possible bug with adjustment parameters

Here you can discuss and share functionality improvements and helper programs to make Context Free better.

Moderators: MtnViewJohn, chris, mtnviewmark

Post Reply
User avatar
kipling
Posts: 91
Joined: Wed Jun 18, 2008 2:36 am

possible bug with adjustment parameters

Post by kipling »

This:

Code: Select all

startshape START[b .5]

shape START{
	recurse([s .95 b .1],CIRCLE)[]
	recurse([s .95 b -.1],TRIANGLE)[x 1]
	recurse([s .95 h -5 r 3 a .05],SQUARE)[x 2 b 1 sat 1 a -1]
	recurse([s .95 h -5 r -5 a -.05],NGON(5))[x 3 b 1 sat 1]
	A[y -1.1]
	B[y -1.1 x 1]
	C[y -1.1 x 2 b 1 sat 1 a -1]
	D[y -1.1 x 3 b 1 sat 1]
	}

shape A{CIRCLE[]	A[s .95 b .1]}
shape B{TRIANGLE[] B[s .95 b -.1]}
shape C{SQUARE[] C[s .95 h -5 r 3 a .05]}
shape D{NGON(5)[] D[s .95 h -5 r -5 a -.05]}

shape recurse(adjustment change, shape thingie)
{
  thingie[]                      // invoke thingie with parameters already bound
  recurse(=) [transform change]  // can be abbreviated as trans
}

path NGON(number n){
	MOVETO(0,.5)
	loop n [r (360/n)] LINETO(0,.5)
	CLOSEPOLY()
	}
produces this
transform-param-bug.png
transform-param-bug.png (42.1 KiB) Viewed 40672 times
So it looks like any saturation/brightness/alpha values in a transform parameter are being pinned to [0,1]. I suppose it shouldn't do this without prior knowledge of this attribute of the shape it is applied to. As the code shows, it does this whether or not it is a valid adjustment to the parent shape.

Hue seems OK.

I haven't checked two-value versions of these adjustments.

I haven't dug into the source code.

I'm using Version 3.0.4 (36) on Mac OS X.

User avatar
MtnViewJohn
Site Admin
Posts: 882
Joined: Fri May 06, 2005 2:26 pm
Location: Mountain View, California
Contact:

Re: possible bug with adjustment parameters

Post by MtnViewJohn »

Ugh. I thought that I handled this situation, but it looks like I didn't. The ultimate cause is mathematical. The world state of a shape has four parts: geometry, Z, time, and color (there is also a random number seed). Geometry, Z, and time adjustments are all associative, [[s 1.05 s 1.05]] is exactly the same as [s (1.05*1.05)]. Hue adjustments are associative as long as you don't use the two value form.

But brightness, saturation and alpha adjustments are not associative. [[b 0.5 b 0.5]] cannot be replaced with a single [b n]. You have to hang onto both of those b 0.5 adjustments and apply them sequentially when it is time to do a shape replacement. This doesn't play nicely at all with CF3 adjustment variables and shape parameters.

If we had known in 2005 that associativity was a useful property for color components then we wouldn't be in this predicament today. But I can fix things so that a single color component change can be stored in a parameter. Some kind of warning will be issued if a design tries to accumulate more than one color change in a variable or parameter.

User avatar
kipling
Posts: 91
Joined: Wed Jun 18, 2008 2:36 am

Re: possible bug with adjustment parameters

Post by kipling »

Confirmed that version 3.0.5 (Mac OS X Context Free.App) now works as expected. Thanks.
transform-param-bug-fixed.png
transform-param-bug-fixed.png (43.51 KiB) Viewed 40636 times
Now I have to go back and remember what I was doing when I found this.

Post Reply