LiveView provides rich, real-time user experiences with server-rendered HTML.

The LiveView programming model is declarative: instead of saying "once event X happens, change Y on the page", events in LiveView are regular messages which may cause changes to its state. Once the state changes, LiveView will re-render the relevant parts of its HTML template and push it to the browser, which updates itself in the most efficient manner. This means developers write LiveView templates as any other server-rendered HTML and LiveView does the hard work of tracking changes and sending the relevant diffs to the browser.

At the end of the day, a LiveView is nothing more than a process that receives events as messages and updates its state. The state itself is nothing more than functional and immutable Elixir data structures. The events are either internal application messages (usually emitted by Phoenix.PubSub) or sent by the client/browser.

LiveView is first rendered statically as part of regular HTTP requests, which provides quick times for "First Meaningful Paint", in addition to helping search and indexing engines. Then a persistent connection is established between client and server. This allows LiveView applications to react faster to user events as there is less work to be done and less data to be sent compared to stateless requests that have to authenticate, decode, load, and encode data on every request. The flipside is that LiveView uses more memory on the server compared to stateless requests.

Use cases

There are many use cases where LiveView is an excellent fit right now:

• Handling of user interaction and inputs, buttons, and forms - such as input validation, dynamic forms, autocomplete, etc;

• Events and updates pushed by server - such as notifications, dashboards, etc;

• Page and data navigation - such as navigating between pages, pagination, etc can be built with LiveView using the excellent live navigation feature set. This reduces the amount of data sent over the wire, gives developers full control over the LiveView life-cycle, while controlling how the browser tracks those changes in state;

There are also use cases which are a bad fit for LiveView:

• Animations - animations, menus, and general events that do not need the server in the first place are a bad fit for LiveView, as they can be achieved purely with CSS and/or CSS transitions;

Life-cycle

A LiveView begins as a regular HTTP request and HTML response, and then upgrades to a stateful view on client connect, guaranteeing a regular HTML page even if JavaScript is disabled. Any time a stateful view changes or updates its socket assigns, it is automatically re-rendered and the updates are pushed to the client.

You begin by rendering a LiveView typically from your router. When LiveView is first rendered, the mount/3 callback is invoked with the current params, the current session and the LiveView socket. As in a regular request, params contains public data that can be modified by the user. The session always contains private data set by the application itself. The mount/3 callback wires up socket assigns necessary for rendering the view. After mounting, render/1 is invoked and the HTML is sent as a regular HTML response to the client.

After rendering the static page, LiveView connects from the client to the server where stateful views are spawned to push rendered updates to the browser, and receive client events via phx- bindings. Just like the first rendering, mount/3 is invoked with params, session, and socket state, where mount assigns values for rendering. However in the connected client case, a LiveView process is spawned on the server, pushes the result of render/1 to the client and continues on for the duration of the connection. If at any point during the stateful life-cycle a crash is encountered, or the client connection drops, the client gracefully reconnects to the server, calling mount/3 once again.

Example

First, a LiveView requires two callbacks: mount/3 and render/1:

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

  def render(assigns) do
    ~L"""
    Current temperature: <%= @temperature %>
    """
  end

  def mount(_params, %{"current_user_id" => user_id}, socket) do
    temperature = Thermostat.get_user_reading(user_id)
    {:ok, assign(socket, :temperature, temperature)}
  end
end

The render/1 callback receives the socket.assigns and is responsible for returning rendered content. You can use Phoenix.LiveView.Helpers.sigil_L/2 to inline LiveView templates. If you want to use Phoenix.HTML helpers, remember to use Phoenix.HTML at the top of your LiveView.

With a LiveView defined, you first define the socket path in your endpoint, and point it to Phoenix.LiveView.Socket:

defmodule MyAppWeb.Endpoint do
  use Phoenix.Endpoint

  socket "/live", Phoenix.LiveView.Socket,
    websocket: [connect_info: [session: @session_options]]

  ...
end

Where @session_options are the options given to plug Plug.Session extracted to a module attribute.

And configure its signing salt in the endpoint:

config :my_app, MyAppWeb.Endpoint,
  ...,
  live_view: [signing_salt: ...]

You can generate a secure, random signing salt with the mix phx.gen.secret 32 task.

Next, decide where you want to use your LiveView.

You can serve the LiveView directly from your router (recommended):

defmodule MyAppWeb.Router do
  use Phoenix.Router
  import Phoenix.LiveView.Router

  scope "/", MyAppWeb do
    live "/thermostat", ThermostatLive
  end
end

You can also live_render from any template:

<h1>Temperature Control</h1>
<%= live_render(@conn, MyAppWeb.ThermostatLive) %>

Or you can live_render your view from any controller:

defmodule MyAppWeb.ThermostatController do
  ...
  import Phoenix.LiveView.Controller

  def show(conn, %{"id" => id}) do
    live_render(conn, MyAppWeb.ThermostatLive)
  end
end

When a LiveView is rendered, all of the data currently stored in the connection session (see Plug.Conn.get_session/1) will be given to the LiveView.

It is also possible to pass additional session information to the LiveView through a session parameter:

# In the router
live "/thermostat", ThermostatLive, session: %{"extra_token" => "foo"}

# In a view
<%= live_render(@conn, MyAppWeb.ThermostatLive, session: %{"extra_token" => "foo"}) %>

Notice the :session uses string keys as a reminder that session data is serialized and sent to the client. So you should always keep the data in the session to a minimum. For example, instead of storing a User struct, you should store the "user_id" and load the User when the LiveView mounts.

Once the LiveView is rendered, a regular HTML response is sent. Next, your client code connects to the server:

import {Socket} from "phoenix"
import LiveSocket from "phoenix_live_view"

let csrfToken = document.querySelector("meta[name='csrf-token']").getAttribute("content")
let liveSocket = new LiveSocket("/live", Socket, {params: {_csrf_token: csrfToken}})
liveSocket.connect()

After the client connects, mount/3 will be invoked inside a spawned LiveView process. At this point, you can use connected?/1 to conditionally perform stateful work, such as subscribing to pubsub topics, sending messages, etc. For example, you can periodically update a LiveView with a timer:

defmodule DemoWeb.ThermostatLive do
  use Phoenix.LiveView
  ...

  def mount(_params, %{"current_user_id" => user_id}, socket) do
    if connected?(socket), do: Process.send_after(self(), :update, 30000)

    case Thermostat.get_user_reading(user_id) do
      {:ok, temperature} ->
        {:ok, assign(socket, temperature: temperature, user_id: user_id)}

      {:error, reason} ->
        {:error, reason}
    end
  end

  def handle_info(:update, socket) do
    Process.send_after(self(), :update, 30000)
    {:ok, temperature} = Thermostat.get_reading(socket.assigns.user_id)
    {:noreply, assign(socket, :temperature, temperature)}
  end
end

We used connected?(socket) on mount to send our view a message every 30s if the socket is in a connected state. We receive the :update message in the handle_info/2 callback, just like in an Elixir GenServer, and update our socket assigns. Whenever a socket's assigns change, render/1 is automatically invoked, and the updates are sent to the client.

Collocating templates

In the examples above, we have placed the template directly inside the LiveView:

defmodule MyAppWeb.ThermostatLive do
  use Phoenix.LiveView

  def render(assigns) do
    ~L"""
    Current temperature: <%= @temperature %>
    """
  end

For larger templates, you can place them in a file in the same directory and same name as the LiveView. For example, if the file above is placed at lib/my_app_web/live/thermostat_live.ex, you can also remove the render/1 definition above and instead put the template code at lib/my_app_web/live/thermostat_live.html.leex.

Alternatively, you can keep the render/1 callback but delegate to an existing Phoenix.View module in your application. For example:

defmodule MyAppWeb.ThermostatLive do
  use Phoenix.LiveView

  def render(assigns) do
    Phoenix.View.render(MyAppWeb.PageView, "page.html", assigns)
  end
end

In all cases, each assign in the template will be accessible as @assign.

Assigns and LiveEEx templates

All of the data in a LiveView is stored in the socket as assigns. The assign/2 and assign/3 functions help store those values. Those values can be accessed in the LiveView as socket.assigns.name but they are accessed inside LiveView templates as @name.

Phoenix.LiveView's built-in templates are identified by the .leex extension (Live EEx) or ~L sigil. They are similar to regular .eex templates except they are designed to minimize the amount of data sent over the wire by splitting static and dynamic parts and tracking changes.

When you first render a .leex template, it will send all of the static and dynamic parts of the template to the client. After that, any change you do on the server will now send only the dynamic parts, and only if those parts have changed.

The tracking of changes is done via assigns. Imagine this template:

<h1><%= expand_title(@title) %></h1>

If the @title assign changes, then LiveView will execute expand_title(@title) and send the new content. If @title is the same, nothing is executed and nothing is sent.

Change tracking also works when accessing map/struct fields. Take this template:

<div id="user_<%= @user.id %>">
  <%= @user.name %>
</div>

If the @user.name changes but @user.id doesn't, then LiveView will re-render only @user.name and it will not execute or resend @user.id at all.

The change tracking also works when rendering other templates as long as they are also .leex templates:

<%= render "child_template.html", assigns %>

The assign tracking feature also implies that you MUST avoid performing direct operations in the template. For example, if you perform a database query in your template:

<%= for user <- Repo.all(User) do %>
  <%= user.name %>
<% end %>

Then Phoenix will never re-render the section above, even if the number of users in the database changes. Instead, you need to store the users as assigns in your LiveView before it renders the template:

assign(socket, :users, Repo.all(User))

Generally speaking, data loading should never happen inside the template, regardless if you are using LiveView or not. The difference is that LiveView enforces this best practice.

LiveEEx pitfalls

There are two common pitfalls to keep in mind when using the ~L sigil or .leex templates.

When it comes to do/end blocks, change tracking is supported only on blocks given to Elixir's basic constructs, such as if, case, for, and friends. If the do/end block is given to a library function or user function, such as content_tag, change tracking won't work. For example, imagine the following template that renders a div:

<%= content_tag :div, id: "user_#{@id}" do %>
  <%= @name %>
  <%= @description %>
<% end %>

LiveView knows nothing about content_tag, which means the whole div will be sent whenever any of the assigns change. This can be easily fixed by writing the HTML directly:

<div id="user_<%= @id %>">
  <%= @name %>
  <%= @description %>
</div>

Another pitfall of .leex templates is related to variables. Due to the scope of variables, LiveView has to disable change tracking whenever variables are used in the template, with the exception of variables introduced by Elixir basic case, for, and other block constructs. Therefore, you must avoid code like this in your LiveEEx:

<% some_var = @x + @y %>
<%= some_var %>

Instead, use a function:

<%= sum(@x, @y) %>

Similarly, do not define variables at the top of your render function:

def render(assigns) do
  sum = assigns.x + assigns.y

  ~L"""
  <%= sum %>
  """
end

Instead explicitly precompute the assign in your LiveView, outside of render:

assign(socket, sum: socket.assigns.x + socket.assigns.y)

Generally speaking, avoid accessing variables inside LiveViews. This also applies to the assigns variable, except when rendering another .leex template. In such cases, it is ok to pass the whole assigns, as LiveView will continue to perform change tracking in the called template:

<%= render "sidebar.html", assigns %>

Similarly, variables introduced by Elixir's block constructs are fine. For example, accessing the post variable defined by the comprehension below works as expected:

<%= for post <- @posts do %>
  ...
<% end %>

As are the variables matched defined in a case or cond:

<%= cond do %>
  <% is_nil(@post) -> %>
    ...
  <% @post -> %>
    ...
<% end %>

To sum up:

1. Avoid passing block expressions to library and custom functions

2. Never do anything on def render(assigns) besides rendering a template or invoking the ~L sigil

3. Avoid defining local variables, except within for, case, and friends

Bindings

Phoenix supports DOM element bindings for client-server interaction. For example, to react to a click on a button, you would render the element:

<button phx-click="inc_temperature">+</button>

Then on the server, all LiveView bindings are handled with the handle_event callback, for example:

def handle_event("inc_temperature", _value, socket) do
  {:ok, new_temp} = Thermostat.inc_temperature(socket.assigns.id)
  {:noreply, assign(socket, :temperature, new_temp)}
end

Click Events

The phx-click binding is used to send click events to the server. When any client event, such as a phx-click click is pushed, the value sent to the server will be chosen with the following priority:

• Any number of optional phx-value- prefixed attributes, such as:

  <div phx-click="inc" phx-value-myvar1="val1" phx-value-myvar2="val2">

  will send the following map of params to the server:

  def handle_event("inc", %{"myvar1" => "val1", "myvar2" => "val2"}, socket) do

  If the phx-value- prefix is used, the server payload will also contain a "value" if the element's value attribute exists.

• When receiving a map on the server, the payload will also include user defined metadata of the client event, or an empty map if none is set. For example, the following LiveSocket client option would send the coordinates and altKey information for all clicks:

  let liveSocket = new LiveSocket("/live", Socket, {
    params: {_csrf_token: csrfToken},
    metadata: {
      click: (e, el) => {
        return {
          altKey: e.altKey,
          clientX: e.clientX,
          clientY: e.clientY
        }
      }
    }
  })

The phx-capture-click event is just like phx-click, but instead of the click event being dispatched to the closest phx-click element as it bubbles up through the DOM, the event is dispatched as it propagates from the top of the DOM tree down to the target element. This is useful when wanting to bind click events without receiving bubbled events from child UI elements. Since capturing happens before bubbling, this can also be important for preparing or preventing behaviour that will be applied during the bubbling phase.

Focus and Blur Events

Focus and blur events may be bound to DOM elements that emit such events, using the phx-blur, and phx-focus bindings, for example:

<input name="email" phx-focus="myfocus" phx-blur="myblur"/>

To detect when the page itself has received focus or blur, phx-window-focus and phx-window-blur may be specified. These window level events may also be necessary if the element in consideration (most often a div with no tabindex) cannot receive focus. Like other bindings, phx-value-* can be provided on the bound element, and those values will be sent as part of the payload. For example:

<div class="container"
    phx-window-focus="page-active"
    phx-window-blur="page-inactive"
    phx-value-page="123">
  ...
</div>

The following window-level bindings are supported:

• phx-window-focus
• phx-window-blur
• phx-window-keydown
• phx-window-keyup

Form Events

To handle form changes and submissions, use the phx-change and phx-submit events. In general, it is preferred to handle input changes at the form level, where all form fields are passed to the LiveView's callback given any single input change. For example, to handle real-time form validation and saving, your template would use both phx_change and phx_submit bindings:

<%= f = form_for @changeset, "#", [phx_change: :validate, phx_submit: :save] %>
  <%= label f, :username %>
  <%= text_input f, :username %>
  <%= error_tag f, :username %>

  <%= label f, :email %>
  <%= text_input f, :email %>
  <%= error_tag f, :email %>

  <%= submit "Save" %>
</form>

Reminder: form_for/3 is a Phoenix.HTML helper. Don't forget to include use Phoenix.HTML at the top of your LiveView, if using Phoenix.HTML helpers. Also, if using error_tag/2, don't forget to import MyAppWeb.ErrorHelpers.

Next, your LiveView picks up the events in handle_event callbacks:

def render(assigns) ...

def mount(_params, _session, socket) do
  {:ok, assign(socket, %{changeset: Accounts.change_user(%User{})})}
end

def handle_event("validate", %{"user" => params}, socket) do
  changeset =
    %User{}
    |> Accounts.change_user(params)
    |> Map.put(:action, :insert)

  {:noreply, assign(socket, changeset: changeset)}
end

def handle_event("save", %{"user" => user_params}, socket) do
  case Accounts.create_user(user_params) do
    {:ok, user} ->
      {:noreply,
       socket
       |> put_flash(:info, "user created")
       |> redirect(to: Routes.user_path(MyAppWeb.Endpoint, MyAppWeb.User.ShowView, user))}

    {:error, %Ecto.Changeset{} = changeset} ->
      {:noreply, assign(socket, changeset: changeset)}
  end
end

The validate callback simply updates the changeset based on all form input values, then assigns the new changeset to the socket. If the changeset changes, such as generating new errors, render/1 is invoked and the form is re-rendered.

Likewise for phx-submit bindings, the same callback is invoked and persistence is attempted. On success, a :noreply tuple is returned and the socket is annotated for redirect with Phoenix.LiveView.redirect/2 to the new user page, otherwise the socket assigns are updated with the errored changeset to be re-rendered for the client.

Note: For proper form error tag updates, the error tag must specify which input it belongs to. This is accomplished with the phx-feedback-for attribute. Failing to add the phx-feedback-for attribute will result in displaying error messages for form fields that the user has not changed yet (e.g. required fields further down on the page).

For example, your MyAppWeb.ErrorHelpers may use this function:

def error_tag(form, field) do
  form.errors
  |> Keyword.get_values(field)
  |> Enum.map(fn error ->
    content_tag(:span, translate_error(error),
      class: "invalid-feedback",
      phx_feedback_for: input_id(form, field)
    )
  end)
end

Now, any DOM container with the phx-feedback-for attribute will receive a phx-no-feedback class in cases where the form fields has yet to receive user input/focus. The following css rules are generated in new projects to hide the errors:

.phx-no-feedback.invalid-feedback, .phx-no-feedback .invalid-feedback {
  display: none;
}

Submitting the form action over HTTP

The phx-trigger-action attribute can be added to a form to trigger a standard form submit on DOM patch to the URL specified in the form's standard action attribute. This is useful to perform pre-final validation of a LiveView form submit before posting to a controller route for operations that require Plug session mutation. For example, in your LiveView template you can annotate the phx-trigger-action with a boolean assign:

<%= f = form_for @changeset, Routes.reset_password_path(@socket, :create),
  phx_submit: :save,
  phx_trigger_action: @trigger_submit %>

Then in your LiveView, you can toggle the assign to trigger the form with the current fields on next render:

def handle_event("save", params, socket) do
  case validate_change_password(socket.assigns.user, params) do
    {:ok, changeset} ->
      {:noreply, assign(socket, changeset: changeset, trigger_submit: true)}

    {:error, changeset} ->
      {:noreply, assign(socket, changeset: changeset)}
    end
end

Number inputs

Number inputs are a special case in LiveView forms. On programmatic updates, some browsers will clear invalid inputs. So LiveView will not send change events from the client when an input is invalid, instead allowing the browser's native validation UI to drive user interaction. Once the input becomes valid, change and submit events will be sent normally.

<input type="number">

This is known to have a plethora of problems including accessibility, large numbers are converted to exponential notation and scrolling can accidentally increase or decrease the number.

As of early 2020, the following avoids these pitfalls and will likely serve your application's needs and users much better. According to https://caniuse.com/#search=inputmode, the following is supported by 90% of the global mobile market with Firefox yet to implement.

<input type="text" inputmode="numeric" pattern="[0-9]*">

Password inputs

Password inputs are also special cased in Phoenix.HTML. For security reasons, password field values are not reused when rendering a password input tag. This requires explicitly setting the :value in your markup, for example:

<%= password_input f, :password, value: input_value(f, :password) %>
<%= password_input f, :password_confirmation, value: input_value(f, :password_confirmation) %>
<%= error_tag f, :password %>
<%= error_tag f, :password_confirmation %>

Key Events

The onkeydown, and onkeyup events are supported via the phx-keydown, and phx-keyup bindings. Each binding supports a phx-key attribute, which triggers the event for the specific key press. If no phx-key is provided, the event is triggered for any key press. When pushed, the value sent to the server will contain the "key" that was pressed, plus any user-defined metadata. For example, pressing the Escape key looks like this:

%{"key" => "Escape"}

To capture additional user-defined metadata, the metadata option for keydown events may be provided to the LiveSocket constructor. For example:

let liveSocket = new LiveSocket("/live", Socket, {
  params: {_csrf_token: csrfToken},
  metadata: {
    keydown: (e, el) => {
      return {
        key: e.key,
        metaKey: e.metaKey,
        repeat: e.repeat
      }
    }
  }
})

To determine which key has been pressed you should use key value. The available options can be found on MDN or via the Key Event Viewer.

By default, the bound element will be the event listener, but a window-level binding may be provided via phx-window-keydown or phx-window-keyup, for example:

def render(assigns) do
  ~L"""
  <div id="thermostat" phx-window-keyup="update_temp">
    Current temperature: <%= @temperature %>
  </div>
  """
end

def handle_event("update_temp", %{"key" => "ArrowUp"}, socket) do
  {:ok, new_temp} = Thermostat.inc_temperature(socket.assigns.id)
  {:noreply, assign(socket, :temperature, new_temp)}
end

def handle_event("update_temp", %{"key" => "ArrowDown"}, socket) do
  {:ok, new_temp} = Thermostat.dec_temperature(socket.assigns.id)
  {:noreply, assign(socket, :temperature, new_temp)}
end

def handle_event("update_temp", _key, socket) do
  {:noreply, socket}
end

Rate limiting events with Debounce and Throttle

All events can be rate-limited on the client by using the phx-debounce and phx-throttle bindings, with the following behavior:

• phx-debounce - Accepts either a string integer timeout value, or "blur". When an int is provided, delays emitting the event by provided milliseconds. When "blur" is provided, delays emitting an input's change event until the field is blurred by the user. Debounce is typically emitted for inputs.

• phx-throttle - Accepts an integer timeout value to throttle the event in milliseconds. Unlike debounce, throttle will immediately emit the event, then rate limit the event at one event per provided timeout. Throttle is typically used to rate limit clicks, mouse and keyboard actions.

For example, to avoid validating an email until the field is blurred, while validating the username at most every 2 seconds after a user changes the field:

<form phx-change="validate" phx-submit="save">
  <input type="text" name="user[email]" phx-debounce="blur"/>
  <input type="text" name="user[username]" phx-debounce="2000"/>
</form>

And to rate limit a volume up click to once every second:

<button phx-click="volume_up" phx-throttle="1000">+</button>

Likewise, you may throttle held-down keydown:

<div phx-window-keydown="keydown" phx-throttle="500">
  ...
</div>

Unless held-down keys are required, a better approach is generally to use phx-keyup bindings which only trigger on key up, thereby being self-limiting. However, phx-keydown is useful for games and other use cases where a constant press on a key is desired. In such cases, throttle should always be used.

Debounce and Throttle special behavior

The following specialized behavior is performed for forms and keydown bindings:

• When a phx-submit, or a phx-change for a different input is triggered, any current debounce or throttle timers are reset for existing inputs.

• A phx-keydown binding is only throttled for key repeats. Unique keypresses back-to-back will dispatch the pressed key events.

LiveView Specific Events

The lv: event prefix supports LiveView specific features that are handled by LiveView without calling the user's handle_event/3 callbacks. Today, the following events are supported:

• lv:clear-flash – clears the flash when sent to the server. If a phx-value-key is provided, the specific key will be removed from the flash.

For example:

<p class="alert" phx-click="lv:clear-flash" phx-value-key="info">
  <%= live_flash(@flash, :info) %>
</p>

Security considerations of the LiveView model

As we have seen, LiveView begins its life-cycle as a regular HTTP request. Then a stateful connection is established. Both the HTTP request and the stateful connection receives the client data via parameters and session. This means that any session validation must happen both in the HTTP request and the stateful connection.

Mounting considerations

For example, if your HTTP request perform user authentication and confirmation on every request via Plugs, such as this:

plug :ensure_user_authenticated
plug :ensure_user_confirmed

Then the mount/3 callback of your LiveView should execute those same verifications:

def mount(params, %{"user_id" => user_id} = _session, socket) do
  socket = assign(socket, current_user: Accounts.get_user!(user_id))

  socket =
    if socket.assigns.current_user.confirmed_at do
      socket
    else
      redirect(socket, to: "/login")
    end

  {:ok, socket}
end

Given almost all mount/3 actions in your application will have to perform these exact steps, we recommend creating a function called assign_defaults/2 or similar, putting it in a new module like MyAppWeb.LiveHelpers, and modifying lib/my_app_web.ex so all LiveViews automatically import it:

def live_view do
  quote do
    # ...other stuff...
    import MyAppWeb.LiveHelpers
  end
end

Then make sure to call it in every LiveView's mount/3:

def mount(params, session, socket) do
  {:ok, assign_defaults(session, socket)}
end

Where MyAppWeb.LiveHelpers can be something like:

defmodule MyAppWeb.LiveHelpers do
  import Phoenix.LiveView

  def assign_defaults(%{"user_id" => user_id}, socket) do
    socket = assign(socket, current_user: Accounts.get_user!(user_id))

    if socket.assigns.current_user.confirmed_at do
      socket
    else
      redirect(socket, to: "/login")
    end
  end
end

One possible concern in this approach is that in regular HTTP requests the current user will be fetched twice: one in the HTTP request and another on mount. You can address this by using the assign_new function, that will reuse any of the connection assigns from the HTTP request:

def assign_defaults(%{"user_id" => user_id}, socket) do
  socket = assign_new(socket, :current_user, fn -> Accounts.get_user!(user_id) end)

  if socket.assigns.current_user.confirmed_at do
    socket
  else
    redirect(socket, to: "/login")
  end
end

Events considerations

It is also important to keep in mind that LiveView are stateful. Therefore, if you load any data on mount/3 and the data itself changes, the data won't be automatically propagated to the LiveView, unless you broadcast those events with Phoenix.PubSub.

Generally speaking, the simplest and safest approach is to perform authorization whenever there is an action. For example, imagine that you have a LiveView for a "Blog", and only editors can edit posts. Therefore, it is best to validate the user is an editor on mount and on every event:

def mount(%{"post_id" => post_id}, session, socket) do
  socket = assign_defaults(session, socket)
  post = Blog.get_post_for_user!(socket.assigns.current_user, post_id)
  {:ok, assign(socket, post: post)}
end

def handle_event("update_post", params, socket) do
  updated_post = Blog.update_post(socket.assigns.current_user, socket.assigns.post, params)
  {:noreply, assign(socket, post: updated_post)}
end

In the example above, the Blog context receives the user on both get and update operations, and always validates accordingly that the user has access, raising an error otherwise.

Disconnecting all instances of a given live user

Another security consideration is how to disconnect all instances of a given live user. For example, imagine the user logs outs, its account is terminated, or any other reason.

Luckily, it is possible to identify all LiveView sockets by setting a "live_socket_id" in the session. For example, when signing in a user, you could do:

conn
|> put_session(:current_user_id, user.id)
|> put_session(:live_socket_id, "users_socket:#{user.id}")

Now all LiveView sockets will be identified and listening to the given live_socket_id. You can disconnect all live users identified by said ID by broadcasting on the topic:

MyAppWeb.Endpoint.broadcast("users_socket:#{user.id}", "disconnect", %{})

Once a LiveView is disconnected, the client will attempt to reestablish the connection, re-executing the mount/3 callback. In this case, if the user is no longer logged in or it no longer has access to its current resource, mount/3 will fail and the user will be redirected to the proper page.

This is the same mechanism provided by Phoenix.Channels. Therefore, if your application uses both channels and LiveViews, you can use the same technique to disconnect any stateful connection.

Compartmentalizing markup and events with render, live_render, and live_component

We can render another template directly from a LiveView template by simply calling render:

render SomeView, "child_template.html", assigns

Where SomeView is a regular Phoenix.View, typically defined in lib/my_app_web/views/some_view.ex and "child_template.html" is defined at lib/my_app_web/templates/some_view/child_template.html.leex. As long as the template has the .leex extension and all assigns are passed, LiveView change tracking will also work across templates.

When rendering a child template, any of the phx-* events in the child template will be sent to the LiveView. In other words, similar to regular Phoenix templates, a regular render call does not start another LiveView. This means render is useful for sharing markup between views.

If you want to start a separate LiveView from within a LiveView, then you can call live_render/3 instead of render/3. This child LiveView runs in a separate process than the parent, with its own mount and handle_event callbacks. If a child LiveView crashes, it won't affect the parent. If the parent crashes, all children are terminated.

When rendering a child LiveView, the :id option is required to uniquely identify the child. A child LiveView will only ever be rendered and mounted a single time, provided its ID remains unchanged. Updates to a child session will be merged on the client, but not passed back up until either a crash and re-mount or a connection drop and recovery. To force a child to re-mount with new session data, a new ID must be provided.

Given that a LiveView runs on its own process, it is an excellent tool for creating completely isolated UI elements, but it is a slightly expensive abstraction if all you want is to compartmentalize markup and events. For example, if you are showing a table with all users in the system, and you want to compartmentalize this logic, rendering a separate LiveView for each user, then using a process per user would likely be too expensive. For these cases, LiveView provides Phoenix.LiveComponent, which are rendered using live_component/3:

<%= live_component(@socket, UserComponent, id: user.id, user: user) %>

Components have their own mount and handle_event callbacks, as well as their own state with change tracking support. Components are also lightweight as they "run" in the same process as the parent LiveView. However, this means an error in a component would cause the whole view to fail to render. See Phoenix.LiveComponent for a complete rundown on components.

To sum it up:

• render - compartmentalizes markup
• live_component - compartmentalizes state, markup, and events
• live_render - compartmentalizes state, markup, events, and error isolation

DOM patching and temporary assigns

A container can be marked with phx-update, allowing the DOM patch operations to avoid updating or removing portions of the LiveView, or to append or prepend the updates rather than replacing the existing contents. This is useful for client-side interop with existing libraries that do their own DOM operations. The following phx-update values are supported:

• replace - the default operation. Replaces the element with the contents
• ignore - ignores updates to the DOM regardless of new content changes
• append - append the new DOM contents instead of replacing
• prepend - prepend the new DOM contents instead of replacing

When using phx-update, a unique DOM ID must always be set in the container. If using "append" or "prepend", a DOM ID must also be set for each child. When appending or prepending elements containing an ID already present in the container, LiveView will replace the existing element with the new content instead appending or prepending a new element.

The "ignore" behaviour is frequently used when you need to integrate with another JS library. The "append" and "prepend" feature is often used with "Temporary assigns" to work with large amounts of data. Let's learn more.

Temporary assigns

By default, all LiveView assigns are stateful, which enables change tracking and stateful interactions. In some cases, it's useful to mark assigns as temporary, meaning they will be reset to a default value after each update. This allows otherwise large but infrequently updated values to be discarded after the client has been patched.

Imagine you want to implement a chat application with LiveView. You could render each message like this:

<%= for message <- @messages do %>
  <p><span><%= message.username %>:</span> <%= message.text %></p>
<% end %>

Every time there is a new message, you would append it to the @messages assign and re-render all messages.

As you may suspect, keeping the whole chat conversation in memory and resending it on every update would be too expensive, even with LiveView smart change tracking. By using temporary assigns and phx-update, we don't need to keep any messages in memory, and send messages to be appended to the UI only when there are new ones.

To do so, the first step is to mark which assigns are temporary and what values they should be reset to on mount:

def mount(_params, _session, socket) do
  socket = assign(socket, :messages, load_last_20_messages())
  {:ok, socket, temporary_assigns: [messages: []]}
end

On mount we also load the initial number of messages we want to send. After the initial render, the initial batch of messages will be reset back to an empty list.

Now, whenever there are one or more new messages, we will assign only the new messages to @messages:

socket = assign(socket, :messages, new_messages)

In the template, we want to wrap all of the messages in a container and tag this content with phx-update. Remember, we must add an ID to the container as well as to each child:

<div id="chat-messages" phx-update="append">
  <%= for message <- @messages do %>
    <p id="<%= message.id %>">
      <span><%= message.username %>:</span> <%= message.text %>
    </p>
  <% end %>
</div>

When the client receives new messages, it now knows to append to the old content rather than replace it.

Live navigation

LiveView provides functionality to allow page navigation using the browser's pushState API. With live navigation, the page is updated without a full page reload.

You can trigger live navigation in two ways:

• From the client - this is done by replacing Phoenix.HTML.Link.link/2 by Phoenix.LiveView.Helpers.live_patch/2 or Phoenix.LiveView.Helpers.live_redirect/2

• From the server - this is done by replacing Phoenix.Controller.redirect/2 calls by Phoenix.LiveView.push_patch/2 or Phoenix.LiveView.push_redirect/2.

For example, in a template you may write:

<%= live_patch "next", to: Routes.live_path(@socket, MyLive, @page + 1) %>

or in a LiveView:

{:noreply, push_patch(socket, to: Routes.live_path(socket, MyLive, page + 1))}

The "patch" operations must be used when you want to navigate to the current LiveView, simply updating the URL and the current parameters, without mounting a new LiveView. When patch is used, the handle_params/3 callback is invoked and the minimal set of changes are sent to the client. See the next section for more information.

The "redirect" operations must be used when you want to dismount the current LiveView and mount a new one. In those cases, an Ajax request is made to fetch the necessary information about the new LiveView, which is mounted in place of the current one within the current layout. While redirecting, a phx-disconnected class is added to the LiveView, which can be used to indicate to the user a new page is being loaded.

At the end of the day, regardless if you invoke link/2, live_patch/2, and live_redirect/2 from the client, or redirect/2, push_patch/2, and push_redirect/2 from the server, the user will end-up on the same page. The difference between those is mostly the amount of data sent over the wire:

• link/2 and redirect/2 do full page reloads

• live_redirect/2 and push_redirect/2 mounts a new LiveView while keeping the current layout

• live_patch/2 and push_patch/2 updates the current LiveView and sends only the minimal diff

An easy rule of thumb is to stick with live_redirect/2 and push_redirect/2 and use the patch helpers only in the cases where you want to minimize the amount of data sent when navigating within the same LiveView (for example, if you want to change the sorting of a table while also updating the URL).

handle_params/3

The handle_params/3 callback is invoked after mount/3 and before the initial render. It is also invoked every time live_patch/2 or push_patch/2 are used. It receives the request parameters as first argument, the url as second, and the socket as third.

For example, imagine you have a UserTable LiveView to show all users in the system and you define it in the router as:

live "/users", UserTable

Now to add live sorting, you could do:

<%= live_patch "Sort by name", to: Routes.live_path(@socket, UserTable, %{sort_by: "name"}) %>

When clicked, since we are navigating to the current LiveView, handle_params/3 will be invoked. Remember you should never trust the received params, so you must use the callback to validate the user input and change the state accordingly:

def handle_params(params, _uri, socket) do
  socket =
    case params["sort_by"] do
      sort_by when sort_by in ~w(name company) -> assign(socket, sort_by: sort)
      _ -> socket
    end

  {:noreply, load_users(socket)}
end

As with other handle_* callback, changes to the state inside handle_params/3 will trigger a server render.

Note the parameters given to handle_params/3 are the same as the ones given to mount/3. So how do you decide which callback to use to load data? Generally speaking, data should always be loaded on mount/3, since mount/3 is invoked once per LiveView life-cycle. Only the params you expect to be changed via live_patch/2 or push_patch/2 must be loaded on handle_params/3.

For example, imagine you have a blog. The URL for a single post is: "/blog/posts/:post_id". In the post page, you have comments and they are paginated. You use live_patch/2 to update the shown comments every time the user paginates, updating the URL to "/blog/posts/:post_id?page=X". In this example, you will access "post_id" on mount/3 and the page of comments on handle_params/3.

Furthermore, it is very important to not access the same parameters on both mount/3 and handle_params/3. For example, do NOT do this:

def mount(%{"post_id" => post_id}, session, socket) do
  # do something with post_id
end

def handle_params(%{"post_id" => post_id, "page" => page}, url, socket) do
  # do something with post_id and page
end

If you do that, because mount/3 is called once and handle_params/3 multiple times, the "post_id" read on mount can get out of sync with the one in handle_params/3. So once a parameter is read on mount, it should not be read elsewhere. Instead, do this:

def mount(%{"post_id" => post_id}, session, socket) do
  # do something with post_id
end

def handle_params(%{"sort_by" => sort_by}, url, socket) do
  post_id = socket.assigns.post.id
  # do something with sort_by
end

Replace page address

LiveView also allows the current browser URL to be replaced. This is useful when you want certain events to change the URL but without polluting the browser's history. This can be done by passing the replace: true option to any of the navigation helpers.

Multiple LiveViews in the same page

LiveView allows you to have multiple LiveViews in the same page by calling Phoenix.LiveView.Helpers.live_render/3 in your templates. However, only the LiveViews defined directly in your router can use the "Live Navigation" functionality described here. This is important because LiveViews work closely with your router, guaranteeing you can only navigate to known routes.

Live layouts

When working with LiveViews, there are usually three layouts to be considered:

• the root layout - this is a layout used by both LiveView and regular views. This layout typically contains the <html> definition alongside the head and body tags. Any content defined in the root layout will remain the same, even as you live navigate across LiveViews

• the app layout - this is the default application layout which is not included or used by LiveViews;

• the live layout - this is the layout which wraps a LiveView and is rendered as part of the LiveView life-cycle

Overall, those layouts are found in templates/layout with the following names:

* root.html.leex
* app.html.eex
* live.html.leex

The "root" layout is shared by both "app" and "live" layouts. It is rendered only on the initial request and therefore it has access to the @conn assign. The root layout must be defined in your router:

plug :put_root_layout, {MyAppWeb.LayoutView, :root}

Alternatively, the root layout can be passed to the live macro of your live routes:

live "/dashboard", MyApp.Dashboard, layout: {MyAppWeb.LayoutView, :root}

The "app" and "live" layouts are often small and similar to each other, but the "app" layout uses the @conn and is used as part of the regular request life-cycle, and the "live" layout is part of the LiveView and therefore has direct access to the @socket.

For example, you can define a new live.html.leex layout with dynamic content. You must use @inner_content where the output of the actual template will be placed at:

<p><%= live_flash(@flash, :notice) %></p>
<p><%= live_flash(@flash, :error) %></p>
<%= @inner_content %>

To use the live layout, update your LiveView to pass the :layout option to use Phoenix.LiveView:

use Phoenix.LiveView, layout: {MyAppWeb.LayoutView, "live.html"}

If you are using Phoenix v1.5, the layout is automatically set when generating apps with the mix phx.new --live flag.

The :layout option does not apply for LiveViews rendered within other LiveViews. If you want to render child live views or opt-in to a layout, use :layout as an option in mount:

  def mount(_params, _session, socket) do
    socket = assign(socket, new_message_count: 0)
    {:ok, socket, layout: {MyAppWeb.LayoutView, "live.html"}}
  end

Note: The layout will be wrapped by the LiveView's :container tag.

Updating the HTML document title

Because the root layout from the Plug pipeline is rendered outside of LiveView, the contents cannot be dynamically changed. The one exception is the <title> of the HTML document. Phoenix LiveView special cases the @page_title assign to allow dynamically updating the title of the page, which is useful when using live navigation, or annotating the browser tab with a notification. For example, to update the user's notification count in the browser's title bar, first set the page_title assign on mount:

  def mount(_params, _session, socket) do
    socket = assign(socket, page_title: "Latest Posts")
    {:ok, socket}
  end

Then access @page_title in the root layout:

<title><%= @page_title %></title>

You can also use Phoenix.LiveView.Helpers.live_title_tag/2 to support adding automatic prefix and suffix to the page title when rendered and on subsequent updates:

<%= live_title_tag @page_title, prefix: "MyApp – " %>

Although the root layout is not updated by LiveView, by simply assigning to page_title, LiveView knows you want the title to be updated:

def handle_info({:new_messages, count}, socket) do
  {:noreply, assign(socket, page_title: "Latest Posts (#{count} new)")}
end

Note: If you find yourself needing to dynamically patch other parts of the base layout, such as injecting new scripts or styles into the <head> during live navigation, then a regular, non-live, page navigation should be used instead. Assigning the @page_title updates the document.title directly, and therefore cannot be used to update any other part of the base layout.

Error and exception handling

As with any other ELixir code, exceptions may happen during the LiveView life-cycle. In this section we will describe how LiveView reacts to errors at different stages.

Expected scenarios

In this section, we will talk about error cases that you expect to happen within your application. For example, a user filling in a form with invalid data is expected. In a LiveView, we typically handle those cases by storing a change in the LiveView state, which causes the LiveView to be re-rendered with the error message.

We may also use flash messages for this. For example, imagine you have a page to manage all "Team members" in an organization. However, if there is only one member left in the organization, they should not be allowed to leave. You may want to handle this by using flash messages:

if MyApp.Org.leave(socket.assigns.current_org, member) do
  {:noreply, socket}
else
  {:noreply, put_flash(socket, :error, "last member cannot leave organization")}
end

However, one may argue that, if the last member of an organization cannot leave it, it may be better to not even show the "Leave" button in the UI when the organization has only one member.

Given the button does not appear in the UI, triggering the "leave" when the organization has now only one member is an unexpected scenario. This means we can probably rewrite the code above to:

true = MyApp.Org.leave(socket.assigns.current_org, member)
{:noreply, socket}

If leave returns false by any chance, it will just raise. Or you can even provide a leave! function that raises a specific exception:

MyApp.Org.leave!(socket.assigns.current_org, member)
{:noreply, socket}

However, what will happen with a LiveView in case of exceptions? Let's talk about unexpected scenarios.

Unexpected scenarios

Elixir developers tend to write assertive code. This means that, if we expect leave to always return true, we can explicitly match on its result, as we did above:

true = MyApp.Org.leave(socket.assigns.current_org, member)
{:noreply, socket}

If leave fails and returns false, an exception is raised. It is common for Elixir developers to use exceptions for unexpected scenarios in their Phoenix applications.

For example, if you are building an application where a user may belong to one or more organizations, when accessing the organization page, you may want to check that the user has access to it like this:

organizations_query = Ecto.assoc(socket.assigns.current_user, :organizations)
Repo.get!(organizations_query, params["org_id"])

The code above builds a query that returns all organizations that belongs to the current user and then validates that the given "org_id" belongs to the user. If there is no such "org_id" or if the user has no access to it, an Ecto.NotFoundError exception is raised.

During a regular controller request, this exception will be converted to a 404 exception and rendered as a custom error page, as detailed here. To understand how a LiveView reacts to exceptions, we need to consider two scenarios: exceptions on mount and during any event.

Exceptions on mount

Given the code on mount runs both on the initial disconnected render and the connected render, an exception on mount will trigger the following events:

Exceptions during disconnected render:

1. An exception on mount is caught and converted to an exception page by Phoenix error views - pretty much like the way it works with controllers

Exceptions during connected render:

1. An exception on mount will crash the LiveView process - which will be logged
2. Once the client has noticed the crash during mount, it will fully reload the page
3. Reloading the page will start a disconnected render, that will cause mount to be invoked again and most likely raise the same exception. Except this time it will be caught and converted to an exception page by Phoenix error views

In other words, LiveView will reload the page in case of errors, making it fail as if LiveView was not involved in the rendering in the first place.

Exceptions on events (handle_info, handle_event, etc)

If the error happens during an event, the LiveView process will crash. The client will notice the error and remount the LiveView - without reloading the page. This is enough to update the page and show the user the latest information.

For example, let's say two users try to leave the organization at the same time. In this case, both of them see the "Leave" button, but our leave function call will succeed only for one of them:

true = MyApp.Org.leave(socket.assigns.current_org, member)
{:noreply, socket}

When the exception raises, the client will remount the LiveView. Once you remount, your code will now notice that there is only one user in the organization and therefore no longer show the "Leave" button. In other words, by remounting, we often update the state of the page, allowing exceptions to be automatically handled.

Note that the choice between conditionally checking on the result of the leave function with an if, or simply asserting it returns true, is completely up to you. If the likelihood of everyone leaving the organization at the same time is low, then you may as well treat it as an unexpected scenario. Although other developers will be more comfortable by explicitly handling those cases. In both scenarios, LiveView has you covered.

Using Gettext for internationalization

For internationalization with gettext, the locale used within your Plug pipeline can be stored in the Plug session and restored within your LiveView mount. For example, after user signs in or preference changes, you can write the locale to the session:

def put_user_session(conn, current_user) do
  locale = get_locale_for_user(current_user)
  Gettext.put_locale(MyApp.Gettext, locale)

  conn
  |> put_session(:user_id, current_user.id)
  |> put_session(:locale, locale)
end

Then in your LiveView mount/3, you can restore the locale:

def mount(_params, %{"locale" => locale}, socket) do
  Gettext.put_locale(MyApp.Gettext, locale)
  {:ok socket}
end

JavaScript client specific

As seen earlier, you start by instantiating a single LiveSocket instance to enable LiveView client/server interaction, for example:

import {Socket} from "phoenix"
import LiveSocket from "phoenix_live_view"

let csrfToken = document.querySelector("meta[name='csrf-token']").getAttribute("content")
let liveSocket = new LiveSocket("/live", Socket, {params: {_csrf_token: csrfToken}})
liveSocket.connect()

All options are passed directly to the Phoenix.Socket constructor, except for the following LiveView specific options:

• bindingPrefix - the prefix to use for phoenix bindings. Defaults "phx-"
• params - the connect_params to pass to the view's mount callback. May be a literal object or closure returning an object. When a closure is provided, the function receives the view's phx-view name.
• hooks – a reference to a user-defined hooks namespace, containing client callbacks for server/client interop. See the interop section below for details.

Debugging Client Events

To aid debugging on the client when troubleshooting issues, the enableDebug() and disableDebug() functions are exposed on the LiveSocket JavaScript instance. Calling enableDebug() turns on debug logging which includes LiveView life-cycle and payload events as they come and go from client to server. In practice, you can expose your instance on window for quick access in the browser's web console, for example:

// app.js
let liveSocket = new LiveSocket(...)
liveSocket.connect()
window.liveSocket = liveSocket

// in the browser's web console
>> liveSocket.enableDebug()

The debug state uses the browser's built-in sessionStorage, so it will remain in effect for as long as your browser session lasts.

Simulating Latency

Proper handling of latency is critical for good UX. LiveView's CSS loading states allow the client to provide user feedback while awaiting a server response. In development, near zero latency on localhost does not allow latency to be easily represented or tested, so LiveView includes a latency simulator with the JavaScript client to ensure your application provides a pleasant experience. Like the enableDebug() function above, the LiveSocket instance includes enableLatencySim(milliseconds) and disableLatencySim() functions which apply throughout the current browser session. The enableLatencySim function accepts an integer in milliseconds for the round-trip-time to the server. For example:

// app.js
let liveSocket = new LiveSocket(...)
liveSocket.connect()
window.liveSocket = liveSocket

// in the browser's web console
>> liveSocket.enableLatencySim(1000)
[Log] latency simulator enabled for the duration of this browser session.
      Call disableLatencySim() to disable

Forms and input handling

The JavaScript client is always the source of truth for current input values. For any given input with focus, LiveView will never overwrite the input's current value, even if it deviates from the server's rendered updates. This works well for updates where major side effects are not expected, such as form validation errors, or additive UX around the user's input values as they fill out a form.

For these use cases, the phx-change input does not concern itself with disabling input editing while an event to the server is in flight. When a phx-change event is sent to the server, the input tag and parent form tag receive the phx-change-loading css class, then the payload is pushed to the server with a "_target" param in the root payload containing the keyspace of the input name which triggered the change event.

For example, if the following input triggered a change event:

<input name="user[username]"/>

The server's handle_event/3 would receive a payload:

%{"_target" => ["user", "username"], "user" => %{"username" => "Name"}}

The phx-submit event is used for form submissions where major side effects typically happen, such as rendering new containers, calling an external service, or redirecting to a new page.

On submission of a form bound with a phx-submit event:

1. The form's inputs are set to readonly
2. Any submit button on the form is disabled
3. The form receives the "phx-submit-loading" class

On completion of server processing of the phx-submit event:

1. The submitted form is reactivated and loses the "phx-submit-loading" class
2. The last input with focus is restored (unless another input has received focus)
3. Updates are patched to the DOM as usual

To handle latent events, any HTML tag can be annotated with phx-disable-with, which swaps the element's innerText with the provided value during event submission. For example, the following code would change the "Save" button to "Saving...", and restore it to "Save" on acknowledgment:

<button type="submit" phx-disable-with="Saving...">Save</button>

You may also take advantage of LiveView's CSS loading state classes to swap out your form content while the form is submitting. For example, with the following rules in your app.css:

.while-submitting { display: none; }
.inputs { display: block; }

.phx-submit-loading {
  .while-submitting { display: block; }
  .inputs { display: none; }
}

You can show and hide content with the following markup:

<form phx-change="update">
  <div class="while-submitting">Please wait while we save our content...</div>
  <div class="inputs">
    <input type="text" name="text" value="<%= @text %>">
  </div>
</form>

Additionally, we strongly recommend including a unique HTML "id" attribute on the form. When DOM siblings change, elements without an ID will be replaced rather than moved, which can cause issues such as form fields losing focus.

Form Recovery following crashes or disconnects

By default, all forms marked with phx-change will recover input values automatically after the user has reconnected or the LiveView has remounted after a crash. This is achieved by the client triggering the same phx-change to the server as soon as the mount has been completed.

Note: if you want to see form recovery working in development, please make sure to disable live reloading in development by commenting out the LiveReload plug in your endpoint.ex file or by setting code_reloader: false in your config/dev.exs. Otherwise live reloading may cause the current page to be reloaded whenever you restart the server, which will discard all form state.

For most use cases, this is all you need and form recovery will happen without consideration. In some cases, where forms are built step-by-step in a stateful fashion, it may require extra recovery handling on the server outside of your existing phx-change callback code. To enable specialized recovery, provide a phx-auto-recover binding on the form to specify a different event to trigger for recovery, which will receive the form params as usual. For example, imagine a LiveView wizard form where the form is stateful and built based on what step the user is on and by prior selections:

<form phx-change="validate_wizard_step" phx-auto-recover="recover_wizard">

On the server, the "validate_wizard_step" event is only concerned with the current client form data, but the server maintains the entire state of the wizard. To recover in this scenario, you can specify a recovery event, such as "recover_wizard" above, which would wire up to the following server callbacks in your LiveView:

def handle_event("validate_wizard_step", params, socket) do
  # regular validations for current step
  {:noreply, socket}
end

def handle_event("recover_wizard", params, socket) do
  # rebuild state based on client input data up to the current step
  {:noreply, socket}
end

To forgo automatic form recovery, set phx-auto-recover="ignore".

Loading state and errors

By default, the following classes are applied to the LiveView's parent container:

• "phx-connected" - applied when the view has connected to the server
• "phx-disconnected" - applied when the view is not connected to the server
• "phx-error" - applied when an error occurs on the server. Note, this class will be applied in conjunction with "phx-disconnected" if connection to the server is lost.

All phx- event bindings apply their own css classes when pushed. For example the following markup:

<button phx-click="clicked" phx-window-keydown="key">...</button>

On click, would receive the phx-click-loading class, and on keydown would receive the phx-keydown-loading class. The css loading classes are maintained until an acknowledgement is received on the client for the pushed event.

In the case of forms, when a phx-change is sent to the server, the input element which emitted the change receives the phx-change-loading class, along with the parent form tag. The following events receive css loading classes:

• phx-click - phx-click-loading
• phx-change - phx-change-loading
• phx-submit - phx-submit-loading
• phx-focus - phx-focus-loading
• phx-blur - phx-blur-loading
• phx-window-keydown - phx-keydown-loading
• phx-window-keyup - phx-keyup-loading

For live page navigation via live_redirect and live_patch, as well as form submits via phx-submit, the JavaScript events "phx:page-loading-start" and "phx:page-loading-stop" are dispatched on window. Additionally, any phx- event may dispatch page loading events by annotating the DOM element with phx-page-loading. This is useful for showing main page loading status, for example:

// app.js
import NProgress from "nprogress"
window.addEventListener("phx:page-loading-start", info => NProgress.start())
window.addEventListener("phx:page-loading-stop", info => NProgress.done())

The info object will contain a kind key, with a value in one of the following events:

• "redirect" - the event was triggered by a redirect
• "patch" - the event was triggered by a patch
• "initial" - the event was triggered by initial page load
• "element" - the event was triggered by a phx- bound element, such as phx-click

For all kinds of page loading events, all but "element" will receive an additional to key in the info metadata pointing to the href associated with the page load.

In the case of an "element" page loading event, the info will contain a "target" key containing the DOM element which triggered the page loading state.

JS Interop and client-controlled DOM

To handle custom client-side JavaScript when an element is added, updated, or removed by the server, a hook object may be provided with the following life-cycle callbacks:

• mounted - the element has been added to the DOM and its server LiveView has finished mounting
• beforeUpdate - the element is about to be updated in the DOM. Note: any call here must be synchronous as the operation cannot be deferred or cancelled.
• updated - the element has been updated in the DOM by the server
• beforeDestroy - the element is about to be removed from the DOM. Note: any call here must be synchronous as the operation cannot be deferred or cancelled.
• destroyed - the element has been removed from the page, either by a parent update, or by the parent being removed entirely
• disconnected - the element's parent LiveView has disconnected from the server
• reconnected - the element's parent LiveView has reconnected to the server

The above life-cycle callbacks have in-scope access to the following attributes:

• el - attribute referencing the bound DOM node,
• viewName - attribute matching the DOM node's phx-view value
• pushEvent(event, payload) - method to push an event from the client to the LiveView server
• pushEventTo(selector, event, payload) - method to push targeted events from the client to LiveViews and LiveComponents.

For example, the markup for a controlled input for phone-number formatting could be written like this:

<input type="text" name="user[phone_number]" id="user-phone-number" phx-hook="PhoneNumber" />

Then a hook callback object could be defined and passed to the socket:

let Hooks = {}
Hooks.PhoneNumber = {
  mounted() {
    this.el.addEventListener("input", e => {
      let match = this.el.value.replace(/\D/g, "").match(/^(\d{3})(\d{3})(\d{4})$/)
      if(match) {
        this.el.value = `${match[1]}-${match[2]}-${match[3]}`
      }
    })
  }
}

let liveSocket = new LiveSocket("/live", Socket, {hooks: Hooks, ...})
...

The hook can push events to the LiveView by using the pushEvent function. Communication with the hook can be done by using data attributes on the container. For example, to implement infinite scrolling, one might do:

<div id="infinite-scroll" phx-hook="InfiniteScroll" data-page="<%= @page %>" />

And then in the client:

Hooks.InfiniteScroll = {
  page() { return this.el.dataset.page },
  mounted(){
    this.pending = this.page()
    window.addEventListener("scroll", e => {
      if(this.pending == this.page() && scrollAt() > 90){
        this.pending = this.page() + 1
        this.pushEvent("load-more", {})
      }
    })
  },
  updated(){ this.pending = this.page() }
}

Note: when using phx-hook, a unique DOM ID must always be set.

For integration with client-side libraries which require a broader access to full DOM management, the LiveSocket constructor accepts a dom option with an onBeforeElUpdated callback. The fromEl and toEl DOM nodes are passed to the function just before the DOM patch operations occurs in LiveView. This allows external libraries to (re)initialize DOM elements or copy attributes as necessary as LiveView performs its own patch operations. The update operation cannot be cancelled or deferred, and the return value is ignored. For example, the following option could be used to add Alpine.js support to your project:

let liveSocket = new LiveSocket("/live", Socket, {
  ...,
  dom: {
    onBeforeElUpdated(from, to){
      if(from.__x){ window.Alpine.clone(from.__x, to) }
    }
  },
})

Endpoint configuration

LiveView accepts the following configuration in your endpoint under the :live_view key:

• :signing_salt (required) - the salt used to sign data sent to the client

• :hibernate_after (optional) - the idle time in milliseconds allowed in the LiveView before compressing its own memory and state. Defaults to 15000ms (15 seconds)