Rework Event System

[ealmloff]

There are three main goals here:

• Access to native event types (Access to raw web_sys::Event #793)
• Binary size improvements (Improve Event System Code Snipping #1163)
• Lazily convert events for performance

This is built on top of the better pointer events introduced in #1394

closes #1163
closes #93

[esimkowitz]

This is freaking awesome! Curious for my own learning: does this place an additional burden on code that clones/copies an event to allocate additional space on the heap? I'm assuming the Box can't be reused by copied instances right?

[esimkowitz]

Or building on that, does this place an additional burden on the transmission system to allocate additional space for every event?

[ealmloff]

This is freaking awesome! Curious for my own learning: does this place an additional burden on code that clones/copies an event to allocate additional space on the heap? I'm assuming the Box can't be reused by copied instances right?

The virtual dom will receive one non-specific web event. That event is wrapped in an Rc so that it will not be cloned. The virtual dom will bubble that event to event handlers. Each event handler will downcast it to the specific event that is needed. Downcasting work will be duplicated if the event bubbles, but it should be fairly cheap. Downcasting in the event handler is what lets us do bundle splitting better.

Once the event is downcast, calling any methods on it will get the needed data from it. For example if you call modifiers, it will build a modifiers struct every time instead of caching it. This is slightly inefficient, but because the old implementation used a depreciated old store of data we were doing this anyway.

Or building on that, does this place an additional burden on the transmission system to allocate additional space for every event?

On the web renderer, we should allocate less in general because we don't get all of the data eagerly when we receive an event. For native/liveview events, we just use the old serialized form of data so it should be the same.

[sdwoodbury]

this branch needs the tray feature. it was added to the fix-event-bubbling branch but not this one.

[ealmloff]

this branch needs the tray feature. it was added to the fix-event-bubbling branch but not this one.

The branch was a few commits behind the main branch. I just pulled in the latest changes. (this happens automatically when the PR is merged)

[esimkowitz]

@ealmloff are you holding this for the 0.5 release? May be good to add a milestone tag for it

[ealmloff]

@ealmloff are you holding this for the 0.5 release? May be good to add a milestone tag for it

Yes, everything breaking is for 0.5