Lavash.Actions.Run (Lavash v0.4.0-rc.4)

Copy Markdown View Source

Post-cascade imperative escape hatch for actions.

Runs AFTER the reactive cascade has settled. Takes a socket, returns a socket. The user can do anything — write state, push events, call LV stream/upload ops, hit external services, log — and the returned socket replaces the current one wholesale.

When to use

run fn socket ->
  # Read post-cascade derived values
  new_msg = %{id: next_id(), summary: socket.assigns.summary}

  # Do socket-level LV ops
  Phoenix.LiveView.stream_insert(socket, :messages, new_msg)
end

Use run for:

  • Socket-level LV ops that aren't pure state writes: stream_insert/4, allow_upload/3, consume_uploaded_entries/3, cancel_upload/3
  • Side effects that need the post-cascade view (logging derived values, sending notifications with computed content)
  • Anything that doesn't fit the declarative ops but needs consistent state

Trade-off: no re-cascade

A run body MAY mutate socket.assigns (via put_state, assign/3, or LV ops that write to the socket). Those writes land — Phoenix's render diff machinery picks them up via __changed__ so the wire diff is correct.

But because run is post-cascade, lavash does NOT re-fire the reactive cascade after the body. Calcs depending on what run wrote will be stale until the next user event.

If you need a derived value of a write, write it in pre_run instead (pre-cascade; cascade then settles the derived values; downstream runs read the result).

Fields

  • :fun - quoted AST of the fn socket -> socket end body. Compiled into a __lavash_run__/3 clause on the user's module.