Hey everyone. I’m working on a service that’s currently LV but will eventually have an app. The interaction will basically be the same. This has had me thinking about how I could avoid duplication with client/server interactions, while still offering a similar responsive/reactive experince on both platforms.
I’d like to avoid having to build a separate API (beyond bare necessities) and LV stack if possible so my question is: can I treat a channel as an ‘API’ interface that the LV and the App could both talk to and does anyone have any experience of this, advice/war stories they can share?
By definition, what liveview does is interactive frontend without writing JS, or having a communication contract.
If you want to change this behavior, you will literally lose all the benefits liveview brings, and IMO is better to stick with a separate frontend with a separate API.
I’m not talking about losing the behaviour, I’m talking about creating a consistent channel interface for sending events between client and server. LV is an abstraction on top of channels already.
I would not classify liveview as an abstraction over channels, I would say that liveview uses phoenix channels to implement the desired behavior, namely abstraction of frontend interaction.
I don’t know these days, but back in the day liveview implementation was marked as internal and it is not recommended to tap into how it works, because implementation details might change in a future release.
No I get that. LV is obviously sending its diffs and such and I wouldn’t want to touch it. I’m just more wondering about something like a separte channel for key persistent state mutations or similar. So you can centralise a channel api between clients for things like e.g add-blah, blah-updated. Maybe more like a shared CQRS or similar? I’m unsure. It just seems like duplication to define how these events are handled twice (once on the channel and once on the LVs).
It just seems like duplication to define how these events are handled twice (once on the channel and once on the LVs).
That sounds rather wrong tbh. Event handlers in LV should handle the web ui specifics, not domain level functionality. You shouldn’t need to duplicate any logic to another API interface. That stuff should be in lower levels of your codebase. Copying boilerplate (e.g. calling into your core modules) is imo the right way to handle this, as eventual divergance of interfaces is to be expected.
