CS465 Program 4: Modeller

• assignment handout | framework code (v. 3)

Usage Information

• Camera Motion: If the perspective camera is selected, alt-dragging in an orthographic view will move the camera's look-at point.
• Camera Motion: If the perspective camera is selected, dragging in an orthographic view will move the camera's eye point.
• General Editing: If a light is selected, dragging in an orthographic view will move the light.
• General Editing: Double clicking on any primitive in the hierarchy will track and dolly the perspective view so that the look at point centers on the object.
• General Editing: Double clicking on any view will maximize that view. Double clicking again will return it to its regular size.
• General Editing: There is an OpenGL enforced 8 light limit.
• File IO: You can save and load scenes for easier testing. The writte file is in ASCII format. Be careful if you hand edit it, as the resulting scene might not be readable by the program afterwards (ie - if you break some assumptions about how many lights exist, where the perspective camera is in the hierarchy, etc.).
• Spline Editing: In the spline editor, ctrl-dragging will track the view.
• Spline Editing: In the spline editor, alt-dragging will zoom the view.
• Spline Editing: In the spline editor, pressing spacebar will cycle through different debug views. These will do things like display the normals, etc.
• Spline Editing: Right clicking on the control points between two segments will disconnect the tangents at those two segments. Right clicking again will reconnect the tangents.
• Spline Editing: Shift-clicking on the polygon will add new points.
• Spline Editing: Shift-clicking on the control points between two segments will join those two segments together, effectively deleting a BezierSegment.

FAQ

• 12:30 pm, 15 November 2004: I'm observing some strange behavior with my rotation transformations. They are doing things like blowing up primitives when the length of the axis is greater than 1, or completely removing them in some cases.
  A: The Matrix4f#setRotation(AxisAngle4f) method is the one used to set the underlying Matrix4f object inside of the Transformation. This is incorrect, the framework code should have been using the Matrix4f#set(AxisAngle4f) method. As an acceptable workaround, try using GL#glRotate(float angle, float x, float y, float z) in the Transformation#apply(GL) method. The posted code does not have this problem anymore.
• 12:30 pm, 15 November 2004: I think there is still a bug with the general transformation reading. It seems to be reading the transpose of the matrix I wrote out.
  A: Yes, the new code I just posted fixed that, although you could change line 187 of TreeSerializer.java from transform.getMatrix().setElement(xCtr, yCtr, readFloat(br)); to transform.getMatrix().setElement(yCtr, xCtr, readFloat(br));.
• 10:15 pm, 13 November 2004: What am I supposed to do when collapsing transformations: combine them with the children or introduce new nodes?
  A: When you push a transformation down, you should combine it with any child transformations (resulting in a general transformation) and create a new transformation above any child Shapes. For example, the hierarchy
  • Rotation
    • Translation
      • Cube
    • Sphere
  upon collapsing the rotation would become
  • General
    • Cube
  • Rotation
    • Sphere
• 10:00 pm, 13 November 2004: What should I do about that human figure model that's mentioned only in the first section of the handout?
  A: Oops! We meant to leave that out of the assignment; it will be part of a future homework instead.
• 10:00 pm, 13 November 2004: I'm confused about what exactly is supposed to happen when the user edits a nonuniform scale. The handout is kind of vague, and the comment in the code is confusing too.
  A: A nonuniform scale has a scale factor along each of the three axes of the coordinate system in which it is applied. We need to take a mouse motion vector and convert it into these three scale factors. The rule you should use: a mouse motion should scale along the axis that is closest to the same 3D direction as the mouse motion, interpreting the mouse motion as a vector on the image plane. That is, if we consider the three scaling axes in world space, the scale should happen only along one axis, and that axis should be the one that has the smallest angle to the mouse motion vector. Consider this FAQ entry to supersede those other two sources.
• 7:30 pm, 13 November 2004: There seems to be a problem with the framework code in regards to reading in General Transformations.
  A: There sure is. Its been fixed in the newest version of the code. Whatever saved files you had created will still work fine.
• 12:30 pm, 12 November 2004: I'm confused in general - the framework code is enormous.
  A: We have added javadoc comments above all of the methods that you need to be filling in. This should provide a quick "sound and completeness" check.
• 12:30 pm, 12 November 2004: The handout says that the framework code should have fully working orthographic camera views, but they seem to be unimplemented
  A: In order to leave in some of the thinking necessary to move the PerspectiveCamera, we had to pull out some of the code for OrthoCamera and leave it broken. To fix this, write the method Camera#convertMotion(Vector2f, Vector3f)
• 12:30 pm, 12 November 2004: How should I write the Camera#converMotion( Vector2f, Vector3f) method in order to make it work like it should for the OrthoCamera?
  A: You should interpret a mouse motion of [x, y] as being magnitudes of motion in the direction of the up and right vector. Both x and y range from [-2, 2], where -2 indicates moving all the way from the top of the canvas to the bottom, or from right to left. Positive motion indicates moving from the bottom/left to the top/right. As a result, you should probably take into account the height and aspect ratio of the camera.
• 12:30 pm, 12 November 2004: How many triangle strips should we use in our sphere? How many triangles per triangle strip? Do we have to cache the points?
  A: Variable, variable, and no. You should make it easy to change the number of divisions in either direction for the sphere, as well as the cylinder, by altering the code. A single class variable at the top will suffice. Caching the points is not required by any means, re-evaluating the points each time is plenty fast.
• 12:30 pm, 12 November 2004: I can't get the vertical mouse motion to quite work out just right. The objects almost follow the mouse, but not quite.
  A: This is a bug in the solution code. The newest code has the fix, or you can just change lines 416 and 417 of GLView.java from "int w = getWidth(); int h = getHeight();" to "int w = canvas.getWidth(); int h = canvas.getHeight();".

OpenGL Information

• The OpenGL blue book - A programming reference
• The OpenGL red book - A programming tutorial
• NeHe - Check out the tutorials on the left
• JOGL Javadoc - Not particularly helpful to learn OpenGL proper, but it will show you how names translate in JOGL
• Winding Orders - This shows how OpenGL interprets the different shapes that you can glBegin()

Setup Information