Live view renders happen twice. There is the first static render that happens as part of an HTTP GET request. The conn is evaluated, a socket built / mounted, the page rendered. After that though the socket goes away entirely, it was just there to address the HTTP get request, there is no persistent connection.
Then the javascript on the page runs, connects via websocket, and does a second render. This time the socket sticks around because you’ve got a persistent connection. Notably though on this run, there is no @conn. In fact it’s a totally separate request, that could even happen to a totally different server from the GET request.
In the GET request, the assigns can be copied into the socket, because the socket is just basically exists to help you avoid writing totally different code for the static render. BUT on the second request, there’s no @conn for the values to be copied from at all. This is why the socket |> assign_new function takes an anonymous function arg. On the GET request assign_new will just use the assign from the @conn. But on the second live request, how can it get its value? It runs the anonymous function.
