lerp interpolates the wrong way?

[gilgamec]

I guess I've never had occasion to use it, but I just noticed that lerp (in Additive) works the opposite way of any other lerp implementation I've seen. (A brief Googling supports this; see C++, GLSL, Unity, Processing.) I'd expect that

lerp t a b = a + t * (b - a);

that is, that lerp 0 a b == a and lerp 1 a b == b, and that is the case in all other implementations. However, the implementation in linear is the opposite of this, i.e. lerp 0 a b == b and lerp 1 a b == a.

Is this mismatch intentional? If so, can it be clearly documented? If not, might it be fixed? I understand that it's longstanding at this point, but maybe something can be done.

[RyanGlScott]

I agree that this is a bit counterintuitive. Moreover, the slerp function uses the opposite convention, where slerp a b 0 = a and slerp a b 1 = b.

Unfortunately, the lerp function is pretty widely used, so I'd feel a bit remorseful about changing the convention all of a sudden. I'd say the best thing to do is to just document this.

[gilgamec]

I'd say the implementation is more than just counterintuitive, it's downright wrong. Like I said, every version I've ever used or could find does it the other way around; so if even the other interpolation function from linear does it that way, I'd say it's clearly a bug and should at least be documented as such.

(I note both functions were written by @ekmett; maybe he has a higher purpose, but I'd guess the lerp implementation is just a longstanding typo.)

[mrehayden1]

Agreed. Can this not be made much clearer in the documentation?

I had some very jittery animations until I figured this out. It defies all my intuition to defined it this way!

[ekmett]

This was a complete brainfart on my part. I'm surprised it has stood this long.

[ekmett]

I'd be open to a major version bump and clear CHANGELOG entry and fixing this.

[RyanGlScott]

I've uploaded linear-1.23 to Hackage with a fix.