I just started working with a scene-centric coordinate system (z = skywards;
xy is level plane; see the following camera) and found that small horizontal
motions of look_at cause the camera to roll over. A look through the docs
turns up the 'sky' vector; if I put  sky <0,0,1> ahead of the look_at, it works
fine.  My suggestion is that the default 'sky' should point in the same direction
as 'up', rather than in 'y' direction. The set of old scenes which would be
broken by this change is likely very small (those which point the 'up' vector
somewhere other than +y, and don't set a 'sky', but use look_at).

camera {
  location  <0.0, -4, -0.5>
  direction 1.5*y        // initially looking along y axis...
  up 1*z
  right     x*image_width/image_height
  look_at   <0.0, 0.0,  0.0> // turns scene upside down!!
}

This is also a right-handed coordinate system. I find the 'scene-centric' coordinates
to
be far more natural, especially when there is no specific view direction; this occurs,
for
instance, in animation with a moving camera, or when the same scene will be shot
from different places; or even when creating a sub-object and you don't know which
way it will be viewed when it is done.

P.S. love the new features, especially sphere_sweep (great for plumbing), photons,
and the parametric surfaces and u-v image map features.

- Greg

Post a reply to this message

Firstly, we sure could use a beta suggestion group, couldn't we?  :-)  This
is off-topic here as long as things are geared toward problems and bugs in
the programming.

I think this could be questionable because of the expected default of up y
happening to be the POV-Ray way of things, and if you didn't use a pure
single axis for any other up direction it might get more complex than just
accomodating the flipping via script workarounds.  Take for example up
<1,0,1> or up <0.1,0,0.9> or any other non-singular axes.  sky is just
trying to remain at the default instead of adhere to all sorts of other
possibilites unless you specifically tell it otherwise (the point of having
sky).  Although it could be argued how up and right would not typically be
non-perpendicular in the first place, but sky is supposed to be the variable
for "up", not a duplicate of 'up'.  Don't know if I'm being logical or
rational about it, just my limited observation.  The actual workings of
camera aren't extremely well-known to me.
I could imagine there might be some way of automating the different
coordinate systems but likely more to it than sky alone involved.



bob h

"Greg Smith" <gsm### [at] alumniuwaterlooca> wrote in message
news:3C7C3BFA.87E7B214@alumni.uwaterloo.ca...
> I just started working with a scene-centric coordinate system (z =
skywards;
> xy is level plane; see the following camera) and found that small
horizontal
> motions of look_at cause the camera to roll over. A look through the docs
> turns up the 'sky' vector; if I put  sky <0,0,1> ahead of the look_at, it
works
> fine.  My suggestion is that the default 'sky' should point in the same
direction
> as 'up', rather than in 'y' direction. The set of old scenes which would
be
> broken by this change is likely very small (those which point the 'up'
vector
> somewhere other than +y, and don't set a 'sky', but use look_at).

Post a reply to this message

On Tue, 26 Feb 2002 21:58:25 -0600, "bob h" <omn### [at] charternet>
wrote:

>Firstly, we sure could use a beta suggestion group, couldn't we?  :-)  This
>is off-topic here as long as things are geared toward problems and bugs in
>the programming.

If there were a povray.beta-test.feature-request group, it would most
probably just be a symlink to /dev/null :)


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message