or smaller than the original in order to improve throughput (avoids a memory copy).

Reason for signed vs unsigned lies in limitations of some languages (incl. .NET), memory pools often don't support 4GB arrays.

Message

Raw message in bytes.
How message is stored depends on serializer.