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

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 provides many features that make it excellent to build rich, real-time user experiences:

• By building on top of Elixir processes and Phoenix.Channels, LiveView scales well vertically (from small to large instances) and horizontally (by adding more instances);

• 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;

• LiveView performs diff tracking. If the LiveView state changes, it will only re-render those changes. This reduces latency and the amount of data sent over the wire;

• LiveView tracks static and dynamic content. Any server-rendered HTML is made of static parts (i.e. that never change) and dynamic ones. On the first render, LiveView sends the static content and in future updates only the modified dynamic content is resent;

• (Coming soon) LiveView uses the Erlang Term Format to send messages to the client. This binary-based format is quite efficient on the server and uses less data over the wire;

• (Coming soon) LiveView includes a latency simulator, which allows you to simulate how your application would behave with greater latency and guides you to provide meaningful feedback to users while they wait for events to be processed;

Furthermore, by keeping a persistent connection between client and server, LiveView applications can 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 but currently you will lose the back/forward button, and the ability to link to pages as you navigate. Support for pushState is on the roadmap;

There are other cases that have limited support but will become first-class as we further develop LiveView:

• Transitions and loading states - the LiveView programming model provides a good foundation for transitions and loading states since any UI change done after a user action is undone once the server sends the update for said action. For example, it is relatively straight-forward to click a button that changes itself in a way that is automatically undone when the update arrives. This is especially important as user feedback when latency is involved. A complete feature set for modelling those states is coming in future versions;

• Optimistic UIs - once we add transitions and loading states, many of the building blocks necessary for building optimistic UIs will be part of LiveView, but since optimistic UIs are about doing work on the client while the server is unavailable, complete support for Optimistic UIs cannot be achieved without also writing JavaScript for the cases the server is not available. See "JS Interop and client-controlled DOM" on how to integrate JS hooks;

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 from your router, controller, or view. When a view 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 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 AppWeb.ThermostatLive do
  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.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.

A separate .leex HTML template can also be rendered within your render/1 callback by delegating to an existing Phoenix.View module in your application. For example:

defmodule AppWeb.ThermostatLive do
  use Phoenix.LiveView

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

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

defmodule AppWeb.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, AppWeb.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 AppWeb.Router do
  use Phoenix.Router
  import Phoenix.LiveView.Router

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

You can also live_render from any template:

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

Or you can live_render your view from any controller:

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

  def show(conn, %{"id" => id}) do
    live_render(conn, AppWeb.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, AppWeb.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. I.e. 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", {params: {_csrf_token: csrfToken}});
liveSocket.connect()

Note: Comprehensive JavaScript client usage is covered in a later section.

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: :timer.send_interval(30000, self(), :update)

    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
    {: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 :update in a handle_info just like a 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.

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 most commonly 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:

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

If the @user assign changes, then LiveView will re-render only the @user.id and @user.name and send them to the browser.

The change tracking also works when rendering other templates, as long as they are also .leex templates and as long as all assigns are passed to the child/inner template:

<%= 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.

Change tracking pitfalls

Although change tracking can considerably reduce the amount of data sent over the wire, there are some pitfalls users should be aware of.

First of all, change tracking can only track assigns. So for example, if you do something such as:

<%= @post.the_whole_content %>

If any other field besides the_whole_content in @post changes for any reason, the_whole_content will be sent downstream. Although this is not generally a problem, if you have large fields that you don't want to resend or if you have one field in particular that changes all the time while others do not, you may want to track them as their own assign.

Another limitation of changing tracking is that it does not work across regular function calls. 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>

Note though this concern does not apply to Elixir's constructs, such as if, case, for, and friends. LiveView always knows how to optimize across those.

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 | phx-click, phx-capture-click | phx-target | | Focus/Blur Events | phx-blur, phx-focus, phx-target | | Form Events | phx-change, phx-submit, phx-target, data-phx-error-for, phx-disable-with | | Key Events | phx-keydown, phx-keyup, phx-target | | Rate Limiting | phx-debounce, phx-throttle | | DOM Patching | phx-update | | JS Interop | phx-hook |

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 contain metadata of the client event, containing all literal keys of the event object, such as a click event's clientX, a keydown event's keyCode, etc.

The phx-capture-click event is just like phx-click, but instead of the click event bubbling up to the closest phx-click element, event capturing is used, where the events propagate inwards from the clicked element. This is useful when wanting to bind a click events without receiving bubbled events from child UI elements.

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>

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} ->
      {:stop,
       socket
       |> put_flash(:info, "user created")
       |> redirect(to: Routes.user_path(AppWeb.Endpoint, AppWeb.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 :stop 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 data-phx-error-for attribute. Failing to add the data-phx-error-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 AppWeb.ErrorHelpers may use this function:

def error_tag(form, field) do
  Enum.map(Keyword.get_values(form.errors, field), fn error ->
    content_tag(:span, translate_error(error),
      class: "help-block",
      data: [phx_error_for: input_id(form, field)]
    )
  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.

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. When pushed, the value sent to the server will contain all the client event object's metadata. For example, pressing the Escape key looks like this:

%{
  "altKey" => false, "charCode" => 0, "code" => "Escape",
  "ctrlKey" => false, "key" => "Escape", "keyCode" => 27,
  "location" => 0, "metaKey" => false, "repeat" => false,
  "shiftKey" => false, "which" => 27
}

By default, the bound element will be the event listener, but a window-level binding may be provided via phx-window-keydown, 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", %{"code" => "ArrowUp"}, socket) do
  {:ok, new_temp} = Thermostat.inc_temperature(socket.assigns.id)
  {:noreply, assign(socket, :temperature, new_temp)}
end

def handle_event("update_temp", %{"code" => "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

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 follow events are supported:

• lv:clear-flash – clears the flash when send 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>

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 "child_template", assigns
render SomeOtherView, "child_template", assigns

If the other template has the .leex extension, LiveView change tracking will also work across templates.

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

One option to address this problem is to render a child LiveView inside a parent LiveView by calling live_render/3 instead of render/3 from the LiveView template. This child LiveView runs in a completely 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, using a separate LiveView, each with its own process, 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

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.
• 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.

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 button click to once every second:

<button phx-click="search" phx-throttle="1000">Search</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 usecases 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.

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 navitation 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/2 by Phoenix.LiveView.Helpers.live_patch/2 or Phoenix.LiveView.Helpers.live_redirect/2

• From the server - this is done by replacing redirect/2 calls by push_patch/2 or 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_redirect(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. 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, the existing root LiveView is shutdown, and an Ajax request is made to request the necessary information about the new LiveView without performing a full static render (which reduces latency and improves performance). Once information is retrieved, the new LiveView is mounted. While redirecting, a phx-disconnected class is added to the root LiveView, which can be used to indicate to the user a new page is being loaded.

live_patch/2, live_redirect/2, push_redirect/2, and push_patch/2 only work for LiveViews defined at the router with the live/3 macro.

handle_params/3

The handle_params/3 callback is invoked after mount/3. 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.

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(%{"organization_id" => org_id}, session, socket) do
  # do something with org_id
end

def handle_params(%{"organization_id" => org_id, "sort_by" => sort_by}, url, socket) do
  # do something with org_id and sort_by
end

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

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

def handle_params(%{"sort_by" => sort_by}, url, socket) do
  # 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.

Live Layouts

Your LiveView will be rendered within the layout specified in your Plug pipeline, such as the default app layout. Assigns defined during mount of the root LiveView are accessible in the layout, but the app layout is never updated after the initial render. For a live layout, you must specify an additional layout to use with your LiveView. For example, your regular app.html template may display a @new_message_count notification, like this:

<!DOCTYPE html>
<html lang="en">
  <head>
    <title><%= @page_title %></title>
  </head>
  <body>
    <div>
      <nav>
        ...
        Messages (<%= @new_message_count %>)
      </nav>
      <%= render @view_module, @view_template, assigns %>
    </div>
  </body>
</html>

To allow the @new_message_count to be be updated by your LiveView, you can move the dynamic content inside a sub-layout, such as app_web/templates/layout/live.html.leex.

First, you would update your app.html layout to keep only the barebones HTML structure:

<!DOCTYPE html>
<html lang="en">
  <head>
    <title>...</title>
    <script>...</script>
  </head>
  <body>
    <%= render @view_module, @view_template, assigns %>
  </body>
</html>

Next, define a new live.html.leex layout with the dynamic content, followed by a render of the inner @live_view_module:

<nav>
  ...
  Messages (<%= @new_message_count %>)
</nav>
<%= @live_view_module.render(assigns) %>

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

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

Or alternatively, you can provide the :layout dynamically as an option in mount:

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

  def handle_info({:new_messages, count}, socket) do
    {:noreply, assign(socket, new_message_count: count)}
  end
end

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

Updating the HTML document title

Because the main 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 app layout:

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

Now, although the app 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.

Disconnecting all instances of a given live user

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_sockets:#{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:

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

It is the same mechanism provided by Phoenix.Socket, so you can use the same approach to disconnect live users and regular channels.

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 liveSocket = new LiveSocket("/live", Socket)
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.

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, a "_target" param will be 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" => %{"name" => "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-loading" class

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

1. The submitted form is reactivated and loses the "phx-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 form submissions, any HTML tag can be annotated with phx-disable-with, which swaps the element's innerText with the provided value during form 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>

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. 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.

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})
...

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

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)