pyopengl-users

[PyOpenGL-Users] Picking using SelectBuffer

[Simon W. <sim...@gm...>]

I'm trying to use glRenderMode(GL_SELECT) to implement picking. The
problem is, almost all picking examples deal with 3D projection
(gluPerspective), whereas I am using glOrtho, so I suspect I need to
do something different.

The below code doesn't work, it accumulates hits in the select buffer,
but they are not the hits I expect. Any pointers on where I might be
going wrong?

Also, is there a simple way to get OpenGL to render a small box on the
screen which shows exactly which region is being picked?

Thanks for any help...=20

Sw.


    def pick(self, position):
        glInitNames()
        glSelectBuffer(64)
        glRenderMode(GL_SELECT)
        glMatrixMode(GL_PROJECTION)
        glPushMatrix()
        glLoadIdentity()
        viewport =3D glGetIntegerv(GL_VIEWPORT)
        n_w,n_h =3D display._screen.dimensions
        glOrtho(-n_w,n_w,-n_h,n_h,1,0)
        gluPickMatrix(position[0],viewport[3]-position[1],3,3,viewport)
        glMatrixMode(GL_MODELVIEW)
        for index, i in enumerate(self.items):
            glPushName(index)
            display._screen.blit(i.item, i.position)
            glPopName()
        glMatrixMode(GL_PROJECTION)
        glPopMatrix()
        glMatrixMode(GL_MODELVIEW)
        glFlush()   =20
        items =3D []
        for hit in glRenderMode(GL_RENDER):
            items =3D items + [self.items[i] for i in hit[2]]
        return items

Re: [PyOpenGL-Users] Picking using SelectBuffer

[Mike C. F. <mcf...@ro...>]

Simon Wittber wrote:

>I'm trying to use glRenderMode(GL_SELECT) to implement picking. The
>problem is, almost all picking examples deal with 3D projection
>(gluPerspective), whereas I am using glOrtho, so I suspect I need to
>do something different.
>  
>
I don't actually use glOrtho, but I'll try to help...

>The below code doesn't work, it accumulates hits in the select buffer,
>but they are not the hits I expect. Any pointers on where I might be
>going wrong?
>  
>
 From the docs for gluPickMatrix:

    The matrix created by gluPickMatrix is multiplied by the current
    matrix just as if glMultMatrix is called with the generated matrix.
    To effectively use the generated pick matrix for picking, first call
    glLoadIdentity to load an identity matrix onto the perspective
    matrix stack. Then call gluPickMatrix, and finally, call a command
    (such as gluPerspective) to multiply the perspective matrix by the
    pick matrix.

which suggests you want the gluOrtho call *after* the gluPickMatrix 
call.  I don't actually recall reversing the order in *my* code, but 
then I haven't written code of this type in a long while.

>Also, is there a simple way to get OpenGL to render a small box on the
>screen which shows exactly which region is being picked?
>  
>
You are creating a rectangular prism centred around the pick location.  
So, do 8 gluUnProjects, one for each corner of the prism, with the near 
face having winz of 0.0 and the far face having winz of 1.0.  That 
should give you a rectangle in model-space coordinates that you can 
render (if I'm working through the logic correctly).

btw, you might want to check around and see if bitmap operations where 
the bitmap position is outside of the picking matrix actually register 
with a selection buffer (i.e. they don't draw if the bitmap position is 
out-of-frustum, do they select when outside the picking frustum?)

Anyway, HTH,
Mike
...

>        glLoadIdentity()
>        viewport = glGetIntegerv(GL_VIEWPORT)
>        n_w,n_h = display._screen.dimensions
>        glOrtho(-n_w,n_w,-n_h,n_h,1,0)
>        gluPickMatrix(position[0],viewport[3]-position[1],3,3,viewport)
>  
>
...

>            display._screen.blit(i.item, i.position)
>
>  
>
...

-- 
________________________________________________
  Mike C. Fletcher
  Designer, VR Plumber, Coder
  http://www.vrplumber.com
  http://blog.vrplumber.com