With the help of ChatGPT 3.5, the documentation of Elixir/Phoenix, and lots of
IO.inspect and trial-and-error, I finally figured out the basics of how a Phoenix LiveComponent works in combination with a LiveView.
Starting from the basic scaffolding that
phx.gen.live created, I ripped out the helper functional component(s) that rendered a list of items in a table, and refactored every list item (a Linux kernel ChangeLog) into its own LiveComponent.
Each such LiveComponent thus has
handle_event functions to deal with clicks on its buttons, such as to handle the “process” click that splits a ChangeLog into separate commits and persists them in an SQLite database through Ecto.
As soon as something changes, the parent LiveView only re-renders the LiveComponent from which the change came.
And, based on the state of entity being rendered (here, a ChangeLog), the render function of the LiveComponent shows different buttons leading to different actions, such as “wipe” to keep the ChangeLog and wipe all its related Commits,
I kept the
handle_event of “delete” (to delete the ChangeLog entirely) on the LiveView, however, to benefit from stream_delete, when the collection of ChangeLogs will get bigger.
I will need to change the sync changelog-to-commits processing into async processing, because some ChangeLogs are 15+ MB large with thousands of commits, and thus the websocket could time-out while waiting for the processing to finish.
The web app will be deployed on a potato-grade server with an anemic AMD E2-9000e 6W-TDP APU that doesn’t even break the 1000 CPU Mark score , so this is a neat little project for me to learn a lot.