Event::MouseMotion underreports movement(?)

[kklibo]

@asny
Observed mouse movement through Event::MouseMotion {delta} (as in examples/texture) seems to be underreporting pixel delta, especially when the mouse is moved quickly.

This effect isn't observable in the demos because it's hard to tell how much rotation should map to a mouse movement. I'm implementing a click-and-drag view interface with an orthographic projection, in which the rendered images should have a fixed position to the cursor as it's dragged across the screen. This does work occasionally, but usually it appears that the delta reported with the MouseMotion event is smaller than it should be, and the graphics don't move far enough to keep up with the cursor, This effect seems to occur more with fast mouse movements, and more in the web version than in the desktop version.

I'm trying to find a mouse motion handling technique that will demonstrate the expected behavior, either just to prove the idea of the underreported motion or as a permanent workaround. Is there an easy way to get cursor position, instead of delta, at the time that a MouseMotion event is handled?

[asny]

Ok that is really weird. On web I just forward the movementX and-Y from the mousemove event, so maybe I somehow skip some events, I have to look into that. Or do you happen to be using touch input? In that case I compute the delta myself from current and previous position so there could be some inaccuracies there. Fortunately, I can easily get the actual position of the mouse, so I will expose that along with the delta for the mouse move event as well as investigate the accuracy of the delta.

[kklibo]

Sure thing, I'll work on it today.

[asny]

Tanks a lot 🙂

[kklibo]

I successfully integrated the new changes into my project. As of three-d 2105acc:

• The aspect ratio issue is resolved in web + desktop (although I haven't tested a resizable web canvas yet)
• Position-based MouseMove event handling works correctly in web + desktop. I tested with window resizes, different aspect ratios, and different zoom levels (with my own project's zoom code handling the orthographic projection resizing, not the new Camera::zoom code that was added).
  Everything looks good.

If the delta-based mouse movement system isn't needed anymore, it might be easier to just remove it.

[asny]

Sounds really great, thanks for testing it! I will release 0.5.0 really soon then.
I will keep the delta, it is very convenient, but either it should be made more precise using the position or there should be a warning that it is not precise.

[asny]

For anyone interested, I made the delta precise by calculating it myself based on the position.