CursorState::Grab does not report mouse delta on macOS

[mystal]

You still get MouseMoved events with the absolute position of the mouse, but it does not change. As a result, it's kind of useless. According to macOS documentation, you should instead check for deltaX and deltaY on the NSEvent.

As far as I can tell, winit doesn't currently report the mouse cursor delta in any events. I think it makes sense to add the mouse delta to the MouseMoved event. Any thoughts?

[Ralith]

#164 introduced API for raw input events, which is more appropriate for things like first-person camera control. Would it make more sense to extend the macos backend with support generating such events? It seems to me that you generally want either high-resolution unfiltered relative mouse movement data or filtered window-space pointer positions, not filtered window-space relative pointer movement, and not both at once.

[LPGhatguy]

@Ralith Will we need a new DeviceEvent variant for mouse movement based on the API you introduced? I see that it exposes an axis and an f64 value, which might not be suitable for platforms like Windows where raw input is reported as one event from the OS

[Ralith]

As discussed on IRC, the intention of DeviceEvent is explicitly that multiple events be generated if a host even contains motion on multiple axes. For example, an X11 device event may represent motion on arbitrarily many axes at once, but instead of introducing a dynamically allocated vector, we just generate N separate motion events.

edit for posterity: @LPGhatguy has convinced me that it's worth having input device class specific event types for the sake of API usability.

[swiftcoder]

This is complete, right? DeviceEvent::MouseMotion has existed for some time now.