Patch 1

[bmschmidt]

See #7

[rreusser]

What would you think instead of

Suggested change

	: 1. - 4.0 * pow(1. - t, 3.0);
	: 1. - 4.0 * pow(abs(1. - t), 3.0);

or my favorite solution, I think,

float cubicInOut(float t) {
  float tm1 = t - 1.0;
  return t < 0.5
    ? 4.0 * t * t * t
    : 1.0 + 4.0 * tm1 * tm1 * tm1;
}

For the reasons mentioned in #7 (comment), namely that even though people shouldn't plug in t < 0 or t > 1, they certainly will, perhaps won't notice they've done it, and it's at least nice if the returned value isn't formally undefined and platform-dependent (which, in my experience usually gets caught years later when a developer gets a new computer, runs the same code, and suddenly things look all wrong).

[bmschmidt]

Thanks for keeping this library up. That makes sense.

I like the tm1 formula with no pows at all better, too. But I like '1-t' instead of 't - 1' because the formulas can stay the same whether it's cubic or quartic, rather than flipping the sign in the final addition. Commits below do it with an explicit if block to avoid an unnecessary subtraction when t < .5.

[bmschmidt]

This also fixes a problem where the visual tests didn't run because the frag shader was misnamed in the code.

[rreusser]

Really nitpicking here, just for the fifth powers. Stack overflow says maybe pow(abs(x), 5.0) is appropriate so that you can let the glsl compiler do the optimization if that's a thing it feels is necessary, but I truly don't think it matters as long as the result is correct, so that I support this PR no matter what. Thanks for working on this!

[rreusser]

Suggested change

	return t*t*t*t*t;
	float t2 = t * t;
	return t2 * t2 * t;

[rreusser]

Suggested change

	return 1.0 - 16.0 * u * u * u * u * u;
	float u2 = u * u;
	return 1.0 - 16.0 * u2 * u2 * u;

[rreusser]

Suggested change

	return 16.0 * t * t * t * t * t;
	float t2 = t * t;
	return 16.0 * t2 * t2 * t;