LiveComponents are a mechanism to compartmentalize state, markup, and events in LiveView.

LiveComponents are defined by using Phoenix.LiveComponent and are used by calling Phoenix.LiveView.Helpers.live_component/1 in a parent LiveView. They run inside the LiveView process but have their own state and life-cycle. For this reason, they are also often called "stateful components". This is a contrast to Phoenix.Component, also known as "function components", which are stateless.

The smallest LiveComponent only needs to define a render/1 function:

defmodule HeroComponent do
  # If you generated an app with mix phx.new --live,
  # the line below would be: use MyAppWeb, :live_component
  use Phoenix.LiveComponent

  def render(assigns) do
    ~H"""
    <div class="hero"><%= @content %></div>
    """
  end
end

A LiveComponent is rendered as:

<.live_component module={HeroComponent} id="hero" content={@content} />

You must always pass the module and id attributes. The id will be available as an assign and it must be used to uniquely identify the component. All other attributes will be available as assigns inside the LiveComponent.

life-cycle

Life-cycle

Stateful components are identified by the component module and their ID. Therefore, two stateful components with the same module and ID are treated as the same component. We often tie the component ID to some application based ID:

<.live_component module={UserComponent} id={@user.id} user={@user} />

When live_component/1 is called, mount/1 is called once, when the component is first added to the page. mount/1 receives the socket as argument. Then update/2 is invoked with all of the assigns given to live_component/1. If update/2 is not defined all assigns are simply merged into the socket. After the component is updated, render/1 is called with all assigns. On first render, we get:

mount(socket) -> update(assigns, socket) -> render(assigns)

On further rendering:

update(assigns, socket) -> render(assigns)

The given id is not automatically used as the DOM ID. If you want to set a DOM ID, it is your responsibility to do so when rendering:

defmodule UserComponent do
  use Phoenix.LiveComponent

  def render(assigns) do
    ~H"""
    <div id={"user-#{@id}"} class="user">
      <%= @user.name %>
    </div>
    """
  end
end

targeting-component-events

Targeting Component Events

Stateful components can also implement the handle_event/3 callback that works exactly the same as in LiveView. For a client event to reach a component, the tag must be annotated with a phx-target. If you want to send the event to yourself, you can simply use the @myself assign, which is an internal unique reference to the component instance:

<a href="#" phx-click="say_hello" phx-target={@myself}>
  Say hello!
</a>

Note that @myself is not set for stateless components, as they cannot receive events.

If you want to target another component, you can also pass an ID or a class selector to any element inside the targeted component. For example, if there is a UserComponent with the DOM ID of "user-13", using a query selector, we can send an event to it with:

<a href="#" phx-click="say_hello" phx-target="#user-13">
  Say hello!
</a>

In both cases, handle_event/3 will be called with the "say_hello" event. When handle_event/3 is called for a component, only the diff of the component is sent to the client, making them extremely efficient.

Any valid query selector for phx-target is supported, provided that the matched nodes are children of a LiveView or LiveComponent, for example to send the close event to multiple components:

<a href="#" phx-click="close" phx-target="#modal, #sidebar">
  Dismiss
</a>

preloading-and-update

Preloading and update

Stateful components also support an optional preload/1 callback. The preload/1 callback is useful when multiple components of the same type are rendered on the page and you want to preload or augment their data in batches.

For each rendering, the optional preload/1 and update/2 callbacks are called before render/1.

So on first render, the following callbacks will be invoked:

preload(list_of_assigns) -> mount(socket) -> update(assigns, socket) -> render(assigns)

On subsequent renders, these callbacks will be invoked:

preload(list_of_assigns) -> update(assigns, socket) -> render(assigns)

To provide a more complete understanding of why both callbacks are necessary, let's see an example. Imagine you are implementing a component and the component needs to load some state from the database. For example:

<.live_component module={UserComponent} id={user_id} />

A possible implementation would be to load the user on the update/2 callback:

def update(assigns, socket) do
  user = Repo.get!(User, assigns.id)
  {:ok, assign(socket, :user, user)}
end

However, the issue with said approach is that, if you are rendering multiple user components in the same page, you have a N+1 query problem. The preload/1 callback helps address this problem as it is invoked with a list of assigns for all components of the same type. For example, instead of implementing update/2 as above, one could implement:

def preload(list_of_assigns) do
  list_of_ids = Enum.map(list_of_assigns, & &1.id)

  users =
    from(u in User, where: u.id in ^list_of_ids, select: {u.id, u})
    |> Repo.all()
    |> Map.new()

  Enum.map(list_of_assigns, fn assigns ->
    Map.put(assigns, :user, users[assigns.id])
  end)
end

Now only a single query to the database will be made. In fact, the preloading algorithm is a breadth-first tree traversal, which means that even for nested components, the amount of queries are kept to a minimum.

Finally, note that preload/1 must return an updated list_of_assigns, keeping the assigns in the same order as they were given.

slots

Slots

LiveComponent can also receive slots, in the same way as a Phoenix.Component:

<.live_component module={MyComponent} id={@data.id} >
  <div>Inner content here</div>
</.live_component>

If the LiveComponent defines an update/2, be sure that the socket it returns includes the :inner_block assign it received.

See the docs for Phoenix.Component for more information.

live-patches-and-live-redirects

Live patches and live redirects

A template rendered inside a component can use Phoenix.LiveView.Helpers.live_patch/2 and Phoenix.LiveView.Helpers.live_redirect/2 calls. The live_patch/2 is always handled by the parentLiveView, as components do not provide handle_params.

managing-state

Managing state

Now that we have learned how to define and use components, as well as how to use preload/1 as a data loading optimization, it is important to talk about how to manage state in components.

Generally speaking, you want to avoid both the parent LiveView and the LiveComponent working on two different copies of the state. Instead, you should assume only one of them to be the source of truth. Let's discuss the two different approaches in detail.

Imagine a scenario where a LiveView represents a board with each card in it as a separate stateful LiveComponent. Each card has a form to allow update of the card title directly in the component, as follows:

defmodule CardComponent do
  use Phoenix.LiveComponent

  def render(assigns) do
    ~H"""
    <form phx-submit="..." phx-target={@myself}>
      <input name="title"><%= @card.title %></input>
      ...
    </form>
    """
  end

  ...
end

We will see how to organize the data flow to keep either the board LiveView or the card LiveComponents as the source of truth.

liveview-as-the-source-of-truth

LiveView as the source of truth

If the board LiveView is the source of truth, it will be responsible for fetching all of the cards in a board. Then it will call live_component/1 for each card, passing the card struct as argument to CardComponent:

<%= for card <- @cards do %>
  <.live_component module={CardComponent} card={card} id={card.id} board_id={@id} />
<% end %>

Now, when the user submits the form, CardComponent.handle_event/3 will be triggered. However, if the update succeeds, you must not change the card struct inside the component. If you do so, the card struct in the component will get out of sync with the LiveView. Since the LiveView is the source of truth, you should instead tell the LiveView that the card was updated.

Luckily, because the component and the view run in the same process, sending a message from the LiveComponent to the parent LiveView is as simple as sending a message to self():

defmodule CardComponent do
  ...
  def handle_event("update_title", %{"title" => title}, socket) do
    send self(), {:updated_card, %{socket.assigns.card | title: title}}
    {:noreply, socket}
  end
end

The LiveView then receives this event using Phoenix.LiveView.handle_info/2:

defmodule BoardView do
  ...
  def handle_info({:updated_card, card}, socket) do
    # update the list of cards in the socket
    {:noreply, updated_socket}
  end
end

Because the list of cards in the parent socket was updated, the parent LiveView will be re-rendered, sending the updated card to the component. So in the end, the component does get the updated card, but always driven from the parent.

Alternatively, instead of having the component send a message directly to the parent view, the component could broadcast the update using Phoenix.PubSub. Such as:

defmodule CardComponent do
  ...
  def handle_event("update_title", %{"title" => title}, socket) do
    message = {:updated_card, %{socket.assigns.card | title: title}}
    Phoenix.PubSub.broadcast(MyApp.PubSub, board_topic(socket), message)
    {:noreply, socket}
  end

  defp board_topic(socket) do
    "board:" <> socket.assigns.board_id
  end
end

As long as the parent LiveView subscribes to the board:<ID> topic, it will receive updates. The advantage of using PubSub is that we get distributed updates out of the box. Now, if any user connected to the board changes a card, all other users will see the change.

livecomponent-as-the-source-of-truth

LiveComponent as the source of truth

If each card LiveComponent is the source of truth, then the board LiveView must no longer fetch the card structs from the database. Instead, the board LiveView must only fetch the card ids, then render each component only by passing an ID:

<%= for card_id <- @card_ids do %>
  <.live_component module={CardComponent} id={card_id} board_id={@id} />
<% end %>

Now, each CardComponent will load its own card. Of course, doing so per card could be expensive and lead to N queries, where N is the number of cards, so we can use the preload/1 callback to make it efficient.

Once the card components are started, they can each manage their own card, without concerning themselves with the parent LiveView.

However, note that components do not have a Phoenix.LiveView.handle_info/2 callback. Therefore, if you want to track distributed changes on a card, you must have the parent LiveView receive those events and redirect them to the appropriate card. For example, assuming card updates are sent to the "board:ID" topic, and that the board LiveView is subscribed to said topic, one could do:

def handle_info({:updated_card, card}, socket) do
  send_update CardComponent, id: card.id, board_id: socket.assigns.id
  {:noreply, socket}
end

With Phoenix.LiveView.send_update/3, the CardComponent given by id will be invoked, triggering both preload and update callbacks, which will load the most up to date data from the database.

cost-of-stateful-components

Cost of stateful components

The internal infrastructure LiveView uses to keep track of stateful components is very lightweight. However, be aware that in order to provide change tracking and to send diffs over the wire, all of the components' assigns are kept in memory - exactly as it is done in LiveViews themselves.

Therefore it is your responsibility to keep only the assigns necessary in each component. For example, avoid passing all of LiveView's assigns when rendering a component:

<.live_component module={MyComponent} {assigns} />

Instead pass only the keys that you need:

<.live_component module={MyComponent} user={@user} org={@org} />

Luckily, because LiveViews and LiveComponents are in the same process, they share the data structure representations in memory. For example, in the code above, the view and the component will share the same copies of the @user and @org assigns.

You should also avoid using stateful components to provide abstract DOM components. As a guideline, a good LiveComponent encapsulates application concerns and not DOM functionality. For example, if you have a page that shows products for sale, you can encapsulate the rendering of each of those products in a component. This component may have many buttons and events within it. On the opposite side, do not write a component that is simply encapsulating generic DOM components. For instance, do not do this:

defmodule MyButton do
  use Phoenix.LiveComponent

  def render(assigns) do
    ~H"""
    <button class="css-framework-class" phx-click="click">
      <%= @text %>
    </button>
    """
  end

  def handle_event("click", _, socket) do
    _ = socket.assigns.on_click.()
    {:noreply, socket}
  end
end

Instead, it is much simpler to create a function component:

def my_button(%{text: _, click: _} = assigns) do
  ~H"""
  <button class="css-framework-class" phx-click={@click}>
    <%= @text %>
  </button>
  """
end

If you keep components mostly as an application concern with only the necessary assigns, it is unlikely you will run into issues related to stateful components.

limitations

Limitations

live-components-require-a-single-html-tag-at-the-root

Live Components require a single HTML tag at the root

Live Components require a single HTML tag at the root. It is not possible to have components that render only text or multiple tags.

svg-support

SVG support

Given components compartmentalize markup on the server, they are also rendered in isolation on the client, which provides great performance benefits on the client too.

However, when rendering components on the client, the client needs to choose the mime type of the component contents, which defaults to HTML. This is the best default but in some cases it may lead to unexpected results.

For example, if you are rendering SVG, the SVG will be interpreted as HTML. This may work just fine for most components but you may run into corner cases. For example, the <image> SVG tag may be rewritten to the <img> tag, since <image> is an obsolete HTML tag.

Luckily, there is a simple solution to this problem. Since SVG allows <svg> tags to be nested, you can wrap the component content into an <svg> tag. This will ensure that it is correctly interpreted by the browser.