glCamera.py :  » Development » PyObjC » trunk » pyobjc » PyOpenGL-2.0.2.01 » OpenGL » Demo » NeHe » lesson44 » Python Open Source

Home
Python Open Source
1.3.1.2 Python
2.Ajax
3.Aspect Oriented
4.Blog
5.Build
6.Business Application
7.Chart Report
8.Content Management Systems
9.Cryptographic
10.Database
11.Development
12.Editor
13.Email
14.ERP
15.Game 2D 3D
16.GIS
17.GUI
18.IDE
19.Installer
20.IRC
21.Issue Tracker
22.Language Interface
23.Log
24.Math
25.Media Sound Audio
26.Mobile
27.Network
28.Parser
29.PDF
30.Project Management
31.RSS
32.Search
33.Security
34.Template Engines
35.Test
36.UML
37.USB Serial
38.Web Frameworks
39.Web Server
40.Web Services
41.Web Unit
42.Wiki
43.Windows
44.XML
Python Open Source » Development » PyObjC 
PyObjC » trunk » pyobjc » PyOpenGL 2.0.2.01 » OpenGL » Demo » NeHe » lesson44 » glCamera.py
# // Code writen by: Vic Hollis 09/07/2003
# // I don't mind if you use this class in your own code. All I ask is 
# // that you give me credit for it if you do.  And plug NeHe while your
# // at it! :P  Thanks go to David Steere, Cameron Tidwell, Bert Sammons,
# // and Brannon Martindale for helping me test all the code!  Enjoy.
# //////////////////////////////////////////////////////////////////////
# // glCamera.h: interface for the glCamera class.
# //////////////////////////////////////////////////////////////////////
# 
# //////////////////////////////////////////////////////////////////////
# // Some minimal additions by rIO.Spinning Kids 
# // For testing flares against occluding objects.
# // Not using proprietary extensions, this is PURE OpenGL1.1
# //
# // Just call the IsOccluded function, passing it the glPoint to check
# //
# //////////////////////////////////////////////////////////////////////
#
# Ported to Python, PyOpenGL by Brian Leair 2004.
# The numarray python module can perform matrix math more effieciently 
# than direct python code. However, for this tutorial the differnce
# in performance isn't huge and it makes for a better tutorial to see
# the math operations directly.

from OpenGL.GL import *
from OpenGL.GLU import *
from OpenGL.GLUT import *
from glPoint import *
from glVector import *
from math import sqrt,fabs

import Numeric
import copy

class glCamera:
  # //////////// CONSTRUCTORS /////////////////////////////////////////
  def __init__ (self):
    # // Initalize all our member varibles.
    self.m_MaxPitchRate      = 0.0;
    self.m_MaxHeadingRate    = 0.0;
    self.m_HeadingDegrees    = 0.0;
    self.m_PitchDegrees      = 0.0;
    self.m_MaxForwardVelocity  = 0.0;
    self.m_ForwardVelocity    = 0.0;
    self.m_GlowTexture          = None;
    # bleair: NOTE that glCamera.cpp has a bug. m_BigGlowTexture isn't initialized.
    # Very minor bug because only in the case where we fail to get an earlier
    # texture will the class potentially read from the uninited memory. Most of
    # the time the field is assigned to straight away in InitGL ().
    self.m_BigGlowTexture       = None;
    self.m_HaloTexture      = None;
    self.m_StreakTexture    = None;
    self.m_MaxPointSize      = 0.0;
    self.m_Frustum = Numeric.zeros ( (6,4), 'f')

    self.m_LightSourcePos     = glPoint ()
    self.m_Position = glPoint ()
    self.m_DirectionVector = glVector ()
    self.m_ptIntersect = glPoint ()




  def __del__ (self):
    self.release ()
    return

  def release (self):
    if (self.m_GlowTexture != None):
      glDeleteTextures (self.m_GlowTexture)
    if (self.m_HaloTexture != None):
      glDeleteTextures (self.m_HaloTexture)
    if (self.m_BigGlowTexture != None):
      glDeleteTextures (self.m_BigGlowTexture)
    if (self.m_StreakTexture != None):
      glDeleteTextures (self.m_StreakTexture)
    return

  def ChangePitch (self, degrees):
    if (fabs (degrees) < fabs (self.m_MaxPitchRate)):
      # // Our pitch is less than the max pitch rate that we 
      # // defined so lets increment it.
      self.m_PitchDegrees += degrees;
    else:
      # // Our pitch is greater than the max pitch rate that
      # // we defined so we can only increment our pitch by the 
      # // maximum allowed value.
      if(degrees < 0):
        # // We are pitching down so decrement
        self.m_PitchDegrees -= self.m_MaxPitchRate;
      else:
        # // We are pitching up so increment
        self.m_PitchDegrees += self.m_MaxPitchRate;

    # // We don't want our pitch to run away from us. Although it
    # // really doesn't matter I prefer to have my pitch degrees
    # // within the range of -360.0f to 360.0f
    if (self.m_PitchDegrees > 360.0):
      self.m_PitchDegrees -= 360.0;
    elif (self.m_PitchDegrees < -360.0):
      self.m_PitchDegrees += 360.0;

    return

  def ChangeHeading (self, degrees):
    if(fabs(degrees) < fabs(self.m_MaxHeadingRate)):
      # // Our Heading is less than the max heading rate that we 
      # // defined so lets increment it but first we must check
      # // to see if we are inverted so that our heading will not
      # // become inverted.
      if (self.m_PitchDegrees > 90 and self.m_PitchDegrees < 270 or 
        (self.m_PitchDegrees < -90 and self.m_PitchDegrees > -270)):
        self.m_HeadingDegrees -= degrees;
      else:
        self.m_HeadingDegrees += degrees;
    else:
      # // Our heading is greater than the max heading rate that
      # // we defined so we can only increment our heading by the 
      # // maximum allowed value.
      if(degrees < 0):
        # // Check to see if we are upside down.
        if ((self.m_PitchDegrees > 90 and self.m_PitchDegrees < 270) or
          (self.m_PitchDegrees < -90 and self.m_PitchDegrees > -270)):
          # // Ok we would normally decrement here but since we are upside
          # // down then we need to increment our heading
          self.m_HeadingDegrees += self.m_MaxHeadingRate;
        else:
          # // We are not upside down so decrement as usual
          self.m_HeadingDegrees -= self.m_MaxHeadingRate;
      else:
        # // Check to see if we are upside down.
        if (self.m_PitchDegrees > 90 and self.m_PitchDegrees < 270 or
          (self.m_PitchDegrees < -90 and self.m_PitchDegrees > -270)):
          # // Ok we would normally increment here but since we are upside
          # // down then we need to decrement our heading.
          self.m_HeadingDegrees -= self.m_MaxHeadingRate;
        else:
          # // We are not upside down so increment as usual.
          self.m_HeadingDegrees += self.m_MaxHeadingRate;
  
    # // We don't want our heading to run away from us either. Although it
    # // really doesn't matter I prefer to have my heading degrees
    # // within the range of -360.0f to 360.0f
    if(self.m_HeadingDegrees > 360.0):
      self.m_HeadingDegrees -= 360.0;
    elif(self.m_HeadingDegrees < -360.0):
      self.m_HeadingDegrees += 360.0;

    return

  # //////////// FUNCTIONS TO CHANGE CAMERA ORIENTATION AND SPEED /////
  def ChangeVelocity(self, vel):
    if(fabs(vel) < fabs(self.m_MaxForwardVelocity)):
      # // Our velocity is less than the max velocity increment that we 
      # // defined so lets increment it.
      self.m_ForwardVelocity += vel;
    else:
      # // Our velocity is greater than the max velocity increment that
      # // we defined so we can only increment our velocity by the 
      # // maximum allowed value.
      if(vel < 0):
        # // We are slowing down so decrement
        self.m_ForwardVelocity -= -self.m_MaxForwardVelocity;
      else:
        # // We are speeding up so increment
        self.m_ForwardVelocity += self.m_MaxForwardVelocity;

    return

  def UpdateFrustum(self):
    """ // I found this code here: http://www.markmorley.com/opengl/frustumculling.html
    // and decided to make it part of
    // the camera class just in case I might want to rotate
    // and translate the projection matrix. This code will
    // make sure that the Frustum is updated correctly but
    // this member is computational expensive with:
    // 82 muliplications, 72 additions, 24 divisions, and
    // 12 subtractions for a total of 190 operations. Ouch! """

    # /* Get the current PROJECTION matrix from OpenGL */
    proj = glGetFloatv( GL_PROJECTION_MATRIX);

    # /* Get the current MODELVIEW matrix from OpenGL */
    modl = glGetFloatv( GL_MODELVIEW_MATRIX);

    # /* Combine the two matrices (multiply projection by modelview) */
    # Careful, Note, that replication is simple scalars is OK, but replicate of objects
    # and lists is very bad.
    clip = [None,] * 16
    # clip = Numeric.zeros ( (16), 'f')
    clip[ 0] = modl[ 0] * proj[ 0] + modl[ 1] * proj[ 4] + modl[ 2] * proj[ 8] + modl[ 3] * proj[12];
    clip[ 1] = modl[ 0] * proj[ 1] + modl[ 1] * proj[ 5] + modl[ 2] * proj[ 9] + modl[ 3] * proj[13];
    clip[ 2] = modl[ 0] * proj[ 2] + modl[ 1] * proj[ 6] + modl[ 2] * proj[10] + modl[ 3] * proj[14];
    clip[ 3] = modl[ 0] * proj[ 3] + modl[ 1] * proj[ 7] + modl[ 2] * proj[11] + modl[ 3] * proj[15];

    clip[ 4] = modl[ 4] * proj[ 0] + modl[ 5] * proj[ 4] + modl[ 6] * proj[ 8] + modl[ 7] * proj[12];
    clip[ 5] = modl[ 4] * proj[ 1] + modl[ 5] * proj[ 5] + modl[ 6] * proj[ 9] + modl[ 7] * proj[13];
    clip[ 6] = modl[ 4] * proj[ 2] + modl[ 5] * proj[ 6] + modl[ 6] * proj[10] + modl[ 7] * proj[14];
    clip[ 7] = modl[ 4] * proj[ 3] + modl[ 5] * proj[ 7] + modl[ 6] * proj[11] + modl[ 7] * proj[15];

    clip[ 8] = modl[ 8] * proj[ 0] + modl[ 9] * proj[ 4] + modl[10] * proj[ 8] + modl[11] * proj[12];
    clip[ 9] = modl[ 8] * proj[ 1] + modl[ 9] * proj[ 5] + modl[10] * proj[ 9] + modl[11] * proj[13];
    clip[10] = modl[ 8] * proj[ 2] + modl[ 9] * proj[ 6] + modl[10] * proj[10] + modl[11] * proj[14];
    clip[11] = modl[ 8] * proj[ 3] + modl[ 9] * proj[ 7] + modl[10] * proj[11] + modl[11] * proj[15];

    clip[12] = modl[12] * proj[ 0] + modl[13] * proj[ 4] + modl[14] * proj[ 8] + modl[15] * proj[12];
    clip[13] = modl[12] * proj[ 1] + modl[13] * proj[ 5] + modl[14] * proj[ 9] + modl[15] * proj[13];
    clip[14] = modl[12] * proj[ 2] + modl[13] * proj[ 6] + modl[14] * proj[10] + modl[15] * proj[14];
    clip[15] = modl[12] * proj[ 3] + modl[13] * proj[ 7] + modl[14] * proj[11] + modl[15] * proj[15];

    # ### Use a shortened name to reference to our camera's Frustum (does 
    # ### not copy anything, just a ref to make code less wordy
    Frustum = self.m_Frustum

    # /* Extract the numbers for the RIGHT plane */
    Frustum[0][0] = clip[ 3] - clip[ 0];
    Frustum[0][1] = clip[ 7] - clip[ 4];
    Frustum[0][2] = clip[11] - clip[ 8];
    Frustum[0][3] = clip[15] - clip[12];

      # /* Normalize the result */
    t = (sqrt( Frustum[0][0] * Frustum[0][0] + \
    Frustum[0][1] * Frustum[0][1] + Frustum[0][2] * Frustum[0][2] ));
    Frustum[0][0] /= t;
    Frustum[0][1] /= t;
    Frustum[0][2] /= t;
    Frustum[0][3] /= t;

    # /* Extract the numbers for the LEFT plane */
    Frustum[1][0] = clip[ 3] + clip[ 0];
    Frustum[1][1] = clip[ 7] + clip[ 4];
    Frustum[1][2] = clip[11] + clip[ 8];
    Frustum[1][3] = clip[15] + clip[12];

    # /* Normalize the result */
    t = sqrt( Frustum[1][0] * Frustum[1][0] + Frustum[1][1] * Frustum[1][1] + Frustum[1][2] * Frustum[1][2] );
    Frustum[1][0] /= t;
    Frustum[1][1] /= t;
    Frustum[1][2] /= t;
    Frustum[1][3] /= t;

    # /* Extract the BOTTOM plane */
    Frustum[2][0] = clip[ 3] + clip[ 1];
    Frustum[2][1] = clip[ 7] + clip[ 5];
    Frustum[2][2] = clip[11] + clip[ 9];
    Frustum[2][3] = clip[15] + clip[13];

    # /* Normalize the result */
    t = sqrt( Frustum[2][0] * Frustum[2][0] + Frustum[2][1] * Frustum[2][1] + Frustum[2][2] * Frustum[2][2] );
    Frustum[2][0] /= t;
    Frustum[2][1] /= t;
    Frustum[2][2] /= t;
    Frustum[2][3] /= t;

    # /* Extract the TOP plane */
    Frustum[3][0] = clip[ 3] - clip[ 1];
    Frustum[3][1] = clip[ 7] - clip[ 5];
    Frustum[3][2] = clip[11] - clip[ 9];
    Frustum[3][3] = clip[15] - clip[13];

      # /* Normalize the result */
    t = sqrt( Frustum[3][0] * Frustum[3][0] + Frustum[3][1] * Frustum[3][1] + Frustum[3][2] * Frustum[3][2] )
    Frustum[3][0] /= t;
    Frustum[3][1] /= t;
    Frustum[3][2] /= t;
    Frustum[3][3] /= t;

     # /* Extract the FAR plane */
     Frustum[4][0] = clip[ 3] - clip[ 2];
     Frustum[4][1] = clip[ 7] - clip[ 6];
     Frustum[4][2] = clip[11] - clip[10];
    Frustum[4][3] = clip[15] - clip[14];

    # /* Normalize the result */
     t = sqrt( Frustum[4][0] * Frustum[4][0] + Frustum[4][1] * Frustum[4][1] + Frustum[4][2] * Frustum[4][2] )
     Frustum[4][0] /= t;
     Frustum[4][1] /= t;
     Frustum[4][2] /= t;
     Frustum[4][3] /= t;

     # /* Extract the NEAR plane */
     Frustum[5][0] = clip[ 3] + clip[ 2];
     Frustum[5][1] = clip[ 7] + clip[ 6];
     Frustum[5][2] = clip[11] + clip[10];
     Frustum[5][3] = clip[15] + clip[14];

     # /* Normalize the result */
     t = sqrt( Frustum[5][0] * Frustum[5][0] + Frustum[5][1] * Frustum[5][1] + Frustum[5][2] * Frustum[5][2] );
     Frustum[5][0] /= t;
     Frustum[5][1] /= t;
     Frustum[5][2] /= t;
     Frustum[5][3] /= t;

    return

  # //////////// FUNCTIONS TO UPDATE THE FRUSTUM //////////////////////
  def UpdateFrustumFaster (self):
    """ // This is the much faster version of the above member 
    // function, however the speed increase is not gained 
    // without a cost. If you rotate or translate the projection
    // matrix then this member will not work correctly. That is acceptable
    // in my book considering I very rarely do such a thing.
    // This function has far fewer operations in it and I 
    // shaved off 2 square root functions by passing in the
    // near and far values. This member has:
    // 38 muliplications, 28 additions, 24 divisions, and
    // 12 subtractions for a total of 102 operations. Still hurts
    // but at least it is decent now. In practice this will 
    // run about 2 times faster than the above function. """

    # /* Get the current PROJECTION matrix from OpenGL */
    proj = glGetFloatv( GL_PROJECTION_MATRIX);
  
    # /* Get the current MODELVIEW matrix from OpenGL */
    modl = glGetFloatv( GL_MODELVIEW_MATRIX);
  
    # /* Combine the two matrices (multiply projection by modelview) 
       # but keep in mind this function will only work if you do NOT
       # rotate or translate your projection matrix                  */
    clip = [0,] * 16
    modl_row1 = modl [0]
    clip[ 0] = modl [0] [0] * proj[0][0];
    clip[ 1] = modl  [0][ 1] * proj[1][1];
    clip[ 2] = modl [0][ 2] * proj[2][2] + modl_row1[ 3] * proj[3][2]
    clip[ 3] = modl [0][ 2] * proj[2][3]
  
    modl_row2 = modl [1]
    clip[ 4] = modl_row2[ 0] * proj[0][0]
    clip[ 5] = modl_row2[ 1] * proj[1][1]
    clip[ 6] = modl_row2[ 2] * proj[2][2] + modl_row2[ 3] * proj[3][2]
    clip[ 7] = modl_row2[ 2] * proj[2][3]
  
    modl_row3 = modl [2]
    clip[ 8] = modl_row3[ 0] * proj[0][0];
    clip[ 9] = modl_row3[ 1] * proj[1][1]
    clip[10] = modl_row3[2] * proj[2][2] + modl_row3[3] * proj[3][2]
    clip[11] = modl_row3[2] * proj[2][3]
  
    modl_row4 = modl [3]
    clip[12] = modl_row4[0] * proj[0][0]
    clip[13] = modl_row4[1] * proj[1][1]
    clip[14] = modl_row4[2] * proj[2][2] + modl_row4[3] * proj[3][2]
    clip[15] = modl_row4[2] * proj[2][3]
  
    # ### Use a shortened name to reference to our camera's Frustum (does 
    # ### not copy anything, just a ref to make code less wordy
    Frustum = self.m_Frustum

    # /* Extract the numbers for the RIGHT plane */
    Frustum[0][0] = clip[ 3] - clip[ 0];
    Frustum[0][1] = clip[ 7] - clip[ 4];
    Frustum[0][2] = clip[11] - clip[ 8];
    Frustum[0][3] = clip[15] - clip[12];
  
    # /* Normalize the result */
    t = sqrt( (Frustum[0][0] * Frustum[0][0]) + (Frustum[0][1] * Frustum[0][1]) + (Frustum[0][2] * Frustum[0][2]) );
    Frustum[0][0] /= t;
    Frustum[0][1] /= t;
    Frustum[0][2] /= t;
    Frustum[0][3] /= t;
  
     # /* Extract the numbers for the LEFT plane */
     Frustum[1][0] = clip[ 3] + clip[ 0];
     Frustum[1][1] = clip[ 7] + clip[ 4];
     Frustum[1][2] = clip[11] + clip[ 8];
     Frustum[1][3] = clip[15] + clip[12];
  
      # /* Normalize the result */
     t = sqrt( Frustum[1][0] * Frustum[1][0] + Frustum[1][1] * Frustum[1][1] + Frustum[1][2] * Frustum[1][2] );
     Frustum[1][0] /= t;
     Frustum[1][1] /= t;
     Frustum[1][2] /= t;
     Frustum[1][3] /= t;
  
    # /* Extract the BOTTOM plane */
    Frustum[2][0] = clip[ 3] + clip[ 1];
    Frustum[2][1] = clip[ 7] + clip[ 5];
    Frustum[2][2] = clip[11] + clip[ 9];
    Frustum[2][3] = clip[15] + clip[13];
  
      # /* Normalize the result */
    t = sqrt( Frustum[2][0] * Frustum[2][0] + Frustum[2][1] * Frustum[2][1] + Frustum[2][2] * Frustum[2][2] );
    Frustum[2][0] /= t;
    Frustum[2][1] /= t;
    Frustum[2][2] /= t;
    Frustum[2][3] /= t;
  
      # /* Extract the TOP plane */
     Frustum[3][0] = clip[ 3] - clip[ 1];
     Frustum[3][1] = clip[ 7] - clip[ 5];
     Frustum[3][2] = clip[11] - clip[ 9];
     Frustum[3][3] = clip[15] - clip[13];
  
      # /* Normalize the result */
     t = sqrt( Frustum[3][0] * Frustum[3][0] + Frustum[3][1] * Frustum[3][1] + Frustum[3][2] * Frustum[3][2] );
     Frustum[3][0] /= t;
     Frustum[3][1] /= t;
     Frustum[3][2] /= t;
     Frustum[3][3] /= t;
  
      # /* Extract the FAR plane */
     Frustum[4][0] = clip[ 3] - clip[ 2];
     Frustum[4][1] = clip[ 7] - clip[ 6];
     Frustum[4][2] = clip[11] - clip[10];
     Frustum[4][3] = clip[15] - clip[14];
  
      # /* Normalize the result */
     t = sqrt( (Frustum[4][0] * Frustum[4][0]) + (Frustum[4][1] * Frustum[4][1]) + (Frustum[4][2] * Frustum[4][2]) );
     Frustum[4][0] /= t;
     Frustum[4][1] /= t;
     Frustum[4][2] /= t;
     Frustum[4][3] /= t;
  
      # /* Extract the NEAR plane */
     Frustum[5][0] = clip[ 3] + clip[ 2];
     Frustum[5][1] = clip[ 7] + clip[ 6];
     Frustum[5][2] = clip[11] + clip[10];
     Frustum[5][3] = clip[15] + clip[14];
  
     # /* Normalize the result */
     t = sqrt( Frustum[5][0] * Frustum[5][0] + Frustum[5][1] * Frustum[5][1] + Frustum[5][2] * Frustum[5][2] );
     Frustum[5][0] /= t;
     Frustum[5][1] /= t;
     Frustum[5][2] /= t;
     Frustum[5][3] /= t;

    return
  
  

  # //////////// FRUSTUM TESTING FUNCTIONS ////////////////////////////
  def SphereInFrustum(self, p, Radius):
    """ // This member function checks to see if a sphere is in
      // the viewing volume.   """

    Frustum = self.m_Frustum
    # // The idea here is the same as the PointInFrustum function.
    if (Radius != 0):
      for i in xrange (6):
      # // If the point is outside of the plane then its not in the viewing volume.
        if(Frustum[i][0] * p.x + Frustum[i][1] * p.y + Frustum[i][2] * p.z + Frustum[i][3] <= -Radius):
          return(False);
    else:
      # // The idea here is the same as the PointInFrustum function.
      for i in xrange (6):
        # // If the point is outside of the plane then its not in the viewing volume.
        if(Frustum[i][0] * p.x + Frustum[i][1] * p.y + Frustum[i][2] * p.z + Frustum[i][3] <= 0):
          return(False);

    return(True);

  def PointInFrustum(self, x,y,z):
    """ // This member fuction checks to see if a point is in
      // the viewing volume. """

    # // The idea behind this algorithum is that if the point
    # // is inside all 6 clipping planes then it is inside our
    # // viewing volume so we can return true.

    Frustum = self.m_Frustum
    # // Loop through all our clipping planes
    for i in xrange (6):
      # // If the point is outside of the plane then its not in the viewing volume.
      if(Frustum[i][0] * x + Frustum[i][1] * y + Frustum[i][2] * z + Frustum[i][3] <= 0):
        return(False);

    return(True);

  # /////////// OCCLUSION TESTING FUNCTIONS ///////////////////////////
  def IsOccluded (self, p):
    # // Now we will ask OGL to project some geometry for us using the gluProject function.
    # // Practically we ask OGL to guess where a point in space will be projected in our current viewport,
    # // using arbitrary viewport and transform matrices we pass to the function.
    # // If we pass to the function the current matrices  (retrievede with the glGet funcs)
    # // we will have the real position on screen where the dot will be drawn.
    # // The interesting part is that we also get a Z value back, this means that 
    # // reading the REAL buffer for Z values we can discover if the flare is in front or
    # // if it's occluded by some objects.
    # ### This function should be a flat function, not a function of the camera as we
    # ### use the immediate GL rendering state entirely.


    # ### Viewport is the rectangle of window pixels that OpenGL is rasterizing into.
    viewport = glGetIntegerv (GL_VIEWPORT);            # //get actual viewport
      mvmatrix = glGetDoublev (GL_MODELVIEW_MATRIX);        # //get actual model view matrix
      projmatrix = glGetDoublev (GL_PROJECTION_MATRIX);      # //get actual projiection matrix

    # // this asks OGL to guess the 2d position of a 3d point inside the viewport
    winx, winy, winz = gluProject(p.x, p.y, p.z, mvmatrix, projmatrix, viewport)
    flareZ = winz;

    # // we read back one pixel from th depth buffer (exactly where our flare should be drawn)
    glPixelStorei(GL_PACK_ALIGNMENT, 1)

    # PyOpenGL 2.0.1.07 bug, Only the type clarified function works.
    # bufferZ = glReadPixels(int(winx), int(winy),1,1,GL_DEPTH_COMPONENT, GL_FLOAT)
    bufferZ = glReadPixelsf(int(winx), int(winy),1,1,GL_DEPTH_COMPONENT)

    # // if the buffer Z is lower than our flare guessed Z then don't draw 
    # // this means there is something in front of our flare
    if (bufferZ [0] [0] < flareZ):
      return True;
    else:
      return False;

  # //////////// FUNCTIONS TO RENDER LENS FLARES //////////////////////
  def RenderLensFlare(self):
    # // Draw the flare only If the light source is in our line of sight (inside the Frustum)
    if (self.SphereInFrustum(self.m_LightSourcePos, 1.0) == True):

      # Vector pointing from the light's position toward the camera's position (the camera might
      # be pointing elsewhere, this vector is pointing from the light to the camera)
      self.vLightSourceToCamera = self.m_Position - self.m_LightSourcePos;    # // Lets compute the vector that points to the camera from
                                            # // the light source.

      Length = self.vLightSourceToCamera.Magnitude ()             # // Save the length we will need it in a minute

      # Move down our look-toward direction vector. Move down the look-toward the same dist. as the
      # distance between camera and the light.
      intersect = self.m_DirectionVector * Length
      self.m_ptIntersect = glPoint (intersect.i, intersect.j, intersect.k)
                                    # // Now lets find an point along the cameras direction
                                    # // vector that we can use as an intersection point. 
                                    # // Lets translate down this vector the same distance
                                    # // that the camera is away from the light source.
      ptIntersect = self.m_ptIntersect
      # Did the motion in the correct direction above, now translate the intersection position 
      # relative to our camera location.
      ptIntersect += self.m_Position;


      self.vLightSourceToIntersect = ptIntersect - self.m_LightSourcePos;    # // Lets compute the vector that points to the Intersect
                                  # // point from the light source
          
      Length = self.vLightSourceToIntersect.Magnitude();    # // Save the length we will need it later.
      self.vLightSourceToIntersect.Normalize();        # // Normalize the vector so its unit length
      vLightSourceToIntersect = self.vLightSourceToIntersect
    
      glEnable(GL_BLEND);                    # // You should already know what this does
      glBlendFunc(GL_SRC_ALPHA, GL_ONE);            # // You should already know what this does
      glDisable(GL_DEPTH_TEST);                # // You should already know what this does
      glEnable(GL_TEXTURE_2D);                # // You should already know what this does
      
      # /////////// Differenet Color Glows & Streaks /////////////////////
      # //RenderBigGlow(1.0f, 1.0f, 1.0f, 1.0f, m_LightSourcePos, 1.0f);
      # //RenderStreaks(1.0f, 1.0f, 0.8f, 1.0f, m_LightSourcePos, 0.7f);
      # //
      # //RenderBigGlow(1.0f, 0.9f, 1.0f, 1.0f, m_LightSourcePos, 1.0f);
      # //RenderStreaks(1.0f, 0.9f, 1.0f, 1.0f, m_LightSourcePos, 0.7f);
      # //////////////////////////////////////////////////////////////////


      # //########################## NEW STUFF ##################################

      if (not self.IsOccluded(self.m_LightSourcePos)):    #  //Check if the center of the flare is occluded
        # // Render the large hazy glow
        self.RenderBigGlow(0.60, 0.60, 0.8, 1.0, self.m_LightSourcePos, 16.0);
        # // Render the streaks
        self.RenderStreaks(0.60, 0.60, 0.8, 1.0, self.m_LightSourcePos, 16.0);
        # // Render the small Glow
        self.RenderGlow(0.8, 0.8, 1.0, 0.5, self.m_LightSourcePos, 3.5);

        pt = glPoint (vLightSourceToIntersect * (Length * 0.1));  # // Lets compute a point that is 20%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.
    
        self.RenderGlow(0.9, 0.6, 0.4, 0.5, pt, 0.6);          # // Render the small Glow

        pt = glPoint (vLightSourceToIntersect * (Length * 0.15));  # // Lets compute a point that is 30%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderHalo(0.8, 0.5, 0.6, 0.5, pt, 1.7);          # // Render the a Halo
    
        pt = glPoint (vLightSourceToIntersect * (Length * 0.175));      # // Lets compute a point that is 35%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderHalo(0.9, 0.2, 0.1, 0.5, pt, 0.83);          # // Render the a Halo

        pt = glPoint (vLightSourceToIntersect * (Length * 0.285));      # // Lets compute a point that is 57%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderHalo(0.7, 0.7, 0.4, 0.5, pt, 1.6);          # // Render the a Halo
    
        pt = glPoint (vLightSourceToIntersect * (Length * 0.2755));      # // Lets compute a point that is 55.1%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderGlow(0.9, 0.9, 0.2, 0.5, pt, 0.8);          # // Render the small Glow

        pt = glPoint (vLightSourceToIntersect * (Length * 0.4775));      # // Lets compute a point that is 95.5%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderGlow(0.93, 0.82, 0.73, 0.5, pt, 1.0);          # // Render the small Glow
    
        pt = glPoint (vLightSourceToIntersect * (Length * 0.49));        # // Lets compute a point that is 98%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderHalo(0.7, 0.6, 0.5, 0.5, pt, 1.4);          # // Render the a Halo

        pt = glPoint (vLightSourceToIntersect * (Length * 0.65));        # // Lets compute a point that is 130%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderGlow(0.7, 0.8, 0.3, 0.5, pt, 1.8);          # // Render the small Glow
    
        pt = glPoint (vLightSourceToIntersect * (Length * 0.63));        # // Lets compute a point that is 126%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderGlow(0.4, 0.3, 0.2, 0.5, pt, 1.4);          # // Render the small Glow

        pt = glPoint (vLightSourceToIntersect * (Length * 0.8));        # // Lets compute a point that is 160%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderHalo(0.7, 0.5, 0.5, 0.5, pt, 1.4);          # // Render the a Halo
    
        pt = glPoint (vLightSourceToIntersect * (Length * 0.7825));      # // Lets compute a point that is 156.5%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.

        self.RenderGlow(0.8, 0.5, 0.1, 0.5, pt, 0.6);          # // Render the small Glow

        pt = glPoint (vLightSourceToIntersect * (Length * 1.0));        # // Lets compute a point that is 200%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderHalo(0.5, 0.5, 0.7, 0.5, pt, 1.7);          # // Render the a Halo
    
        pt = glPoint (vLightSourceToIntersect * (Length * 0.975));      # // Lets compute a point that is 195%
        pt += self.m_LightSourcePos;                # // away from the light source in the
                                      # // direction of the intersection point.    
    
        self.RenderGlow(0.4, 0.1, 0.9, 0.5, pt, 2.0);          # // Render the small Glow

      glDisable(GL_BLEND );                      # // You should already know what this does
      glEnable(GL_DEPTH_TEST);                    # // You should already know what this does
      glDisable(GL_TEXTURE_2D);                    # // You should already know what this does
    return

  def RenderHalo (self, r, g, b, a, p, scale):
    self.RenderFlareTexture (self.m_HaloTexture, r, g, b, a, p, scale)
    return

  def RenderGlow (self, r, g, b, a, p, scale):
    self.RenderFlareTexture (self.m_GlowTexture, r, g, b, a, p, scale)
    return

  def RenderBigGlow (self, r, g, b, a, p, scale):
    self.RenderFlareTexture (self.m_BigGlowTexture, r, g, b, a, p, scale)
    return

  def RenderStreaks (self, r, g, b, a, p, scale):
    self.RenderFlareTexture (self.m_StreakTexture, r, g, b, a, p, scale)
    return

  def RenderFlareTexture (self, tex_ID, r, g, b, a, p, scale):
    # bleair: Duplicate functions all the same except for the texture to bind to.

    q = []
    q.append (glPoint ())
    q.append (glPoint ())
    q.append (glPoint ())
    q.append (glPoint ())
    # // Basically we are just going to make a 2D box
    # // from four points we don't need a z coord because
    # // we are rotating the camera by the inverse so the 
    # // texture mapped quads will always face us.

    q[0].x = (p.x - scale);                      # // Set the x coordinate -scale units from the center point.
    q[0].y = (p.y - scale);                      # // Set the y coordinate -scale units from the center point.
    
    q[1].x = (p.x - scale);                      # // Set the x coordinate -scale units from the center point.
    q[1].y = (p.y + scale);                      # // Set the y coordinate scale units from the center point.
    
    q[2].x = (p.x + scale);                      # // Set the x coordinate scale units from the center point.
    q[2].y = (p.y - scale);                      # // Set the y coordinate -scale units from the center point.
    
    q[3].x = (p.x + scale);                      # // Set the x coordinate scale units from the center point.
    q[3].y = (p.y + scale);                      # // Set the y coordinate scale units from the center point.
    
    glPushMatrix();                          # // Save the model view matrix
    glTranslatef(p.x, p.y, p.z);                  # // Translate to our point
    glRotatef(-self.m_HeadingDegrees, 0.0, 1.0, 0.0);
    glRotatef(-self.m_PitchDegrees, 1.0, 0.0, 0.0);
    glBindTexture(GL_TEXTURE_2D, tex_ID);              # // Bind to the Big Glow texture
    glColor4f(r, g, b, a);                      # // Set the color since the texture is a gray scale
  
    glBegin(GL_TRIANGLE_STRIP);                    # // Draw the Big Glow on a Triangle Strip
    glTexCoord2f(0.0, 0.0);          
    glVertex2f(q[0].x, q[0].y);
    glTexCoord2f(0.0, 1.0);
    glVertex2f(q[1].x, q[1].y);
    glTexCoord2f(1.0, 0.0);
    glVertex2f(q[2].x, q[2].y);
    glTexCoord2f(1.0, 1.0);
    glVertex2f(q[3].x, q[3].y);
    glEnd();                    
    glPopMatrix();                          # // Restore the model view matrix
    return




  def SetPrespective (self):
    # Matrix = [0] * 16           # // A (list) array to hold the model view matrix.

    # However the MODELVIEW was oriented, we now rotate it based upon our Camer object's state.
    # // Going to use glRotate to calculate our direction vector
    glRotatef(self.m_HeadingDegrees, 0.0, 1.0, 0.0);    # turn your head left/right (around y axe)
    glRotatef(self.m_PitchDegrees, 1.0, 0.0, 0.0);      # nod your head up/down (around x axe)

    # // Get the resulting matrix from OpenGL it will have our
    # // direction vector in the 3rd row.
    Matrix = glGetFloatv(GL_MODELVIEW_MATRIX);

    # // Get the direction vector from the matrix. Element 10 must
    # // be inverted!
    self.m_DirectionVector.i = Matrix[2] [0]  #[8];
    self.m_DirectionVector.j = Matrix[2] [1]  #[9];
    self.m_DirectionVector.k = -Matrix[2] [2]   #[10];

    # #### bleair: no need to do this as this. Previous rotates already here (because
    # #### all invocations have the modelview at identity.
    # #### Suspect this was just a bit of code that was mvoed up and not deleted here.
    # // Ok erase the results of the last computation.
    glLoadIdentity();

    # // Rotate the scene to get the right orientation.
    glRotatef(self.m_PitchDegrees, 1.0, 0.0, 0.0);
    glRotatef(self.m_HeadingDegrees, 0.0, 1.0, 0.0);

    # // A vector to hold our cameras direction * the forward velocity
    # // we don't want to destory the Direction vector by using it instead.
    # // Scale the direction by our speed.
    v = copy.copy (self.m_DirectionVector);
    v *= self.m_ForwardVelocity;

    # // Increment our position by the vector
    self.m_Position.x += v.i;
    self.m_Position.y += v.j;
    self.m_Position.z += v.k;

    # // Translate to our new position.
    glTranslatef(-self.m_Position.x, -self.m_Position.y, -self.m_Position.z);
    return
    

"""
  //////////// MEMBER VARIBLES //////////////////////////////////////
  glVector vLightSourceToCamera, vLightSourceToIntersect;
  glPoint ptIntersect, pt;
  GLsizei m_WindowHeight;
  GLsizei m_WindowWidth;
  GLuint m_StreakTexture;
  GLuint m_HaloTexture;
  GLuint m_GlowTexture;
  GLuint m_BigGlowTexture;
  GLfloat m_MaxPointSize;
  GLfloat m_Frustum[6][4];
  glPoint m_LightSourcePos;
  GLfloat m_MaxPitchRate;
  GLfloat m_MaxHeadingRate;
  GLfloat m_HeadingDegrees;
  GLfloat m_PitchDegrees;
  GLfloat m_MaxForwardVelocity;
  GLfloat m_ForwardVelocity;
  glPoint m_Position;
  glVector m_DirectionVector;
"""

w__w_w___.__j_a_va2__s.___c___om__ | Contact Us
Copyright 2009 - 12 Demo Source and Support. All rights reserved.
All other trademarks are property of their respective owners.