t_quat::set: return a zero rotation instead of a zero quaternion #987
Conversation
It could be argued that the default constructor for quaternion should
return a zero rotation i.e. {0, 0, 0, 1}, but I decided not to go
there.
|
Is this solving a particular problem? What happens with the unmodified code? The default vsg::quat should probably match what glm does but just had a look at glm and left wondering WTF. So many headers for such basic functionality, not at all clear what is going on from a quick glance. Ahhh one of the main reasons I opted to implement maths classes part of the VSG, code base has got to be readable to be maintainable. |
|
I been thinking about it, I'm leaning toward making zero rotation the default rather than 0,0,0,0 "invalid" rotation that we have right now. Still a bit torn though... |
|
Do not be torn. Please initialize it as zero-rotation. I would have felt
surprised it were not a unit-quaternion.
[0, 0, 0, 1] - 0 rotation valid quaternion.
[0, 0, 0, 0] - NOT a quaternion
[nan, nan, nan, nan] - Could be a quaternion, but atleast things break
immediately.
…On Wed, Oct 4, 2023 at 1:23 PM Robert Osfield ***@***.***> wrote:
I been thinking about it, I'm leaning toward making zero rotation the
default rather than 0,0,0,0 "invalid" rotation that we have right now.
Still a bit torn though...
—
Reply to this email directly, view it on GitHub
<#987 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AHWFHIO47DRM7VQ7GUYHNVDX5WLRJAVCNFSM6AAAAAA5SZOFDOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONBXGMZTKNBRHA>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
--
S <https://www.nodein.com/>uresh K. Kannan, Ph.D.,
Chief Scientist, Nodein
https://www.nodein.com
|
It's fixing the case of quat(0.0f, vec3(0.0f , 0.0f, 0.0f)) to return a 0 rotation. While this could be sensibly specified as undefined, Open Scene Graph set it to a zero rotation. Actually, Quat() in OSG results in (0, 0, 0, 1), so maybe that's the answer here too.
I don't care about glm 😄 , but OSG compatibility might be worth considering. |
I don't agree that [0,0,0,0] is not a valid quaternion. It's certainly not a rotation, but it can meaningfully be used in quaternion multiplication. There's a performance cost in having these primitive math types have non-trivial default constructors, but I suppose that we are over that. |
|
Agreed that the zero element is a valid quaternion. As long as it is clear
that when a Quat is constructed it does not necessarily have magnitude one
and hence not a quaternion that represents a rotation.
…On Wed, Oct 4, 2023 at 7:20 PM Tim Moore ***@***.***> wrote:
Do not be torn. Please initialize it as zero-rotation. I would have felt
surprised it were not a unit-quaternion. [0, 0, 0, 1] - 0 rotation valid
quaternion. [0, 0, 0, 0] - NOT a quaternion [nan, nan, nan, nan] - Could be
a quaternion, but atleast things break immediately.
… <#m_34790264754041740_>
On Wed, Oct 4, 2023 at 1:23 PM Robert Osfield *@*.*> wrote: I been
thinking about it, I'm leaning toward making zero rotation the default
rather than 0,0,0,0 "invalid" rotation that we have right now. Still a bit
torn though... — Reply to this email directly, view it on GitHub <#987
(comment)
<#987 (comment)>>,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AHWFHIO47DRM7VQ7GUYHNVDX5WLRJAVCNFSM6AAAAAA5SZOFDOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONBXGMZTKNBRHA
<https://github.com/notifications/unsubscribe-auth/AHWFHIO47DRM7VQ7GUYHNVDX5WLRJAVCNFSM6AAAAAA5SZOFDOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONBXGMZTKNBRHA>
. You are receiving this because you are subscribed to this thread.Message
ID: @.*>
-- S https://www.nodein.com/uresh K. Kannan, Ph.D., Chief Scientist,
Nodein https://www.nodein.com
I don't agree that [0,0,0,0] is not a valid quaternion. It's certainly not
a rotation, but it can meaningfully be used in quaternion multiplication.
There's a performance cost in having these primitive math types have
non-trivial default constructors, but I suppose that we are over that.
—
Reply to this email directly, view it on GitHub
<#987 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AHWFHIIUHPQPFW7SATOINXLX5XVMZAVCNFSM6AAAAAA5SZOFDOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONBXG43TQMJWGQ>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
It could be argued that the default constructor for quaternion should return a zero rotation i.e. {0, 0, 0, 1}, but I decided not to go there.