Validate position DOMPointInit

[Manishearth]

Given that it's a float, comparison with 1 may not be the best idea, but I feel like erroring here is marginally better than ignoring invalid w values.

[klausw]

I think people have a somewhat irrational (no pun intended) distrust of floating point arithmetic. If you start out with w=1, and apply matrix transforms involving add/multiply of values that are also exactly 0 or 1 (as is the case for the bottom row of a rigid transform matrix), the result will still exactly equal 1.0. Floating point arithmetic is exact for integers as long as you don't exceed the mantissa range (up to 2^52 or so for doubles), and the typical things you'd be doing with w values wouldn't lead to an approximately-but-not-exactly-1.0 value.

[Manishearth]

Yeah, this was the assumption I was making too, floating point should be fine for ints here. And the fix for this error is straightforward, you can always set the last value to 1 if you need.

[toji]

Hm... so this would explicitly prevent me from doing something along the lines of the following, right?

let transform = new XRRigidTransform({x: 1, y: 2, z: 3}, someOrientation);

Because that actually seems pretty convenient.

[Manishearth]

It would allow it as long as someOrientation has a unit value 1 -- if it doesn't you have a bug anyway since it's going to get overwritten

[Manishearth]

Oh, sorry, position. That is a problem, hmm. Maybe provide our own init dicts?

[Manishearth]

It seems like w=1 by default, so that looks like it would still work?

[toji]

Oh really? Could you link to the relevant spec text? (I'm gonna try and hunt it down too.)

[toji]

I guess this is the right link to cite: https://www.w3.org/TR/geometry-1/#DOMPoint

And in light of that, I'm good with this being a strict validation. Thanks!