Ever wondered how to perfectly fill an outline with touching circles? I did -- after seeing the example image in Solving design effect - image made of perfectly arranged circles.
Select any path -- but no live text, please; you have to convert it to outlines and select each character in turn. Then run the script.
It displays a simple dialog, where you can set a maximum and minimum circle size as a percentage of the selected object size. In addition, you can select either a plain basic color, or select any of your current Swatch Groups; in that case, each of the circles are filled with a random color from that group.
The script may take a while to run. Usually, only a couple of seconds for a simple rectangle, but it may run into minutes for objects with lots of curves and/or holes. I didn't have the guts to run it on a vectorized world map, to recreate Mario "Quasimondo" Klingemann's Foam World Map; but, in theory? Possible.
This image only took a few minutes:
Works great in CS5 here.
How can I download this script?
Thanks for so generously sharing your talent!
1. The Relative Size determination wasn't really very good. Fixed. (Hopefully)
2. Holes in polygons threw everything off and were handled quite badly. Now proper (non-self intersecting!) polygons with holes in them are properly triangulated -- code ought to be a bit faster ... A propos: every single path inside an outer one is considered a 'hole', regardless of what Illustrator says. Don't try with two overlapping paths either; only use a simple outer path, with simple holes -- something like a '0' or '8' shape is good.
3. Added a useful parameter: minimal distance between two circles. This is only between circles, they still may touch edges.
Here is a sample: a big 'A' over a big 'B'. Separated, then filled 'A' only parts with greens, 'B' only parts with reds, and parts that are in both with browns. Then cut out the entire combined path out of the background rectangle and filled that with yellows.
(I meant to create a Protanopia-type Color Blindness test, but I guess I should have checked what they actually looked like before trying Still, it's an interesting image.)
.. as is this
Where is France, where is the United Kingdom and – of capital importance – where is Texel on that map?
Here is Texel, Kurt:
I don't know how you missed it.
There is no pleasing some people Texel is on that map, but your screen resolution is just too low! (And if you believe that, you might also believe the same goes for France and the Brit Isles.)
Here it is, in red; I also marked my own geospatial location with a green dot.
By the way, it seems I wasn't entirely accurate with my statement "now polygons with holes are properly handled." I'm working on a new version where -- again -- they should really be processed right ... If I don't make any significant progress with the next few problems I'm tackling now, I'll put this version online as well. In the mean time, just don't try "complicated" polygons.
I've found another slow-down reason: now that each point gets correctly placed inside the polygon, it still has to check the distance to each and every edge -- that's because a circle may not grow any larger than the distance to the nearest edge. There are several ways to circumvent this ... or so I think. I'm looking into BSP trees -- or maybe quadtrees --, and I might as well find out if I can get a working a polygon midline algorithm.
If I can get this fixed, it might be worthwhile to take a look at speeding up the point-to-circle test: "is any random point not inside an existing circle, and far away of it to get drawn at a sufficient size (plus any inter-circle spacing)?"
Rest assured, all of this is actually really big fun for me!
A thought regarding your issue of checking distance - If there is a maximum circle size, can you limit the distance each circle checks to that? Maybe not the best answer, but it would save at least a good chunk of the time, right?
I'm sure it's just that simple, anyhow. 😉
Thanks for this bit of awesomeness.
I wasn't sure on how to proceed with speeding up the main time-guzzling parts -- checking for the outer edges, and checking for circle-circle distance -- so I asked on Usenet in the comp.graphics.algorithms group. The regulars there are a varied bunch, experts in just about any theoretical part re: graphics. I got two good leads already: creating a Voronoi diagram of the outer polygon, and using Possion Disk Sampling to generate the circles themselves.
Graphic theory is utterly fascinating -- my script might be a nice introduction for (hopefully) the few Illy users interested in the backgrounds
Are you JW?
Just found the script a few days ago and i want to use it for a Ishihara Colortest.
Everything works fine - either filling the background with cutted out letters.
I get this error message:
Can you please help me asap? I need it for a project on Tuesday, 1st November.
Thanks a lot! Lisa
What do you do to get that error message? Looking at your images, it seems to work correctly.
Perhaps there is a single stray point somewhere in your selection – a "path" that consists of a single point only? That could explain the error message.
I get this in CS3 when I run this script:
A quick peek in the ESTK under Illustrator seems to indicate CS3 can't script swatch groups ...
The idea of using a swatch group was to get random colored circles -- if you don't need that anyway, I can take out the entire swatch group selection to make a CS3-compatible version.
That would be excellent, thanks.
Peter, I made a quick change in the script -- now it checks for a version "pre-14" (which is, so it seems, the official name of CS4). If so, it doesn't attempt to add color swatch groups and always will use Black. I have no CS3 over here, so I can't tell if that's the only thing that needs changing. You could wait for tomorrow or test it yourself: http://www.jongware.com/binaries/CircleFill-1-2.zip
Jay: creepy! If I find one like that crawling around I'll make sure to trap it in a box and mail it to ya!
Nope, it still doesn't work. I think it's the same error message, line 20. By the way, that's a dud link: should be CircleFill-1.2, not 1-2.
Message was edited by: tromboniator
Three minutes behind a real CS3 and I got it. Just ignore color groups for pre-CS4 versions ...
Nope, it still doesn't work. I think it's the same error message, line 20.
Yeah, changed it at one place where it should have been the other.
By the way, that's a dud link: should be CircleFill-1.2, not 1-2.
Yeah, it was gettin' late!
http://www.jongware.com/binaries/CircleFill-1.2.zip is fixed, ought to work now for CS3.
Wunderbar! Thank you!
Thanks for circles and beetlebum -- great together. (Got to get back to real work now!)
Im trying to use this in CS5 but when I change the text to outlines then i try and run the script I get this notice and then it never works
Well, I can't test with '5, so it can be just about everything.
Can you test if it works with a simple shape -- a circle or a rectangle? If it does, there is something different about the convert-text-to-outline system in CS5.
The text is just grouped (even if it's a single letter) that's all. Ungroup it and everything works fine.