taitbryan.axangle2euler obviously doesn't work (or I missed something obvious)

[petaflot]

here is a bunch of values (points on a half-circle) for the vector (vec), theta rotation around that vector is always zero, norm is always 1

vec 1.0 0.0 0.0
ax = (-0.0, 0.0, -0.0)

vec 0.9510565162951535 0.3090169943749474 0.0
ax = (-0.0, 0.0, -0.0)

vec 0.8090169943749475 0.5877852522924731 0.0
ax = (-0.0, 0.0, -0.0)

vec 0.5877852522924731 0.8090169943749475 0.0
ax = (-0.0, 0.0, -0.0)

vec 0.30901699437494745 0.9510565162951535 0.0
ax = (-0.0, 0.0, -0.0)

vec 6.123233995736766e-17 1.0 0.0
ax = (-0.0, 0.0, -0.0)

I obviously expect something more useful to come out of axangle2euler( vec, 0 )

[matthew-brett]

Could you say more about what you're expecting? Here's me running what I think you're running:

transforms3d.taitbryan.axangle2euler([0.30901699437494745, 0.9510565162951535, 0.0], 0)

which gives (-0.0, 0.0, -0.0). That seems right to me - rotating zero radians around any axis will give zero rotations in Euler angles. Did I misunderstand your question?

[petaflot]

the axii itself is rotated. does this make sense?

[matthew-brett]

What do you mean by "the axis itself is rotated"? That it should be rotated? In which case, no, the meaning there is that the rotation is of 0 degrees around the axis - so the axis is static.

[petaflot]

consider a vector of length 1

this vector, rotated around alpha-beta-gamma Euler angles is a point in space (with orientation relative to the horizon). I am looking for a way to express with vector not in Euler angles, but Tait-Bryan angles. maybe this picture helps?

[matthew-brett]

Sorry - I think I'm still not following you - you have a vector $\vec{p}$, and you want to express that vector as Euler angles (in Tait-Bryan convention), where the angles are the rotation from the unit vector in x (1, 0, 0) to $\vec{p}$?

[petaflot]

I don't follow you either, so I'll try to rephrase.

In your lib, I was expecting to find a taitbryan2euler() function. Like... I don't even want to have to use an [x,y,z] vector.

[petaflot]

as I understand, the sequence-order in which the rotations are applied changes massively. it seems Wikipedia refers to it as gun-order angles¹ ; I chose to use this because it seems to me the most intuitive way to represent a quaternion for a human being walking on a "flat" surface (as per my drawing, echo and phi are equal to zero), and eventually throwing a stone at something (phi = 0).

Does this make more sense?

Footnotes

1. https://en.wikipedia.org/wiki/Euler_angles#Others ↩

[matthew-brett]

Just to check we agree - Euler angles are a generic term for triplets of rotation angles. Tait-Bryan angles are just Euler angles using a particular convention for the axes. You said you have a point, p, I've called that $\vec{p}$. You want the Tait-Bryan (=Euler in a particular convention) angles to get from - somewhere - to p. Where is somewhere? I was assuming it was just the unit axis in x.

[petaflot]

somewhere is the origin, facing towards infinity on x axis (north as per my drawing), orientation parallel to the horizon

[matthew-brett]

Right - and given the length of the vector is irrelevant to the rotation, you do in fact want the angles to rotate the unit vector in x to the vector $\vec{p}$ as I had guessed. Does this help?

[petaflot]

apparently you edited your first reply, which read

Did you misunderstand your question?

possibly :S

Does this help?

it just confirms that going by trial-and-error is exactly what I wanted to avoid

[matthew-brett]

The algorithm proposed was:

 rotx = Math.atan2( y, z );
 if (z >= 0) {
    roty = -Math.atan2( x * Math.cos(rotx), z );
 }else{
    roty = Math.atan2( x * Math.cos(rotx), -z );
 }

That's not a guess - but do you mean - you are not sure it's correct? I can check it more thoroughly if you like.

[petaflot]

I can check it more thoroughly if you like.

please, don't do this yet. see moble/quaternion#218