This document is about: QUANTUM 3

DSL (game state)


Quantum requires components and other runtime game state data types to be declared with its own DSL (domain-specific-language).

These definitions are written into text files with the .qtn extension. The Quantum compiler will parse them into an AST, and generate partial C# struct definitions for each type (definitions can be split across as many files if needed, the compiler will merge them accordingly).

The goal of the DSL is to abstract away from the developer the complex memory alignment requirements imposed by Quantum's ECS sparse set memory model, required to support the deterministic predict/rollback approach to simulation.

This code generation approach also eliminates the need to write "boiler-plate" code for type serialization (used for snapshots, game saves, killcam replays), checksumming and other functions, like printing/dumping frame data for debugging purposes.

In order to create a new .qtn file to the project, open the context menu on Unity's Project tab and click in Create/Quantum/Qtn, or simply create a new file with the .qtn extenstion.


Components are special structs that can be attached to entities, and used for filtering them (iterating only a subset of the active entities based on its attached components). This is a basic example definition of a component:


component Action
    FP Cooldown;
    FP Power;

These will be turned into regular C# structs. Labelling them as components (like above) will generate the appropriate code structure (marker interface, id property, etc).

Aside from custom components, Quantum comes with several pre-built ones:

  • Transform2D/Transform3D: position and rotation using Fixed Point (FP) values;
  • PhysicsCollider, PhysicsBody, PhysicsCallbacks, PhysicsJoints (2D/3D): used by Quantum's stateless physics engines;
  • PathFinderAgent, SteeringAgent, AvoidanceAgent, AvoidanceObstacle: navmesh-based path finding and movement.


Structs can be defined in both the DSL and C#.

DSL Defined

The Quantum DSL also allows the definition of regular structs (just like components, memory alignment, and helper functions will be taken care of):


struct ResourceItem
    FP Value;
    FP MaxValue;
    FP RegenRate;

The fields will be ordered in alphabetical order when the struct is generated. If you need / want to have them appear in a specific order, you will have to define the structs in C# (see section below).

This would let you use the "Resources" struct as a type in all other parts of the DSL, for example using it inside a component definition:


component Resources
    ResourceItem Health;
    ResourceItem Strength;
    ResourceItem Mana;

The generated struct is partial and can be extended in C# if so desired.

CSharp Defined

You can define structs in C# as well; however, in this case you will have to manually define:

  • The memory layout of the struct (using LayoutKind.Explicit)
  • Add a const int SIZE to the struct containing its size.
  • Implement the Serialize function.


public struct Foo {
  public const int SIZE = 12; // the size in bytes of all members in bytes.
  public int A;
  public int B;
  public int C;
  public static unsafe void Serialize(void* ptr, FrameSerializer serializer)
      var foo = (Foo*)ptr;

When using C# defined structs in the DSL (e.g. inside components), you will have to manually import the struct definition.

import struct Foo(12);

N.B.: The import does not support constants in the size; you will have to specify the exact numerical value each time.

Components vs. Structs

An important question is why and when should components be used instead of regular structs (components, in the end, are also structs).

Components contain generated meta-data that turns them into a special type with the following features:

  • Can be attached directly to entities;
  • Used to filter entities when traversing the game state (next chapter will dive into the filter API);

Components can accessed, used or passed as parameters as either pointers or as value types, just like any other struct.

Dynamic Collections

Quantum's custom allocator exposes blittable collections as part of the rollback-able game state. Collections only support support blittable types (i.e. primitive and DSL-defined types).

To manage collection, the Frame API offers 3 methods for each:

  • Frame.AllocateXXX: To allocate space for the collection on the heap.
  • Frame.FreeXXX: To free/deallocate the collection's memory.
  • Frame.ResolveXXX: To access the collection by resolving the pointer it.

Note: After freeing a collection, it HAS TO be nullified by setting it to default. This is required for serialization of the game state to work properly. Omitting the nullification will result in indeterministic behavior and desynchronisation. As alternative to freeing a collection and nullifying its Ptrs manually, it possible to use the FreeOnComponentRemoved attribute on the field in question.

Important Notes

  • Several components can reference the same collection instance.
  • Dynamic collections are stored as references inside components and structs. They therefore have to to be allocated when initializing them, and more importantly, freed when they are not needed any more. If the collection is part of a component, two options are available:
    • implement the reactive callbacks ISignalOnAdd<T> and ISignalOnRemove<T> and allocate/free the collections there. (For more information on these specific signals, see the Components page in the ECS section of the Manual); or,
    • use the [AllocateOnComponentAdded] and [FreeOnComponentRemoved] attributes to let Quantum handle the allocation and deallocation when the component is added and removed respectively.
  • Quantum do NOT pre-allocate collections from prototypes, unless there is at least value. If the collection is empty, the memory has to be manually allocated.
  • Attempting to free a collection more than once will throw an error and puts the heap in an invalid state internally.


Dynamic lists can be defined in the DSL using list<T> MyList.

component Targets {
  list<EntityRef> Enemies;

The basic API methods for dealing with these Lists are:

  • Frame.AllocateList<T>()
  • Frame.FreeList(QListPtr<T> ptr)
  • Frame.ResolveList(QListPtr<T> ptr)

Once resolved, a list can be iterated over or manipulated with all the expected API methods of a list such as Add, Remove, Contains, IndexOf, RemoveAt, [], etc... .

To use the list in the component of type Targets defined in the code snippet above, you could create the following system:


namespace Quantum
  public unsafe class HandleTargets : SystemMainThread, ISignalOnComponentAdded<Targets>, ISignalOnComponentRemoved<Targets>
        public override void Update(Frame f) 
      foreach (var (entity, component) in f.GetComponentIterator<Targets>()) { 
        // To use a list, you must first resolve its pointer via the frame
        var list = f.ResolveList(component.Enemies);

        // Do stuff

        public void OnAdded(Frame f, EntityRef entity, Targets* component)
            // allocating a new List (returns the blittable reference type - QListPtr)
        component->Enemies = f.AllocateList<EntityRef>();
        public void OnRemoved(Frame f, EntityRef entity, Targets* component)
            // A component HAS TO de-allocate all collection it owns from the frame data, otherwise it will lead to a memory leak.
            // receives the list QListPtr reference.
            // All dynamic collections a component points to HAVE TO be nullified in a component's OnRemoved
            // EVEN IF is only referencing an external one!
            // This is to prevent serialization issues that otherwise lead to a desynchronisation.
            component->Enemies = default;


Dictionaries can be declared in the DSL like so dictionary<key, value> MyDictionary.


component Hazard{
    dictionary<EntityRef, Int32> DamageDealt;

The basic API methods for dealing with these dictionaries are:

  • Frame.AllocateDictionary<K,V>()
  • Frame.FreeDictionary(QDictionaryPtr<K,V> ptr)
  • Frame.ResolveDictionary(QDictionaryPtr<K,V> ptr)

Just like with any other dynamic collection it is mandatory to allocate it before using it, as well as de-allocate it from the frame data and nullified it once the dictionary is no longer used. See the example provided in the section about lists here above.


HashSets can be declared in the DSL like so hash_set<T> MyHashSet.


component Nodes{
    hash_set<FP> ProcessedNodes;

The basic API methods for dealing with these dictionaries are:

  • Frame.AllocateHashSet(QHashSetPtr<T> ptr, int capacity = 8)
  • Frame.FreeHashSet(QHashSetPtr<T> ptr)
  • Frame.ResolveHashSet(QHashSetPtr<T> ptr)

Just like with any other dynamic collection it is mandatory to allocate it before using it, as well as de-allocate it from the frame data and nullified it once the hash set is no longer used. See the example provided in the section about lists here above.

Enums, Unions and Bitsets


Enums can be used to define a set of named contant values. Use the "enum" keyword to define its name and data. This example defines a simple enum, EDamageType, and how to use it as field in a struct:


enum EDamageType {
    None, Physical, Magic

struct StatsEffect {
    EDamageType DamageType;

Enums are treated as integer constants on CodeGen by default, starting from 0. However, it is possible to explicitly assign integer values to Enum members if needed. Also, in scenarios where memory usage is a concern, it is possible to reduce the memory footpring of enum values by using a diffrent type as the underlying type, such as byte. Here's an example:


enum EModifierOperation : Byte
    None = 0,
    Add = 1,
    Subtract = 2

The flags keyword is used with Enum types to indicate that each value in the Enum represents a distinct bit flag. This allows you to combine multiple Enum values using bitwise operations, it is a way to represent sets of related options or states.


flags ETeamStatus : Byte

Using the flags keyword also code-generates utility methods for the enum type which are more performant than regular System.Enum methods such as IsFlagSet(), which is more performance than System.Enum.HasFlag() as it avoids the need of value type boxing.


C-like unions can be generated as well. The union type overlaps in memory the data layout of all the involved structs. Here is an example of how to declare a union in the DSL:


struct DataA
    FPVector2 Foo;

struct DataB
    FP Bar;

union Data
    DataA A;
    DataB B;

Unions can be declared as part of a component:


component ComponentWithUnion {
    Data ComponentData;

Internally, the union type Data contains the logic necessary for switching between the union types as they are accessed. Here are a few usage examples:


private void UseWarriorAttack(Frame f)
    var character = f.Unsafe.GetPointer<Character>(entity);
    character->Data.Warrior->ImpulseDirection = FPVector3.Forward;

private void ResetSpellcasterMana(Frame f)
    var character = f.Unsafe.GetPointer<Character>(entity);
    character->Data.Spellcaster->Mana = FP._10;

When using unions, it is possible to use only one of the multiple internal types it contains, or it can be switched dynamically, in runtime, by simply accessing specific internal types. In the code snippets above, accessing the Warrior and Spellcaster pointers already changed the union type internally.

It is also possible to check what is the currently used union type by using the Field property and some of code-gen'd constants:


private bool IsWarrior(CharacterData data)
    return data.Field == CharacterData.WARRIOR;


Bitsets can be used to declared fixed-size memory blocks for any desired purpose (for example fog-of-war, grid-like structures for pixel perfect game mechanics, etc.):


struct FOWData
    bitset[256] Map;


In Quantum, the runtime input exchanged between clients is also declared in the DSL. This example defines a simple movement vector and a Fire button as input for a game:


    FPVector2 Movement;
    button Fire;

The input struct is polled every tick and sent to the server (when playing online).

For more information about input, such as best practices and recommended approaches to optimization, refer to this page: Input


Signals are function signatures used as a decoupled inter-system communication API (a form of publisher/subscriber API). This would define a simple signal (notice the special type entity_ref - these will be listed at the end of this chapter):


signal OnDamage(FP damage, entity_ref entity);

This would generate the following interface (that can be implemented by any System):


public interface ISignalOnDamage
    public void OnDamage(Frame f, FP damage, EntityRef entity);

Signals are the only concept which allows the direct declaration of a pointer in Quantum's DSL, so passing data by reference can be used to modify the original data directly in their concrete implementations:


signal OnBeforeDamage(FP damage, Resources* resources);

Notice this allows the passing of a component pointer (instead of the entity reference type).


Events are a fine-grained solution to communicate what happens inside the simulation to the rendering engine / view (they should never be used to modify/update part of the game state). Use the "event" keyword to define its name and data:

Find detailed information about events in the Frame Events Manual.

Define an event using the Quantum DSL

event MyEvent{
    int Foo;

Trigger the event from the simulation



And subscribe and consume the event in Unity


QuantumEvent.Subscribe(listener: this, handler: (MyEvent e) => Debug.Log($"MyEvent {e.Foo}"));


It is possible to define globally accessible variables in the DSL. Globals can be declared in any .qtn file by using the global scope.


global {
    // Any type that is valid in the DSL can also be used.
    FP MyGlobalValue;

Like all things DSL-defined, global variables are part of the state and are fully compatible with the predict-rollback system.

Variables declared in the global scope are made available through the Frame API. They can be accessed (read/write) from any place that has access to the frame - see the Systems document in the ECS section.

N.B.: An alternative to global variables are the Singleton Components; for more information please refer to the Components page in the ECS section of the manual.

Special Types

Quantum has a few special types that are used to either abstract complex concepts (entity reference, player indexes, etc.), or to protect against common mistakes with unmanaged code, or both. The following special types are available to be used inside other data types (including in components, also in events, signals, etc.):

  • player_ref: represents a runtime player index (also cast to and from Int32). When defined in a component, can be used to store which player controls the associated entity (combined with Quantum's player-index-based input).
  • entity_ref: because each frame/tick data in quantum resides on a separate memory region/block (Quantum keeps a a few copies to support rollbacks), pointers cannot be cached in-between frames (nor in the game state neither in Unity scripts). An entity ref abstracts an entity's index and version properties (protecting the developer from accidentally accessing deprecated data over destroyed or reused entity slots with old refs).
  • asset_ref<AssetType>: rollback-able reference to a data asset instance from the Quantum asset database (please refer to the data assets chapter).
  • list<T>, dictionary<K,T>: dynamic collection references (stored in Quantum's frame heap). Only supports blittable types (primitives + DSL-defined types).
  • array<Type>[size]: fixed sized "arrays" to represent data collections. A normal C# array would be a heap-allocated object reference (it has properties, etc.), which violates Quantum's memory requirements, so the special array type generates a pointer based simple API to keep rollback-able data collections inside the game state;

A Note On Assets

Assets are a special feature of Quantum that let the developer define data-driven containers (normal classes, with inheritance, polymorphic methods, etc.) that end up as immutable instances inside an indexed database. The "asset" keyword is used to assign an (existing) class as a data asset that can have references assigned inside the game state (please refer to the Data Assets chapter to learn more about features and restrictions):


asset CharacterData; // the CharacterData class is partially defined in a normal C# file by the developer

The following struct show some valid examples of the types above (sometimes referencing previously defined types):


struct SpecialData
    player_ref Player;
    entity_ref Character;
    entity_ref AnotherEntity;
    asset_ref<CharacterData> CharacterData;
    array<FP>[10] TenNumbers;

Available Types

When working in the DSL, you can use a variety of types. Some are pre-imported by the parsers, while others need to be manually imported.

By default

Quantum's DSL parser has a list of pre-imported cross-platform deterministic types that can be used in the game state definition:

  • Boolean / bool - internally gets wrapped in QBoolean which works identically (get/set, compare, etc...)
  • Byte
  • SByte
  • UInt16 / Int16
  • UInt32 / Int32
  • UInt64 / Int64
  • FP
  • FPVector2
  • FPVector3
  • FPMatrix
  • FPQuaternion
  • PlayerRef / player_ref in the DSL
  • EntityRef / entity_ref in the DSL
  • LayerMask
  • NullableFP / FP? in the DSL
  • NullableFPVector2 / FPVector2? in the DSL
  • NullableFPVector3 / FPVector3? in the DSL
  • QString is for UTF-16 (aka Unicode in .NET)
  • QStringUtf8 is always UTF-8
  • Hit
  • Hit3D
  • Shape2D
  • Shape3D
  • Joint, DistanceJoint, SpringJoint and HingeJoint

Note on QStrings: N represents the total size of the string in bytes minus 2 bytes used for bookkeeping. In other words QString<64> will use 64 bytes for a string with a max byte length of 62 bytes, i.e. up to 31 UTF-16 characters.

Manual Import

If you need a type that is not listed in the previous section, you will have to import it manually when using it in QTN files.

Namespaces / Types outside of Quantum

To import types defined in other namespaces, you can use the following syntax:


import MyInterface;
import MyNameSpace.Utils;

For an enum the syntax is as follows:


import enum MyEnum(underlying_type);

// This syntax is identical for Quantum specific enums
import enum Shape3DType(byte);

Built-In Quantum Type and Custom Type

When importing a Quantum built-in type or a custom type, the struct size is predefined in their C# declaration. It is therefore important to add some safety measures.


namespace Quantum {
  public struct Foo {
    public const int SIZE = sizeof(Int32) * 2;
    public Int32 A;
    public Int32 B;


#define FOO_SIZE 8 // Define a constant value with the known size of the struct
import struct Foo(8);

To ensure the expected size of the struct is equal to the actual size, it is recommended to add an Assert as shown below in one of your systems.


public unsafe class MyStructSizeCheckingSystem : SystemMainThread{    
    public override void OnInit(Frame f)
        Assert.Check(Constants.FOO_SIZE == Foo.SIZE);

If the size of the built-in struct changes during an upgrade, this Assert will throw and allow you to update the values in the DSL.


Quantum supports several attributes to present parameters in the Inspector.

The attributes are contained within the Quantum.Inspector namespace.

Attribute Parameters Description
DrawIf string fieldName
long value
CompareOperator compare
HideType hide
Displays the property only if the condition evaluates to true.

fieldName = the name of the property to evaluate.
value = the value used for comparison.
compare = the comparison operation to be performed Equal, NotEqual, Less, LessOrEqual, GreaterOrEqual or Greater.
hide = the field's behavior when the expression evaluates to False:Hide or ReadOnly.

For more information on compare and hide, see below.
Header string header Adds a header above the property.

header = the header text to display.
HideInInspector - Serializes the field and hides the following property in the Unity inspector.
Layer - Can only be applied to type int.
Will call EditorGUI.LayerField on the field.
Optional string enabledPropertyPath Allows to turn the display of a property on/off.

enabledPropertyPath = the path to the bool used to evaluate the toggle.
Space - Adds a space above the property
Tooltip string tooltip Displays a tool tip when hovering over the property.

tooltip = the tip to display.
int length
int minLength
int maxLength
Using length allows to define the size of a an array.
Using minLength and maxLength allows to define a range for the size in the Inspector.
The final size can then be set in the Inspector.
(minLength and maxLength are inclusive)
ExcludeFromPrototype - Can be applied to both a component and component fields.
- Field: Excludes field from a the prototype generated for the component.
- Component: No prototype will be generated for this component.
PreserveInPrototype - Added to a type marks it as usable in prototypes and prevents prototype class from being emit.
Added to a field only affects a specific field. Useful for simple [Serializable] structs as it avoids having to use _Prototype types on Unity side.
AllocateOnComponentAdded - Can be applied to dynamic collections.
This will allocate memory for the collection if it has not already been allocated when the component holding the collection is added to an entity.
FreeOnComponentRemoved - Can be applied to dynamic collections and Ptrs.
This will deallocate the associated memory and nullify the Ptr held in the field when the component is removed.
IMPORTANT: Do NOT use this attribute in combination with cross-referenced collections as it only nullifies the Ptr held in that particular field and the others will be pointing to invalid memory.

The Attributes can be used in both C# and qtn files unless otherwise specified; however, there are some syntactic differences.

Use in CSharp

In C# files, attributes can be used and concatenated like any other attribute.


// Multiple single attributes
[Header("Example Array")][Tooltip("min = 1\nmax = 20")] public FP[] TestArray = new FP[20];

// Multiple concatenated attributes
[Header("Example Array"), Tooltip("min = 1\nmax = 20")] public FP[] TestArray = new FP[20];

Use in qtn

In qtn files, the usage of single attributes remains the same as in C#.


[Header("Example Array")] array<FP>[20] TestArray;

When combining multiple attributes, they have to be concatenated.


[Header("Example Array"), Tooltip("min = 1\nmax = 20")] array<FP>[20] TestArray;

Compiler Options

The following compiler options are currently available to be used inside Quantum's DSL files (more will be added in the future):


// pre defining max number of players (default is 6, absolute max is 64)
#pragma max_players 16

// increase the component count from 256 to 512
#pragma max_components 512

// numeric constants (useable inside the DSL by MY_NUMBER and useable in code by Constants.MY_NUMBER)
#define MY_NUMBER 10

// overriding the base class name for the generated constants (default is "Constants")
#pragma constants_class_name MyFancyConstants

Custom FP Constants

You can also define custom FP constants inside Quantum's DSL files. ex:


// in a DSL file
#define Pi 3.14

Then, Quantum codegen will generate the corresponding constant in the FP struct:


// 3.14
FP constant = Constants.Pi;

It will also generate the corresponding raw value as well:


// 3.14 Raw
var rawConstant = Constants.Raw.Pi;
Back to top