Entities and Components

Introduction

Entities and components are the main building blocks of any ECS application. This manual describes the behavior of entities and components in Flecs.

Entities

Entities are uniquely identifiable objects in a game or simulation. In a real time strategy game, there may be entities for the different units, buildings, UI elements and particle effects, but also for example the camera, world and player.

By itself, an entity is just a unique identifier that does not contain any state. You cannot tell from the entity's type what kind of entity it is, as the type is just "entity".

In Flecs, an entity is represented by a 64 bit integer, which is also how it is exposed in the C API:

typedef uint64_t ecs_entity_t;

The first 32 bits of the entity are the "identifying" part: it is the number that uniquely identifies an object in the game. The remaining 32 bits are used by Flecs for liveliness tracking and various other purposes. This article has more details on the exact bit representation of entities.

Zero ("`0`") is the value that is reserved for invalid ids. Functions that return an entity id may return 0 to indicate that the function failed. 0 can also be used to indicate the absence of an entity, for example when a lookup by name didn't find anything.

Creation

The following example shows how to create a new entity without any components:

• C

  ecs_entity_t my_entity = ecs_new(world);

• C++

  flecs::entity my_entity = world.entity();

• C#

  Entity myEntity world.Entity();

• Rust

  let my_entity = world.entity();

Empty entities don't have any components, and are therefore not matched by any queries or systems. They won't show up in the tree view of the explorer, as it is query based.

The id of the first returned entity is not 1, but likely a much higher number like 500. The reason for this is that Flecs creates a bunch of builtin entities, and reserves some of the id space for components (more on that later).

Deletion

The following examples show how to delete entities:

• C

  ecs_delete(world, my_entity);

• C++

  my_entity.destruct();

• C#

  myEntity.Destruct();

• Rust

  my_entity.destruct();

After an entity is deleted, it can no longer be used with most ECS operations. The deleted entity will be made available for reuse, so that the next time a new entity is created the deleted id can be recycled.

Whenever an id is recycled, Flecs increases a "version" counter in the upper 32 bits of the entity identifier. This can look strange, as it makes recycled entity ids large, often larger than 4 billion!. It is however 100% expected behavior, and happens as soon after the first entity is deleted.

The reason this happens is so that the Flecs API can tell whether an entity id is alive or not. Consider the following example, where "v" denotes the version part of the entity id:

• C

  ecs_entity_t e1 = ecs_new(world); // Returns 500v0
  ecs_delete(world, e1); // Recycle 500
   
  ecs_entity_t e2 = ecs_new(world); // Returns 500v1
   
  ecs_add(world, e1, Npc); // Fails, 500v0 is not alive
  ecs_add(world, e2, Npc); // OK, 500v1 is alive

• C++

  flecs::entity e1 = world.entity(); // Returns 500v0
  e1.destruct(); // Recycles 500
   
  flecs::entity e2 = world.entity(); // Returns 500v1
   
  e1.add<Npc>(); // Fails, 500v0 is not alive
  e2.add<Npc>(); // OK, 500v1 is alive

• C#

  Entity e1 = world.Entity(); // Returns 500v0
  e1.Destruct(); // Recycles 500
   
  Entity e2 = world.Entity(); // Returns 500v1
   
  e1.Add<Npc>(); // Fails, 500v0 is not alive
  e2.Add<Npc>(); // OK, 500v1 is alive

• Rust

  let e1 = world.entity(); // Returns 500v0
  e1.destruct(); // Recycles 500
  let e2 = world.entity(); // Returns 500v1
  // Fails, 500v0 is not alive
  e1.add::<Npc>();
  // OK, 500v1 is alive
  e2.add::<Npc>();

It is valid to delete an already deleted entity:

• C

  ecs_entity_t e1 = ecs_new(world);
  ecs_delete(world, e1);
  ecs_delete(world, e1); // OK: post condition is satisfied

• C++

  flecs::entity e1 = world.entity();
  e1.destruct();
  e1.destruct(); // OK: post condition is satisfied

• C#

  Entity e1 = world.Entity();
  e1.Destruct();
  e1.Destruct(); // OK: post condition is satisfied

• Rust

  let e1 = world.entity();
  e1.destruct();
  e1.destruct(); // OK: post condition is satisfied

Clearing

An entity can be cleared, which means all components are removed without making the entity not alive. Clearing an entity is more efficient than removing components one by one. An example:

• C

  ecs_clear(world, my_entity);

• C++

  my_entity.clear();

• C#

  myEntity.Clear();

• Rust

  my_entity.clear();

Liveliness Checking

Applications can use the is_alive operation to test if an entity is alive or not. The entity passed to is_alive must be a valid entity, passing 0 will throw an error. An example:

• C

  ecs_entity_t e1 = ecs_new(world);
  ecs_entity_t e2 = ecs_new(world);
  ecs_delete(world, e1);
   
  ecs_is_alive(world, e1); // False
  ecs_is_alive(world, e2); // True

• C++

  flecs::entity e1 = world.entity();
  flecs::entity e2 = world.entity();
  e1.destruct();
   
  e1.is_alive(); // False
  e2.is_alive(); // True

• C#

  Entity e1 = world.Entity();
  Entity e2 = world.Entity();
  e1.Destruct();
   
  e1.IsAlive(); // False
  e2.IsAlive(); // True

• Rust

  let e1 = world.entity();
  let e2 = world.entity();
  e1.destruct();
  e1.is_alive(); // False
  e2.is_alive(); // True

The API also provides an is_valid operation. Whereas is_alive tests if an entity is alive, is_valid tests if the entity can be used with operations, and may be used with invalid (0) entity ids. An example:

• C

  ecs_entity_t e1 = ecs_new(world);
  ecs_entity_t e2 = ecs_new(world);
  ecs_delete(world, e1);
   
  ecs_is_valid(world, e1); // False
  ecs_is_valid(world, e2); // True
  ecs_is_valid(world, 0); // False

• C++

  flecs::entity e1 = world.entity();
  flecs::entity e2 = world.entity();
  e1.destruct();
   
  e1.is_valid(); // False
  world.entity(0).is_valid(); // False

• C#

  Entity e1 = world.Entity();
  Entity e2 = world.Entity();
  e1.Destruct();
   
  e1.IsValid(); // False
  e2.IsValid(); // True
  world.Entity(0).IsValid(); // False

• Rust

  let e1 = world.entity();
  let e2 = world.entity();
  e1.destruct();
  e1.is_valid(); // False
  world.entity_from_id(0).is_valid(); // False

Manual Ids

Sometimes it can be useful to have direct control over which id an entity assumes. One example of this is to make sure that both sides of a networked application use the same entity id. Applications are allowed to do this, as long as the entity id is not yet in use, and it is made alive. The following example shows how to do this:

• C

  ecs_make_alive(world, 1000); // Make id 1000 alive

• C++

  flecs::entity e = world.make_alive(1000);

• C#

  Entity e = world.MakeAlive(1000);

• Rust

  let e = world.make_alive(1000);

The make_alive operation guarantees that the provided entity id will be alive after the operation finishes. The operation will fail if the provided entity id is already in use with a different version.

Manual Versioning

Applications can manually override the version of an entity with the set_version operation. This can be useful when for example synchronizing networked ids. An example:

• C

  ecs_set_version(world, versioned_id);

• C++

  world.set_version(versioned_id);

• C#

  world.SetVersion(versionedId);

• Rust

  //TODO does not exist yet
  //world.set_version(versioned_id);

Ranges

An application can instruct Flecs to issue ids from a specific offset and up to a certain limit with the ecs_set_entity_range operation. This example ensures that id generation starts from id 5000:

• C

  ecs_set_entity_range(world, 5000, 0);

• C++

  world.set_entity_range(5000, 0);

• C#

  world.SetEntityRange(5000, 0);

• Rust

  world.set_entity_range(5000, 0);

If the last issued id was higher than 5000, the operation will not cause the last id to be reset to 5000. An application can also specify the highest id that can be generated:

• C

  ecs_set_entity_range(world, 5000, 10000);

• C++

  world.set_entity_range(5000, 10000);

• C#

  world.SetEntityRange(5000, 10000);

• Rust

  world.set_entity_range(5000, 10000);

If invoking ecs_new would result in an id higher than 10000, the application would assert. If 0 is provided for the maximum id, no upper bound will be enforced.

Note that at the moment setting the range does not affect recycled ids. It is therefore possible that ecs_new returns an id outside of the specified range if a recycle-able id is available. This is an issue that will be addressed in future versions.

It is possible for an application to enforce that entity operations are only allowed for the configured range with the ecs_enable_range_check operation:

• C

  ecs_enable_range_check(world, true);

• C++

  world.enable_range_check();

• C#

  world.EnableRangeCheck(true);

• Rust

  world.enable_range_check(true);

This can be useful for enforcing simple ownership behavior, where different id ranges are mapped out for different owners (often game clients or servers).

Names

Entity can be given names, which allows them to be looked up on the world. An example:

• C

  ecs_entity_t e = ecs_entity(world, { .name = "MyEntity" });
   
  if (e == ecs_lookup(world, "MyEntity")) {
      // true
  }
   
  printf("%s\n", ecs_get_name(world, e));

• C++

  flecs::entity e = world.entity("MyEntity");
   
  if (e == world.lookup("MyEntity")) {
      // true
  }
   
  std::cout << e.name() << std::endl;

• C#

  Entity e = world.Entity("MyEntity");
   
  if (e == world.Lookup("MyEntity")) {
      // true
  }
   
  Console.WriteLine(e.Name());

• Rust

  let e = world.entity_named("MyEntity");
  if e == world.lookup("MyEntity") {
      // true
  }
  println!("{}", e.name());

Names are namespaced. They are similar to namespaced types in a programming language, or to files on a file system. If an entity is a child of a parent, a lookup will have to include the name of the parent:

• C

  ecs_entity_t p = ecs_entity(world, { .name = "Parent" });
  ecs_entity_t e = ecs_entity(world, { .name = "Child", .parent = p });
   
  if (e == ecs_lookup(world, "Parent.Child")) {
      // true
  }

• C++

  flecs::entity p = world.entity("Parent");
  flecs::entity e = world.entity("Child").child_of(p);
   
  if (e == world.lookup("Parent::Child")) {
      // true
  }

• C#

  Entity p = world.Entity("Parent");
  Entity e = world.Entity("Child").ChildOf(p);
   
  if (e == world.Lookup("Parent.Child")) {
      // true
  }

• Rust

  let p = world.entity_named("Parent");
  let e = world.entity_named("Child").child_of_id(p);
  if e == world.lookup("Parent::Child") {
      // true
  }

Lookups can be relative to an entity. The following example shows how to lookup an entity relative to a parent:

• C

  ecs_entity_t p = ecs_entity(world, { .name = "Parent" });
  ecs_entity_t e = ecs_entity(world, { .name = "Child", .parent = p });
   
  if (e == ecs_lookup_from(world, p, "Child")) {
      // true
  }

• C++

  flecs::entity p = world.entity("Parent");
  flecs::entity e = world.entity("Child").child_of(p);
   
  if (e == p.lookup("Child")) {
      // true
  }

• C#

  Entity p = world.Entity("Parent");
  Entity e = world.Entity("Child").ChildOf(p);
   
  if (e == p.Lookup("Child")) {
      // true
  }

• Rust

  let p = world.entity_named("Parent");
  let e = world.entity_named("Child").child_of_id(p);
  if e == p.lookup("Child") {
      // true
  }

An application can request the name and path of an entity. The name is the direct name given to an entity, without the names of the parent. The path is the name and names of the entity's parent, separated by a separator. If an entity has no parents, the name and the path are the same. An example:

• C

  ecs_entity_t p = ecs_entity(world, { .name = "Parent" });
  ecs_entity_t e = ecs_entity(world, { .name = "Child", .parent = p });
   
  // Returns name, result does not need to be freed
  const char *name = ecs_get_name(world, e);
  printf("%s\n", name); // Child
   
  // Returns path, result must be freed
  char *path = ecs_get_path(world, e);
  printf("%s\n", path) // Parent.Child
  ecs_os_free(path);

• C++

  flecs::entity p = world.entity("Parent");
  flecs::entity e = world.entity("Child").child_of(p);
   
  // Returns entity name, does not allocate
  std::cout << e.name() << std::endl; // Child
   
  // Returns entity path, does allocate
  std::cout << e.path() << std::endl; // Parent.Child

• C#

  Entity p = world.Entity("Parent");
  Entity e = world.Entity("Child").ChildOf(p);
   
  // Returns entity name
  Console.WriteLine(e.Name()); // Child
   
  // Returns entity path
  Console.WriteLine(e.Path()); // Parent.Child

• Rust

  let p = world.entity_named("Parent");
  let e = world.entity_named("Child").child_of_id(p);
  // Returns entity name, does not allocate
  println!("{}", e.name()); // Child
  // Returns entity path, does allocate
  println!("{}", e.path().unwrap()); // Parent.Child

Names must be unique. There can only be one entity with a specific name in the scope of a parent. Entities can be annotated with a doc name, which does not need to be unique. See the doc addon for more details.

When the name for an existing entity is used during the creation of a named entity, the existing entity is returned. An example:

• C

  ecs_entity_t e1 = ecs_entity(world, { .name = "Parent.Child" });
  ecs_entity_t e2 = ecs_entity(world, { .name = "Parent.Child" });
   
  if (e1 == e2) {
      // true
  }

• C++

  flecs::entity e1 = world.entity("Parent::Child");
  flecs::entity e2 = world.entity("Parent::Child");
   
  if (e1 == e2) {
      // true
  }

• C#

  Entity e1 = world.Entity("Parent.Child");
  Entity e2 = world.Entity("Parent.Child");
   
  if (e1 == e2) {
      // true
  }

• Rust

  let e1 = world.entity_named("Parent::Child");
  let e2 = world.entity_named("Parent::Child");
  if e1 == e2 {
      // true
  }

Entity names can be changed after the entity is created, as long as the specified name is unique (e.g. there isn't a sibling entity with the same name). An example:

• C

  ecs_entity_t e = ecs_entity(world, { .name = "Foo" });
   
  // Change name
  ecs_set_name(world, e, "Bar");

• C++

  flecs::entity e = world.entity("Foo");
   
  // Change name
  e.set_name("Bar");

• C#

  Entity e = world.Entity("Foo");
   
  e.SetName("Bar");

• Rust

  let e = world.entity_named("Foo");
  // Change name
  e.set_name("Bar");

Entity names can be plain numbers:

• C

  ecs_entity_t ten = ecs_entity(world, { .name = "10" });
  ecs_entity_t twenty = ecs_entity(world, { .name = "20" });

• C++

  flecs::entity ten = world.entity("10");
  flecs::entity twenty = world.entity("20");

• C#

  Entity ten = world.Entity("10");
  Entity twenty = world.Entity("20");

• Rust

  let ten = world.entity_named("10");
  let twenty = world.entity_named("20");

Entity names can be used to refer directly to an entity id by prefixing them with a #. For example, looking up #1 will return the entity with id 1. This allows applications that work with names to treat anonymous entities the same as named entities.

Disabling

Entities can be disabled which prevents them from being matched with queries. This can be used to temporarily turn off a feature in a scene during game play. It is also used for Flecs systems, which can be disabled to temporarily remove them from a schedule. The following example shows how to disable and reenable an entity:

• C

  ecs_entity_t e = ecs_new(world);
   
  // Enable entity
  ecs_enable(world, e, true);
   
  // Disable entity
  ecs_enable(world, e, false);

• C++

  flecs::entity e = world.entity();
   
  // Enable entity
  e.enable();
   
  // Disable entity
  e.disable();

• C#

  Entity e = world.Entity();
   
  // Enable entity
  e.Enable();
   
  // Disable entity
  e.Disable();

• Rust

  let e = world.entity();
  // Enable entity
  e.enable_self();
  // Disable entity
  e.disable_self();

Entity disabling can be combined with prefabs to create lists of entities that can be disabled with a single operation. The following example shows how to disable three entities with a single operation:

• C

  // Three entities to disable
  ecs_entity_t e1 = ecs_new(world);
  ecs_entity_t e2 = ecs_new(world);
  ecs_entity_t e3 = ecs_new(world);
   
  // Create prefab that has the three entities
  ecs_entity_t p = ecs_new_w_id(world, EcsPrefab);
  ecs_add_id(world, p, e1);
  ecs_add_id(world, p, e2);
  ecs_add_id(world, p, e3);
   
  // Disable entities
  ecs_enable(world, p, false);
   
  // Enable entities
  ecs_enable(world, p, true);

• C++

  // Three entities to disable
  flecs::entity e1 = world.entity();
  flecs::entity e2 = world.entity();
  flecs::entity e3 = world.entity();
   
  // Create prefab that has the three entities
  flecs::entity p = world.prefab();
  p.add(e1);
  p.add(e2);
  p.add(e3);
   
  // Disable entities
  p.disable();
   
  // Enable entities
  p.enable();

• C#

  // Three entities to disable
  Entity e1 = world.Entity();
  Entity e2 = world.Entity();
  Entity e3 = world.Entity();
   
  // Create prefab that has the three entities
  Entity p = world.Prefab();
  p.Add(e1);
  p.Add(e2);
  p.Add(e3);
   
  // Disable entities
  p.Disable();
   
  // Enable entities
  p.Enable();

• Rust

  // Three entities to disable
  let e1 = world.entity();
  let e2 = world.entity();
  let e3 = world.entity();
   
  // Create prefab that has the three entities
  let p = world.prefab();
  p.add_id(e1);
  p.add_id(e2);
  p.add_id(e3);
   
  // Disable entities
  p.disable_self();
   
  // Enable entities
  p.enable_self();

This also works with prefab hierarchies, as shown in the following example:

• C

  // Three entities to disable
  ecs_entity_t e1 = ecs_new(world);
  ecs_entity_t e2 = ecs_new(world);
  ecs_entity_t e3 = ecs_new(world);
   
  // Create prefab hierarchy with the three entities
  ecs_entity_t p1 = ecs_new_w_id(world, EcsPrefab);
  ecs_add_id(world, p1, e1);
   
  ecs_entity_t p2 = ecs_new_w_id(world, EcsPrefab);
  ecs_add_pair(world, p2, EcsIsA, p1);
  ecs_add_id(world, p2, e2);
  ecs_add_id(world, p2, e3);
   
  // Disable e1, e2, e3
  ecs_enable(world, p2, false);
   
  // Enable e1
  ecs_enable(world, p1, true);

• C++

  // Three entities to disable
  flecs::entity e1 = world.entity();
  flecs::entity e2 = world.entity();
  flecs::entity e3 = world.entity();
   
  // Create prefab hierarchy with the three entities
  flecs::entity p1 = world.prefab()
      .add(e1);
   
  flecs::entity p2 = world.prefab()
      .is_a(p1)
      .add(e2)
      .add(e3);
   
  // Disable e1, e2, e3
  p2.disable();
   
  // Enable e1
  p1.enable();

• C#

  // Three entities to disable
  Entity e1 = world.Entity();
  Entity e2 = world.Entity();
  Entity e3 = world.Entity();
   
  // Create prefab hierarchy with the three entities
  Entity p1 = world.Prefab()
      .Add(e1);
   
  Entity p2 = world.Prefab()
      .IsA(p1)
      .Add(e2)
      .Add(e3);
   
  // Disable e1, e2, e3
  p2.Disable();
   
  // Enable e1
  p1.Enable();

• Rust

  // Three entities to disable
  let e1 = world.entity();
  let e2 = world.entity();
  let e3 = world.entity();
   
  // Create prefab hierarchy with the three entities
  let p1 = world.prefab().add_id(e1);
  let p2 = world.prefab().is_a_id(p1).add_id(e2).add_id(e3);
   
  // Disable e1, e2, e3
  p2.disable_self();
   
  // Enable e1
  p1.enable_self();

Entity disabling works by adding a Disabled tag to the entity, which can also be added manually with regular add/remove operations. An example:

• C

  ecs_add_id(world, e, EcsDisabled);

• C++

  e.add(flecs::Disabled);

• C#

  e.Add(Ecs.Disabled);

• Rust

  e.add::<flecs::Disabled>();

Components

A component is something that is added to an entity. Components can simply tag an entity ("this entity is an `Npc`"), attach data to an entity ("this entity is at `Position` `{10, 20}`") and create relationships between entities ("bob `Likes` alice") that may also contain data ("bob `Eats` `{10}` apples").

To disambiguate between the different kinds of components in Flecs, they are named separately in Flecs:

Name	Has Data	Is Pair
Tag	No	No
Component	Yes	No
Relationship	No	Yes
Relationship component	Yes	Yes

Here, "has data" indicates whether a component attaches data to the entity. This means that in addition to asking whether an entity has a component, we can also ask what the value of a component is.

A pair is a component that's composed out of two elements, such as "Likes, alice" or "Eats, apples". See the Relationship manual for more details.

Operations

The following table provides the base set of operations that Flecs offers for components:

Operation	Description
add	Adds component to entity. If entity already has the component, add does nothing. Requires that the component is default constructible.
remove	Removes component from entity. If entity doesn't have the component, remove does nothing.
get	Returns a immutable reference to the component. If the entity doesn't have the component, get returns nullptr.
get_mut	Returns a mutable reference to the component. If the entity doesn't have the component, get_mut returns nullptr.
ensure	Returns a mutable reference to the component. ensure behaves as a combination of add and get_mut.
emplace	Returns a mutable reference to the component. If the entity doesn't have the component, emplace returns a reference to unconstructed memory. This enables adding components that are not default constructible.
modified	Emits a modified event for a component. This ensures that OnSet observers and on_set hooks are invoked, and updates change detection administration.
set	Sets the value of a component. set behaves as a combination of ensure and modified. set does not take ownership of the provided value.

The following component lifecycle diagram shows how the different operations mutate the storage and cause hooks and observers to be invoked:

Component Lifecycle

Components are Entities

In an ECS framework, components need to be uniquely identified. In Flecs this is done by making each component is its own unique entity. If an application has a component Position and Velocity, there will be two entities, one for each component. Component entities can be distinguished from "regular" entities as they have a Component component. An example:

• C

  // Register component (see below)
  ECS_COMPONENT(world, Position);
   
  // The ecs_id() macro can be used in C to obtain the component id from a type
  ecs_entity_t pos = ecs_id(Position);
   
  // Component entities have the Component component
  const EcsComponent *comp_data = ecs_get(world, pos, EcsComponent);
   
  printf("{size: %d, alignment: %d}\n", 
      comp_data->size, 
      comp_data->alignment);

• C++

  // Get the entity for the Position component
  flecs::entity pos = world.component<Position>();
   
  // Component entities have the Component component
  const flecs::Component *comp_data = pos.get<flecs::Component>();
   
  std::cout << "{size: " << comp_data->size << ", " 
            << comp_data->alignment << "}\n";

• C#

  // Get the entity for the Position component
  Entity pos = world.Component<Position>();
   
  // Component entities have the Component component
  ref readonly flecs.EcsComponent compData = ref e.Get<flecs.EcsComponent>();
   
  Console.WriteLine($"Size: {compData.size}, Alignment: {compData.alignment}");

• Rust

  // Get the entity for the Position component
  let pos = world.component::<Position>();
  // Component entities have the Component component
  pos.get::<&flecs::Component>(|comp_data| {
      println!(
          "size: {}, alignment: {}",
          comp_data.size, comp_data.alignment
      );
  });

All of the APIs that apply to regular entities also apply to component entities. Many Flecs features leverage this. It is for example possible to customize component behavior by adding tags to components. The following example shows how a component can be customized to use sparse storage by adding a Sparse tag to the component entity:

• C

  // Register a sparse component
  ECS_COMPONENT(world, Position);
   
  ecs_add_id(world, ecs_id(Position), EcsSparse);

• C++

  // Register a sparse component
  world.component<Position>().add(flecs::Sparse);

• C#

  // Register a sparse component
  world.Component<Position>().Entity.add(Ecs.Sparse);

• Rust

  // Register a sparse component
  world.component::<Position>().add_trait::<flecs::Sparse>();

These kinds of tags are called "traits". To see which kinds of traits you can add to components to customize their behavior, see the component traits manual.

Registration

Components must be registered before they can be used. The following sections describe how component registration works for the different language bindings.

• C

  In C, the easiest way to register a component is with the ECS_COMPONENT macro:

  ECS_COMPONENT(world, Position);
   
  ecs_entity_t e = ecs_new_w(world, Position); // OK

  However, if you try to use the component from another function or across files, you will find that this doesn't work. To fix this, the component has to be forward declared. The following example shows how to forward declare a component that is used from multiple C files:

  // position.h
  extern ECS_COMPONENT_DECLARE(Position);

  // position.c
  #include "position.h"
   
  ECS_COMPONENT_DECLARE(Position);
   
  void register_components(ecs_world_t *world) {
      ECS_COMPONENT_DEFINE(world, Position);
  }

  // something.c
  #include "position.h"
   
  void do_something(ecs_world_t *world) {
      ecs_entity_t e = ecs_new_w(world, Position); // OK
  }
   
  int main(int argc, char *argv[]) {
      ecs_world_t *world = ecs_init();
   
      register_components(world);
   
      do_something(world);
   
      ecs_fini(world);
  }

  Note that this is just an example, and that typically you would not create a file per component. A more common way to organize components is with modules, which often register multiple components, amongst others. The following code shows an example with a module setup:

  // movement.h
  extern ECS_COMPONENT_DECLARE(Position);
  extern ECS_COMPONENT_DECLARE(Velocity);
   
  void MovementImport(ecs_world_t *world);

  // movement.c
  #include "movement.h"
   
  ECS_COMPONENT_DECLARE(Position);
  ECS_COMPONENT_DECLARE(Velocity);
   
  void MovementImport(ecs_world_t *world) {
      ECS_MODULE(world, Movement);
   
      ECS_COMPONENT_DEFINE(world, Position);
      ECS_COMPONENT_DEFINE(world, Velocity);
  }

  // something.c
  #include "movement.h"
   
  void do_something(ecs_world_t *world) {
      ecs_new_w(world, Position); // OK
  }
   
  int main(int argc, char *argv[]) {
      ecs_world_t *world = ecs_init();
   
      ECS_IMPORT(world, Movement);
   
      do_something(world);
   
      ecs_fini(world);
  }

• C++

  In C++ components are automatically registered upon first usage. The following example shows how:

  int main(int argc, char *argv[]) {
      flecs::world world;
   
      flecs::entity e1 = world.entity()
          .set(Position{10, 20}) // Position registered here
          .set(Velocity{1, 2}); // Velocity registered here
   
      flecs::entity e1 = world.entity()
          .set(Position{10, 20}) // Position already registered
          .set(Velocity{1, 2}); // Velocity already registered
  }

  Components can be registered in advance, which can be done for several reasons:

  • Makes it easier to see which components are used by an application
  • No unexpected registration code that suddenly runs in the middle of a frame
  • Component needs to be setup with traits, reflection data, hooks etc.

  To register a component in advance, do:

  world.component<Position>();

  In general it is recommended to register components in advance, and to only use automatic registration during prototyping. Applications can enforce manual registration by defining the FLECS_CPP_NO_AUTO_REGISTRATION macro at compile time, when building Flecs. This will cause a panic whenever a component is used that was not yet registered. Disabling auto registration also improves performance of the C++ API.

  A convenient way to organize component registration code is to use Flecs modules. An example:

  struct movement {
      movement(flecs::world& world) {
          world.component<Position>();
          world.component<Velocity>();
      }
  };
   
  int main(int argc, char *argv[]) {
      flecs::world world;
   
      world.import<movement>();
  }

• C#

  In C# components are automatically registered upon first usage. The following example shows how:

  public static void Main()
  {
      using World world = World.Create();
   
      Entity e1 = world.Entity()
          .Set(new Position(10, 20)) // Position registered here
          .Set(new Velocity(1, 2));  // Velocity registered here
   
      Entity e2 = world.Entity()
          .Set(new Position(10, 20)) // Position already registered
          .Set(new Velocity(1, 2));  // Velocity already registered
  }

  Components can be registered in advance, which can be done for several reasons:

  • Makes it easier to see which components are used by an application
  • No unexpected registration code that suddenly runs in the middle of a frame
  • Component needs to be setup with traits, reflection data, hooks etc.

  To register a component in advance, do:

  world.Component<Position>();

  In general it is recommended to register components in advance, and to only use automatic registration during prototyping.

  A convenient way to organize component registration code is to use Flecs modules. An example:

  public struct Movement : IFlecsModule
  {
      public void InitModule(World world)
      {
          world.Component<Position>();
          world.Component<Velocity>();
      }
  }
   
  public static void Main()
  {
      using World world = World.Create();
   
      world.Import<Movement>();
  }

• Rust

  In Rust components are automatically registered upon first usage. The following example shows how:

  fn main() {
      let world = World::new();
      let e1 = world
          .entity()
          .set(Position { x: 10.0, y: 20.0 }) // Position registered here
          .set(Velocity { x: 1.0, y: 2.0 }); // Velocity registered here
   
      let e2 = world
          .entity()
          .set(Position { x: 10.0, y: 20.0 }) // Position already registered
          .set(Velocity { x: 1.0, y: 2.0 }); // Velocity already registered
  }

  Components can be registered in advance, which can be done for several reasons:

  • Makes it easier to see which components are used by an application
  • No unexpected registration code that suddenly runs in the middle of a frame
  • Component needs to be setup with traits, reflection data, hooks etc.

  To register a component in advance, do:

  world.component::<Position>();

  In general it is recommended to register components in advance, and to only use automatic registration during prototyping. Applications can enforce manual registration by activating the rust feature flecs_manual_registration. This will cause a panic whenever a component is used that was not yet registered. Disabling auto registration also improves performance of the Rust API.

  A convenient way to organize component registration code is to use Flecs modules. An example:

  use flecs_ecs::prelude::*;
   
  #[derive(Component)]
  struct Movement;
   
  impl Module for Movement {
      fn module(world: &World) {
          world.module::<Movement>("Movement");
          // Define components, systems, triggers, ... as usual. They will be
          // automatically created inside the scope of the module.
      }
  }
   
  let world = World::new();
  world.import::<Movement>();

Runtime Type Registration

In some cases, typically when using scripting, components must be registered for types that do not exist at compile time. In Flecs this is possible by calling the ecs_component_init function. This function returns a component entity, which can then be used with regular ECS operations. An example:

• C

  // Register component
  ecs_entity_t comp = ecs_component(world, {
      .type = {
          .size = 8,
          .alignment = 8
      }
  });
   
  ecs_entity_t e = ecs_new(world);
   
  // Add component
  ecs_add_id(world, e, comp);
   
  // Get component
  const void *ptr = ecs_get_id(world, e, comp);

• C++

  ecs_component_desc_t desc = {0};
  desc.type.size = 8;
  desc.type.alignment = 8;
  flecs::entity_t comp = ecs_component_init(world, &desc);
   
  flecs::entity e = world.entity();
   
  // Add component
  e.add(comp);
   
  // Get component
  const void *ptr = e.get(comp);

• C#

  TODO

• Rust

  TODO

Often when registering a component at runtime, it needs to be described with reflection data so it can be serialized & deserialized. See the "reflection/runtime_component" example on how to do this.

When registering a component, it is good practice to give it a name. This makes it much easier to debug the application with tools like the explorer. To create a component with a name, we can pass a named entity to the ecs_component_init function. An example:

• C

  // Register named component
  ecs_entity_t comp = ecs_component(world, {
      .entity = ecs_entity(world, { .name = "MyComponent" }),
      .type = {
          .size = 8,
          .alignment = 8
      }
  });
   
  ecs_entity_t e = ecs_new(world);
   
  // Add component
  ecs_add_id(world, e, comp);
   
  // Get component
  const void *ptr = ecs_get_id(world, e, comp);

• C++

  ecs_component_desc_t desc = {0};
  desc.entity = world.entity("MyComponent");
  desc.type.size = 8;
  desc.type.alignment = 8;
  flecs::entity_t comp = ecs_component_init(world, &desc);
   
  flecs::entity e = world.entity();
   
  // Add component
  e.add(comp);
   
  // Get component
  const void *ptr = e.get(comp);

• C#

  TODO

• Rust

  TODO

See the documentation for ecs_component_desc_t for more component registration options.

Unregistration

A component can be unregistered from a world by deleting its entity. When a component is deleted, by default it will be removed from all entities that have the component. This behavior is customizable, see the "cleanup traits" section in the component trait manual.

The following example shows how to unregister a component:

• C

  ECS_COMPONENT(world, Position);
   
  // Create entity with Position
  ecs_entity_t e = ecs_new_w(world, Position);
   
  // Unregister the component
  ecs_delete(world, ecs_id(Position));
   
  // Position is removed from e

• C++

  flecs::entity pos = world.component<Position>();
   
  // Create entity with Position
  flecs::entity e = world.entity().add<Position>();
   
  // Unregister the component
  pos.destruct();
   
  // Position is removed from e

• C#

  Entity pos = world.Component<Position>();
   
  // Create entity with Position
  Entity e = world.Entity().Add<Position>();
   
  // Unregister the component
  pos.Destruct();
   
  // Position is removed from e

• Rust

  let pos = world.component::<Position>();
   
  // Create entity with Position
  let e = world.entity().add::<Position>();
   
  // Unregister the component
  pos.destruct();
   
  // Position is removed from e

Singletons

Singletons are components for which only a single instance exists on the world. They can be accessed on the world directly and do not require providing an entity. Singletons are useful for global game resources, such as game state, a handle to a physics engine or a network socket. An example:

• C

  ECS_COMPONENT(world, TimeOfDay);
   
  // Set singleton
  ecs_singleton_set(world, TimeOfDay, { 0.5 });
   
  // Unregister the component
  const TimeOfDay *t = ecs_singleton_get(world, TimeOfDay);

• C++

  // Set singleton
  world.set<TimeOfDay>({ 0.5 });
   
  // Get singleton
  const TimeOfDay *t = world.get<TimeOfDay>();

• C#

  // Set singleton
  world.Set(new TimeOfDay(0.5));
   
  // Get singleton
  ref readonly TimeOfDay t = ref world.Get<TimeOfDay>();

• Rust

  // Set singleton
  world.set(TimeOfDay { value: 0.5 });
   
  // Get singleton
  world.get::<&TimeOfDay>(|time| println!("{}", time.value));

Singletons are implemented as components that are added to themselves. The rationale for this design is that it allows for reusing all of the existing logic implemented for entities and components, as for the storage there is no fundamental difference between a singleton and a regular component. Additionally, it spreads out the storage of singletons across multiple entities which is favorable for performance when compared to a single entity that contains all singleton components. Since the component's entity is known during the singleton lookup, it makes most sense to use this as the entity on which to store the data.

The following example shows how the singleton APIs are equivalent to using the regular APIs with the component entity:

• C

  // Set singleton
  ecs_singleton_set(world, TimeOfDay, { 0.5 });
   
  // Equivalent to:
  ecs_set(world, ecs_id(TimeOfDay), TimeOfDay, {0.5});

• C++

  // Set singleton
  world.set<TimeOfDay>({ 0.5 });
   
  // Equivalent to:
  world.component<TimeOfDay>().set(TimeOfDay{ 0.5 })

• C#

  // Set singleton
  world.Set(new TimeOfDay(0.5));
   
  // Equivalent to:
  world.Component<TimeOfDay>().Entity.Set(new TimeOfDay(0.5));

• Rust

  // Set singleton
  world.set(TimeOfDay { value: 0.5 });
   
  // Equivalent to:
  world.component::<TimeOfDay>().set(TimeOfDay { value: 0.5 });

Disabling

Components can be disabled, which prevents them from being matched by queries. When a component is disabled, it behaves as if the entity doesn't have it. Only components that have the CanToggle trait can be disabled (see the component traits manual for more details). The following example shows how to disable a component:

• C

  // Register toggle-able component
  ECS_COMPONENT(world, Position);
  ecs_add_id(world, ecs_id(Position), EcsCanToggle);
   
  ecs_entity_t e = ecs_insert(world, ecs_value(Position, {10, 20}));
   
  // Disable component
  ecs_enable_component(world, e, Position, false);
  ecs_is_enabled(world, e, Position); // False
   
  // Enable component
  ecs_enable_component(world, e, Position, true);
  ecs_is_enabled(world, e, Position); // True

• C++

  // Register toggle-able component
  world.component<Position>().add(flecs::CanToggle);
   
  flecs::entity e = world.entity().set(Position{10, 20});
   
  // Disable component
  e.disable<Position>();
  e.is_enabled<Position>(); // False
   
  // Enable component
  e.enable<Position>();
  e.is_enabled<Position>()  // True

• C#

  ecs.Component<Position>().Entity
      .Add(Ecs.CanToggle);
   
  Entity e = world.Entity()
      .Set<Position>(new(10, 20));
   
  // Disable component
  e.Disable<Position>();
  e.IsEnabled<Position>(); // False
   
  // Enable component
  e.Enable<Position>();
  e.IsEnabled<Position>(); // True

• Rust

  // Register toggle-able component
  world
      .component::<Position>()
      .add_trait::<flecs::CanToggle>();
   
  let e = world.entity().set(Position { x: 10.0, y: 20.0 });
   
  // Disable component
  e.disable::<Position>();
   
  e.is_enabled::<Position>(); // False
   
  // Enable component
  e.enable::<Position>();
   
  e.is_enabled::<Position>(); // True

Disabling a component is a cheaper alternative to removing it from an entity, as it relies on setting a bit vs. moving an entity to another table. Component toggling can also be used to restore a component with its old value.