omnimessage_lustre
Seamless server communication using Lustre messages
A main challenge of maintaining a client side application is state synchronization. Usually, a complete separation is ideal, where client state is managed exclusively at the client, and server state exclusively at the server. There’s nuance to this take, but it’s a good general rule to follow.
Lustre exceles at this with having both the power of a full-blown single page application, and the flexibility of using server components for a LiveView/HTMX like solution to state fully owned by the server.
This approach becomes brittle when the state is shared between both the client
and server. In a chat app, for example, the messages are state owned by the
server, but you still need to display a message as Sending...
before the
server knows it exists. This is usually where LiveView or HTMLX will recommend
you sprinkle some Javascript. Lustre has an easier time dealing with this, but
it can become quite cumbersome to manage the different HTTP requests and
websocket connections.
This is where Omnimessage comes in. When paired with a properly set up server (see omnimessage_server), all you need to do to communicate is dispatch messages in Lustre:
import omnimessage_lustre as omniclient
pub fn chat_component() {
// Instead of lustre.component, use:
omniclient.component(
init,
update,
view,
dict.new(),
// this is for encoding/decoding, supplied by you
encoder_decoder,
// this transfers the encoded messages
transports.websocket("http://localhost:8000/omni-app-ws"),
TransportState,
)
}
// Divide you messages wisely:
pub type Msg {
UserSendDraft
UserUpdateDraftMessageContent(content: String)
// Messages from the client
ClientMessage(ClientMessage)
// Messages from the server
ServerMessage(ServerMessage)
// Messages about transport health
TransportState(transports.TransportState(json.DecodeError))
}
fn update(model: Model, msg: Msg) -> #(Model, effect.Effect(Msg)) {
case msg {
// Merge strategy
ServerMessage(shared.ServerUpsertMessages(server_messages)) -> {
let messages =
model.messages
// Omnimessage shines when you're OK with server being source of truth
|> dict.merge(server_messages)
#(Model(..model, messages:), effect.none())
}
// ...handle the rest of the messages
}
}
// Then in your view, all you need to do is:
html.form()
|> event.on_submit(UserSendChat)
// That message will go to both the client, that can use it to disaply the chat
// in a sending state, and to the server, which can handle the new message
// and reply with an updated, correct state.
Further documentation can be found at https://hexdocs.pm/omnimessage_lustre.