Gilles Tran <tra### [at] inapginrafr> wrote:
> Uh? I'm not sure I understand why this is a "plague"... The stretch/squeeze
> issue is such a common headache for newcomers that it made its way into FAQ
> 9.1.2.4, which is the primary source of the image_width/image_height trick.

  The problem I see with this "trick" is that it's unpredictable.

  Even though a perfect model of the scene models everything which could
be viewed from the camera location (looking at any direction), it's quite
common that to save work people just model the visible part of the
scene (visible with the current camera location and look_at).
  What happens if he (or worse, someone else) decides to render the image
with a different aspect ratio? The image doesn't get squeezed, but at the
cost of either extra scene (which may not be modelled at all) coming into
view or part of the original scene going out of view (which may leave
important details out of the image).
  Both of these phenomena may be as bad as a squeezed image because they
can potentially destroy the look of the image by introducing modelling
artifacts (lack of proper modelling of the scene) or by leaving out important
details of the image.

  Another problem is that the automatic control of aspect ratio can only
be done by adding/removing extra image vertically. (Making image_width
and image_height change the 'up' vector does not help: The elongation of
the image is still done vertically.)
  One could think "my scene is ok regardless of how much the viewing plane
is elongated horizontally... but why is 'right x*image_width/image_height'
doing it the wrong way? What should I do?".
  The only way to control the aspect ratio horizontally would be to make
the 'angle' change as well. Are you willing to guess the correct formula?
  Scenes are quite often designed so that they look good even if a
"widescreen shot" is done of them. It's not unusual that modelling
extra scene vertically is skipped.

  Even though this trick may sound pretty handy at first, personally I'm
not completely happy with it.
  IMHO this trick should not be freely preached without a good warning
about its (often surprising) side-effects.

-- 
#macro N(D)#if(D>99)cylinder{M()#local D=div(D,104);M().5,2pigment{rgb M()}}
N(D)#end#end#macro M()<mod(D,13)-6mod(div(D,13)8)-3,10>#end blob{
N(11117333955)N(4254934330)N(3900569407)N(7382340)N(3358)N(970)}//  - Warp -

Post a reply to this message