Introductory docs on event loop

[jmaargh]

Fixes #2736

• [x] Tested on all platforms changed
• [x] Added an entry to CHANGELOG.md if knowledge of this change could be valuable to users
• [x] Updated documentation to reflect any user-facing changes, including notes of platform-specific behavior

I ended up changing more text than I'd originally planned at the root level of the docs. Happy to make large changes/reversions if the maintainers would prefer - but to me this reads as a much better "what I would have wanted to know before starting" than what there was.

I also took the opportunity to fix up some docs I found along the way which I thought could use some love, specifically:

• map_user_event()
• Event::DeviceEvent

[jmaargh]

Anybody keen to review this? I'd love to get something merged while I have some time on my hands

[jmaargh]

@madsmtm @daxpedda Thanks both for the reviews, very helpful comments.

I think I've addressed everything (either with changes or comments). I haven't changed the diagram at all, but can do if still desired.

[madsmtm]

Thanks for the changes!

I think @kchibisov is working on changing how things work in https://github.com/rust-windowing/winit/pull/2896, so I'd like to hear from them and @rib before moving forwards (let's give them a week or so).

[rib]

There certainly are some changes in the pipeline that are likely to affect this but, so long as we feel like this isn't immutable then maybe it's ok to land and we can update it when things change.

I just tried opening the diagram in https://app.diagrams.net/ and although I guess it's a proprietary web app it seems simple enough to use.

Main changes I'm hoping to see though would be:

RedrawEventsCleared can probably just be removed, since the concept doesn't really make sense imho if we no longer guarantee an explicit ordering for when we emit redraw events relative to other 'main' events.

With the direction discussed for the pump_events/run_ondemand changes then MainEventsCleared is becoming more like an AboutToWait - i.e. it's more like a RunLoop observer event for tracking wake up and wait transitions, and should probably get renamed.

LoopDestroyed should probably be renamed to LoopExiting as was discussed in relation with the run_ondemand changes.

[rib]

I'd also like to reiterate the thanks for looking at this @jmaargh, having some visual docs for this seems like it would be really nice to have, sorry for dropping the ball with giving feedback after the initial issue discussion

[kchibisov]

I'm sort of concerned that the diagram isn't really accurate, given that we have backends delivering RedrawRequested regardless, we really can't guarantee the ordering like we mention for some backends, not that we really want to.

I'm also not even sure that we'll keep MainEventsCleared, it seems like we do don't guarantee when the RedrawRequested arrives, we don't really need it, and AboutToPoll should be enhough.

[daxpedda]

I just tried opening the diagram in https://app.diagrams.net/ and although I guess it's a proprietary web app it seems simple enough to use.

I think, if it's not too much work for you @jmaargh, we should shift to Mermaid, which has native support in GitHub, Rustdoc integration and MdBook integration.

It's completely open source and has an online editor that's fairly easy to use.

[jmaargh]

I'm sort of concerned that the diagram isn't really accurate, given that we have backends delivering RedrawRequested regardless, we really can't guarantee the ordering like we mention for some backends, not that we really want to.

I'm afraid I don't understand how this makes the diagram (or the text) inaccurate? If RedrawRequested is always delivered on a platform, then that is covered by the "platform ... requests it" text. All the diagram says is that RedrawRequested, when it occurs, will occur after MainEventsCleared. Could you explain a bit why you feel these docs need to change?

[jmaargh]

I think, if it's not too much work for you @jmaargh, we should shift to Mermaid, which has native support in GitHub, Rustdoc integration and MdBook integration.

No worries, I'll get that done soon.

[jmaargh]

@rib As the behaviour changes, I'm very happy for this documentation to change with it. I'd be very happy to be involved with making those updates if that's helpful (unfortunately I can't guarantee an SLA though!)

Unless there's a definite major change to the API/behaviour that's imminent, my preference would be to get this merged as describing the current behaviour. The docs can always (and should) change as the API does.