README.md

Welcome to GBX.NET 2!

A general purpose library for Gbx files - data from Nadeo games like Trackmania or Shootmania, written in C#/.NET. It supports high performance serialization and deserialization of 200+ Gbx classes.

GBX.NET is not just a library that can read or write Gbx files, but also a modding platform that connects all Nadeo games together. It is exceptionally useful for bulk Gbx data processing, for example, when you want to fix broken sign locator URLs on hundreds of maps.

For any questions, open an issue, join the GameBox Sandbox Discord server or message me via Discord DM: bigbang1112

Navigation

• Supported games
• File types
• Gbx Explorer 2
• Lua support
• Framework support
• Preparation
  • Create a new GBX.NET project (lightweight)
  • Create a new GBX.NET project (Visual Studio Code)
  • Create a new GBX.NET project (Visual Studio)
• IMPORTANT INFO about the LZO and Gbx compression
• Usage (simple examples)
  • Load a map and display block count per block name
  • Modify and save a map
  • Processing multiple Gbx types
  • Read a large amount of replay metadata quickly
• Tool framework
• Clarity
  • Differences between Gbx.Parse/Header/Node
  • Do not repeat gbx.Node.[any] too often!
  • Game Version Interfaces!
• Optimization
  • Trimming (tree shaking)
  • Explicit vs. Implicit parse
  • Only header parse
  • NativeAOT
  • Asynchronous
• Benchmarks
• Build
  • Nightly builds
• License
• Special thanks
• Alternative Gbx parsers

Supported games

Many essential Gbx files from many games are supported:

• Trackmania (2020), July 2024 update
• ManiaPlanet 4(.1), TM2/SM
• Trackmania Turbo
• ManiaPlanet 3, TM2/SM
• ManiaPlanet 2, TM2/SM
• ManiaPlanet 1, TM2
• TrackMania Forever, Nations/United
• Virtual Skipper 5
• TrackMania United
• TrackMania Nations ESWC
• TrackMania Sunrise eXtreme
• TrackMania Original
• TrackMania Sunrise
• TrackMania Power Up
• TrackMania (1.0)

File types

Here are some of the known file types to start with:

Latest extension	Class	Can read	Can write	Other extension/s
Map.Gbx	CGameCtnChallenge	Yes	Yes	Challenge.Gbx
Replay.Gbx	CGameCtnReplayRecord	Yes	No
Ghost.Gbx	CGameCtnGhost	Yes	Yes
Clip.Gbx	CGameCtnMediaClip	Yes	Yes
Item.Gbx	CGameItemModel	Yes	Yes	Block.Gbx
Mat.Gbx	CPlugMaterialUserInst	Yes	Yes
Mesh.Gbx	CPlugSolid2Model	Yes	Yes
Shape.Gbx	CPlugSurface	Yes	Yes
Macroblock.Gbx	CGameCtnMacroBlockInfo	Yes	Yes
LightMapCache.Gbx	CHmsLightMapCache	No	No
SystemConfig.Gbx	CSystemConfig	Yes	Yes
FidCache.Gbx	CMwRefBuffer	Yes	Yes
Scores.Gbx	CGamePlayerScore	Yes	No

Full list of supported file types is available in the SUPPORTED GBX FILE TYPES.

Gbx Explorer 2

Gbx Explorer 2 will be a complete remake of the old Gbx Explorer, fixing most of what the old version lacked. It is planned to be a relatively huge project, so it may take time to do.

Old Gbx Explorer is compatible with GBX.NET 2, but some features have been stripped off.

Lua support

GBX.NET 2 will support dynamic Lua script execution around the .NET code, to simplify the library usage even further.

It should be supported to run offline locally through a Gbx Lua Interpreter tool and in Gbx Explorer 2 in web browser, PWA, and Photino build.

The goal is to also make it viable for NativeAOT.

Framework support

Due to the recently paced evolution of .NET, framework support has been limited only to a few ones compared to GBX.NET 1:

• .NET 8
• .NET 6
• .NET Standard 2.0

You can still use GBX.NET 2 on the old .NET Framework, but the performance of the library could be degraded. Depending on the needs, I can add explicit .NET Framework support, but so far there haven't been ones not replaceable with .NET Standard 2.0.

Preparation

Using the NuGet packages is recommended.

Create a new GBX.NET project (lightweight)

1. Install .NET SDK 8.
   • Windows: here or winget install Microsoft.DotNet.SDK.8 (make sure you have WinGet installed)
   • Linux (just SDK)
2. Create directory for your project (anywhere), go inside it.
3. Create new console project: dotnet new console
4. Add the GBX.NET 2 NuGet package: dotnet add package GBX.NET
5. (optional) Add the GBX.NET.LZO 2 NuGet package: dotnet add package GBX.NET.LZO
6. Open Program.cs with your favorite text editor: code . -g Program.cs (for example)
7. Write code - see Examples (simple).
8. Use dotnet run to run the app.

Steps 2-8:

mkdir MyGbxProject
cd MyGbxProject
dotnet new console
dotnet add package GBX.NET
dotnet add package GBX.NET.LZO
code . -g Program.cs
dotnet run

Create a new GBX.NET project (Visual Studio Code)

1. Install C# Dev Kit extension.
2. Click on Create .NET Project button, or press Ctrl+Shift+P, type .NET: New Project.
3. Select Console App and create your project.
4. Open a new terminal and type dotnet add package GBX.NET to add GBX.NET 2.
5. (optional) Add the GBX.NET.LZO 2 NuGet package: dotnet add package GBX.NET.LZO
6. Write code - see Usage (simple examples).
7. Run and debug as usual, select C# if prompted.

Create a new GBX.NET project (Visual Studio)

1. Create a new Console project
2. Under your project in Solution Explorer, right-click on Dependencies and select Manage NuGet packages...
3. Search GBX.NET and click install
4. Write code - see Usage (simple examples).

IMPORTANT INFO about the LZO and Gbx compression

Reading or writing compressed Gbx files require to include the GBX.NET.LZO 2 library (or any other implementation that uses the ILzo interface). This is how you can include it:

Command line:

dotnet add package GBX.NET.LZO

C# code:

using GBX.NET;
using GBX.NET.LZO;

Gbx.LZO = new Lzo();

You should run this line of code only once for the whole program lifetime.

The compression logic is split up from the read/write logic to allow GBX.NET 2 library to be distributed under the MIT license, as Oberhumer distributes the open source version of LZO under the GNU GPL v3. Therefore, using GBX.NET.LZO 2 requires you to license your project under the GNU GPL v3, see License.

Gbx header is not compressed and can contain useful information (icon data, replay time, ...), and also many of the internal Gbx files from Pak files are not compressed, so you can avoid LZO for these purposes.

Usage (simple examples)

Load a map and display block count per block name

Required packages: GBX.NET, GBX.NET.LZO

This project example expects you to have <ImplicitUsings>enable</ImplicitUsings>. If this does not work for you, add using System.Linq;.

using GBX.NET;
using GBX.NET.Engines.Game;
using GBX.NET.LZO;

Gbx.LZO = new Lzo();

var map = Gbx.ParseNode<CGameCtnChallenge>("Path/To/My.Map.Gbx");

foreach (var block in map.GetBlocks().GroupBy(x => x.Name))
{
    Console.WriteLine($"{block.Key}: {block.Count()}");
}

This will print out all blocks on the map and their count. This code can potentially crash for at least 3 reasons:

1. The Gbx file is not a map. See Explicit vs. Implicit parse in the Optimization part.
2. There's a Gbx exception. See Exceptions in GBX.NET 2 (TBD).
3. There's a file system problem.

Modify and save a map

Required packages: GBX.NET, GBX.NET.LZO

GBX.NET's strength is in its ability to modify Gbx files and save them back. This example shows how you can change the map name of a map:

using GBX.NET;
using GBX.NET.Engines.Game;
using GBX.NET.LZO;

Gbx.LZO = new Lzo();

var gbx = Gbx.Parse<CGameCtnChallenge>("Path/To/My.Map.Gbx");
var map = gbx.Node; // See Clarity section for more info

map.MapName = "My new map name";

gbx.Save("Path/To/MyNew.Map.Gbx");

The trick here is that the Gbx properties are saved in the gbx object variable (Gbx class). These properties ensure that the Gbx file is saved correctly across all Trackmania versions.

If you were to go with ParseNode in this case, this would not work for TMF and older games, but it is still possible if you specify the Gbx parameters in the Save method:

map.Save("Path/To/MyNew.Map.Gbx", new()
{
    PackDescVersion = 2 // Latest known PackDesc version in TMF
});

For TMS or TMN ESWC, you would have to specify ClassIdRemapMode for example:

map.Save("Path/To/MyNew.Map.Gbx", new()
{
    ClassIdRemapMode = ClassIdRemapMode.Id2006
    PackDescVersion = 1
});

These save parameters depend on the game of choice, but since Trackmania 2, this does not matter and you can safely use Save method on CMwNod without additional settings.

For more details, see Differences between Gbx.Parse/Header/Node in the Clarity section.

Processing multiple Gbx types

Required packages: GBX.NET, GBX.NET.LZO

This project example expects you to have <ImplicitUsings>enable</ImplicitUsings>. If this does not work for you, add using System.Linq;.

This example shows how you can retrieve ghost objects from multiple different types of Gbx:

using GBX.NET;
using GBX.NET.Engines.Game;
using GBX.NET.LZO;

Gbx.LZO = new Lzo();

var node = Gbx.ParseNode("Path/To/My.Gbx");

var ghost = node switch
{
    CGameCtnReplayRecord replay => replay.GetGhosts().FirstOrDefault(),
    CGameCtnMediaClip clip => clip.GetGhosts().FirstOrDefault(),
    CGameCtnGhost g => g,
    _ => null
};

if (ghost is null)
{
    Console.WriteLine("This Gbx file does not have any ghost.");
}
else
{
    Console.WriteLine("Time: {0}", ghost.RaceTime);
}

Using pattern matching with non-generic Parse methods is a safer approach (no exceptions on different Gbx types), but less trim-friendly, see Explicit vs. Implicit parse in the Optimization section.

Read a large amount of replay metadata quickly

Required packages: GBX.NET

In case you only need the most basic information about many of the most common Gbx files (maps, replays, items, ...), do not read the full Gbx file, but only the header part. It is a great performance benefit for disk scans.

Note

Reading only the header also does not infect you with GNU GPL v3 and you can use licenses compatible with MIT. Header is not compressed with LZO.

This project example expects you to have <ImplicitUsings>enable</ImplicitUsings>. If this does not work for you, add using System.IO;.

using GBX.NET;
using GBX.NET.Engines.Game;

foreach (var filePath in Directory.EnumerateFiles("Path/To/My/Directory", "*.Replay.Gbx", SearchOption.AllDirectories))
{
    try
    {
        DisplayBasicReplayInfo(filePath);
    }
    catch (Exception ex)
    {
        Console.WriteLine($"Gbx exception occurred {Path.GetFileName(filePath)}: {ex}");
    }
}

void DisplayBasicReplayInfo(string filePath)
{
    var nodeHeader = Gbx.ParseHeaderNode(filePath);

    if (nodeHeader is CGameCtnReplayRecord replay)
    {
        Console.WriteLine($"{replay.MapInfo}: {replay.Time}");
    }
}

This code should only crash in case of a file system problem. Other problems will be printed out in the console.

Note

It is still valuable to parse the full Gbx even when you just want a piece of information available in header, because body info overwrites header info. So you can use the benefit of full parse to fool people tackling with the Gbx header.

Tool framework

Tool framework (GBX.NET.Tool* libraries) is a simple way to create rich tools that can be adapted to different environments.

Currently supported environments:

• Console (GBX.NET.Tool.CLI)

Planned environments:

• Blazor (GBX.NET.Tool.Blazor) - both server and WebAssembly
• HTTP server (GBX.NET.Tool.Server)
• Discord bot (GBX.NET.Tool.Discord.Bot)

The tool framework guides you with this project structure:

graph LR
    A(GBX.NET.Tool) --> B(YourToolProject)
    C(GBX.NET.Tool.CLI) --> D(YourToolProjectCLI)
    B --> D
    A --> C
    E(GBX.NET) --> A
    F(GBX.NET.LZO) --> C
    E --> F
    G(Spectre.Console) --> C

Loading

• Structure: Tool library (YourToolProject) and at least 1 implementation application of it (YourToolProjectCLI).
• Tool library references GBX.NET.Tool and implementation application references GBX.NET.Tool.CLI.
• GBX.NET.Tool.CLI currently uses LZO by default, no need to reference it additionally.

Tool library has a primary tool class that implements ITool interface. There should be only one.

Tool class accepts input through constructors (best one is picked according to input provided by implementation). The tool can output as "produce" (IProductive), which creates objects without mutating the input (for example, create MediaTracker clip from replay inputs), or "mutate" (IMutative) which creates objects while also causing changes to the input (for example, modifying a map without having to recreate it again).

Samples are available here.

Clarity

This section describes best practices to keep your projects clean when using GBX.NET 2.

Differences between Gbx.Parse/Header/Node

Gbx files contain many different parameters that are not exactly part of the game objects. We commonly use ParseNode or ParseHeaderNode to simplify the access level, as Gbx parameters are usually unnecessary to know about, but they have to be present to ensure consistent serialization.

You can still save nodes into Gbx files by using the Save method - be careful specifying the Gbx parameters correctly, like the class ID mappings (wrap/unwrap).

• Gbx.Parse
  • Reads the full Gbx file, saves the Node and many of its parameters to the Gbx object, and returns the Gbx object.
• Gbx.ParseNode
  • Same as Gbx.Parse, except only the Node itself is returned and Gbx object parameters are discarded and garbage collected, except for GbxRefTable, which is referenced further down the nodes.
  • Can return null on unknown Gbx file, while Gbx.Parse can't.
• Gbx.ParseHeader
  • Reads only the uncompressed Gbx header part, saves the Node and most of its parameters to the Gbx object, and returns the Gbx object.
• Gbx.ParseHeaderNode
  • Same as Gbx.ParseHeader, except only the Node itself is returned and Gbx object parameters are discarded and garbage collected.
  • Can return null on unknown Gbx file, while Gbx.ParseHeader can't.

Do not repeat gbx.Node.[any] too often!

It is fairly common to see people repeat the gbx.Node.something instead of saving the gbx.Node to a new, shorter variable to improve code clarity. The original guide was to rather refer to using Gbx.ParseNode, however, in GBX.NET 2, some Gbx parameters can be lost by using Gbx.ParseNode, so for simplification, Gbx.Parse is recommended to use when the goal is to modify the file.

Note

That doesn't mean you cannot use Gbx.ParseNode for Gbx modification. In fact there's no difference since TM2 for saving nodes retrieved using Gbx.Parse or Gbx.ParseNode. In TMUF and older versions, when you're using Gbx.ParseNode, you may need to specify a few write parameters that were stored in the original Gbx object (which got discarded).

If you're accessing a lot of main node members, prefer saving gbx.Node into an additional variable.

var gbx = Gbx.Parse<CGameCtnChallenge>("Path/To/My.Map.Gbx");
var map = gbx.Node;

Console.WriteLine(map.MapName);
Console.WriteLine(map.AuthorLogin);
Console.WriteLine(map.AuthorNickname);
Console.WriteLine(map.BuildVersion);

It looks better than this:

var gbx = Gbx.Parse<CGameCtnChallenge>("Path/To/My.Map.Gbx");

Console.WriteLine(gbx.Node.MapName);
Console.WriteLine(gbx.Node.AuthorLogin);
Console.WriteLine(gbx.Node.AuthorNickname);
Console.WriteLine(gbx.Node.BuildVersion);

Game Version Interfaces!

Interfaces (short name of Game Version Interfaces) is a new upcoming feature of GBX.NET 2 where you can scope the Gbx classes for specific Trackmania/Shootmania games to hide unrelated properties and avoid large amount of null checks. These null checks will be done for you behind the scenes and will throw exceptions if they are "exceptional" for the game version you pick.

Currently, it is split into 3 ideas:

• Chunk Game Version Specification (partially available now)
  • Every CMwNod has a member GameVersion that can guess from which game the Gbx file is, or where it could be supported.
• Generated Game Version Interfaces (still WIP)
• Generated Builders (still WIP)

Since 2.0.1, the experimentation of Game Version Interfaces started on the CGameCtnChallenge and CGameCtnBlock classes. Example:

using GBX.NET;
using GBX.NET.Engines.Game;
using GBX.NET.Interfaces.Game;

var map = Gbx.ParseNode<CGameCtnChallenge, IGameCtnChallengeTMF>();

foreach (var block in map.GetBlocks())
{
    Console.WriteLine(block.Name);
    // Console.WriteLine(block.Color); -- block color is not available in TMF
}

Optimization

GBX.NET 2 introduced rich optimization techniques to improve both performance and compiled size of your applications.

The goal of these optimizations is to prove that GBX.NET is not "too big" for anything small.

Trimming (tree shaking)

On publish (the final build), you can trim out unused code by using this property in .csproj:

<PropertyGroup>
    <PublishTrimmed>true</PublishTrimmed>
</PropertyGroup>

The library does not load anything dynamically and does not use reflection, so this is fully supported.

GBX.NET is a huge library when everything is included (over 1.5MB), so please use this whenever it's possible. Code was written to be very trimmable, so the difference is huge (much bigger than in GBX.NET v1).

Note

Expect this to work only with dotnet publish.

However, in case you wanna use reflection on GBX.NET, it is strongly recommended to simply turn off trimming of this library. In case of Blazor WebAssembly specifically, it's worth noting that the release build trims automatically, so in case you're using reflection, modify your library reference:

<PackageReference Include="GBX.NET">
    <IsTrimmable>false</IsTrimmable> <!-- add this line -->
</PackageReference>

In case this is not enough, you can specify TrimmerRootAssembly on the project you're building that this library should absolutely not be trimmed:

<ItemGroup>
    <TrimmerRootAssembly Include="GBX.NET" />
</ItemGroup>

Explicit vs. Implicit parse

In the past, the difference between these two used to be only to reduce the amount of written code by the consumer and making the type more strict, the performance was exactly the same.

GBX.NET 2 changes this majorly by making the Explicit parse way more explicit. The Explicit parse runs through a slightly modified code that does slightly simpler things than the Implicit parse:

• The nodes are instantiated right away using the generics.
  • If the node type is more of a base type, there's a check for all inherited classes, in a much smaller switch statement than the Implicit parse uses.
  • If the type cannot match, exception is thrown.
• Smaller switch statement of classes allows effective trimming that can reduce the library size much more than it is able to with the Implicit parse.
• You cannot use Explicit parse on unknown Gbx files.
• Explicit parse is still considered fairly experimental and it might sometimes fail its job.

Explicit parse:

Gbx<CGameCtnChallenge> gbxMap = Gbx.Parse<CGameCtnChallenge>("Path/To/My.Map.Gbx");
CGameCtnChallenge map = Gbx.ParseNode<CGameCtnChallenge>("Path/To/My.Map.Gbx");
Gbx<CGameCtnChallenge> gbxMap = Gbx.ParseHeader<CGameCtnChallenge>("Path/To/My.Map.Gbx");
CGameCtnChallenge gbxMap = Gbx.ParseHeaderNode<CGameCtnChallenge>("Path/To/My.Map.Gbx");

The Implicit parse cannot guess the type right away, but it does not fail on unknown Gbx files. It is also less effective with library trimming.

Implicit parse:

Gbx gbxMap = Gbx.Parse("Path/To/My.Map.Gbx");
CMwNod map = Gbx.ParseNode("Path/To/My.Map.Gbx");
Gbx gbxMap = Gbx.ParseHeader("Path/To/My.Map.Gbx");
CMwNod gbxMap = Gbx.ParseHeaderNode("Path/To/My.Map.Gbx");

To figure out a type, use pattern matching or casting.

Only header parse

As mentioned earlier, you can largely speed up Gbx reading in case your information is available in the header part of Gbx.

However, if the information is something serious, you should still validate it against the body, in other words: read the full Gbx, which this process will use the information from the body instead, and overwrite what was read in the header.

NativeAOT

GBX.NET fully supports NativeAOT, and it is highly recommended to use its potential on smaller-sized applications:

<PropertyGroup>
    <PublishAot>true</PublishAot>
</PropertyGroup>

It also automatically trims the application (no need for <PublishTrimmed>true</PublishTrimmed>).

On basic GBX.NET applications, native compilation has a couple of improvements:

• Reduces trimmed standalone binary size from ~7MB to 2.8MB.
• Startup time is reduced from 50ms to 0.5ms (JIT is removed, so you should be only bottlenecked by disk speed).
• If you're using only the MIT License libraries, you can make your project harder to reverse engineer.
• The app feels generally lighter, but can be slightly slower for long-running process than a runtime app with JIT (very small difference).

Note

Expect this to work only with dotnet publish.

Asynchronous

Reading Gbx files asynchronously is currently only partially supported, but can be already more effective in networking scenarios.

• Asynchronous reading currently only takes effect in the Gbx body part.
• It is used on reading the compressed LZO buffer, which is quite large, so async can have a positive effect.
• It is also planned to use async reading on Gbx map thumbnail JPEG buffer.
• It is not planned to use async reading while reading the contents of decompressed body, as it's already fully there in memory and it could rather hurt the performance than improve it.

The library uses sync method generators to simplify async method definitions without duplicating code.

Benchmarks

TODO

Build

Note

You don't need to build the solution/repository to use GBX.NET, NuGet packages have been made for you. This is only for internal development purposes.

The solution has more than 30 projects for different purposes, make sure you have enough disk space, the full build could take up a couple hundreds of megabytes.

Make sure you have these framework SDKs available:

• .NET 8
• .NET 6
• .NET Standard 2.0

Visual Studio 2022 should be able to install those with default installation settings. Using Visual Studio 2019 will likely not work.

You should also have .NET WebAssembly Build Tools installed additionally to build the full solution. It is required for Gbx Explorer to work properly, as it uses native LZO implementation compiled into WebAssembly.

In Visual Studio, you can just use Build Solution and everything should build. JetBrains Rider has been tested and also works.

In .NET CLI, run dotnet build on the solution (.sln) level.

Nightly builds

Every 5AM UTC, there is a new build from the dev branch published on https://nuget.gbx.tools/ of the base GBX.NET package (soon also GBX.NET.Tool*).

Go to %appdata%/NuGet and modify the NuGet.Config file to include the package source:

<?xml version="1.0" encoding="utf-8"?>
<configuration>
  <packageSources>
    <add key="nuget.gbx.tools" value="https://nuget.gbx.tools/v3/index.json" /> <!-- add this -->
  </packageSources>
</configuration>

Then you can select the nuget.gbx.tools package source and fetch nightly builds from there.

In the past, nightly builds were pushed to GitHub Packages which required you to provide access tokens to be able to read the packages. Nightly builds are no longer pushed to GitHub Packages.

License

GBX.NET 2 is licensed under multiple licenses, depending on the part of the project. Here are the licenses and their directories:

• MIT License
  • Src/GBX.NET
  • Src/GBX.NET.Crypto
  • Src/GBX.NET.Hashing
  • Src/GBX.NET.Json
  • Src/GBX.NET.NewtonsoftJson
  • Src/GBX.NET.Imaging
  • Src/GBX.NET.Imaging.ImageSharp
  • Src/GBX.NET.Imaging.SkiaSharp
  • Src/GBX.NET.Tool
  • Src/GBX.NET.ZLib
  • Src/GBX.NET.Lua
  • Generators
• GNU GPL v3 License
  • Src/GBX.NET.LZO
  • Src/GBX.NET.Tool.CLI
  • Samples
  • Tools
• The Unlicense
  • Resources

The Unlicense also applies on information gathered from the project (chunk structure, parse examples, data structure, wiki information, markdown).

If you use the LZO compression library, you must license your project under the GNU GPL v3.

Special thanks

Without these people, this project wouldn't be what it is today (ordered by impact):

• Stefan Baumann (Solux)
• Melissa (Miss)
• florenzius
• Kim
• tilman
• schadocalex
• James Romeril
• frolad (Juice)
• Mika Kuijpers (TheMrMiku)
• donadigo

And many thanks to every bug reporter!

Alternative Gbx parsers

• gbx-py by schadocalex (advanced read+write Gbx parser specialized on TM2020 and custom items)
• gbx-ts by thaumictom (read-only Gbx parser for TypeScript)
• ManiaPlanetSharp by Solux (C# toolkit for accessing ManiaPlanet data, including read-only Gbx parser used by ManiaExchange)
• pygbx by Donadigo (read-only Gbx parser for Python)