translating @redFrick's tweets

[Roger Allen]

I don't know where I got the pointer to this guys tweets, but he's making some amazingly dense, fun tweets of Supercollider code.  I picked this one to translate and learned a few things from the attempt.  I thought I'd write it up here for those who might be interested.

The tweet is here https://twitter.com/redFrik/status/333317729073381377 

play{a=LFSaw;mean({|i|Ringz.ar(Blip.ar(a.ar(i+1/[3,4])>(a.ar(i+1/8)+1)*25+50,i+[2,3])*a.ar(i+1/50,i/25),i+1*99,0.1)}!50)/5}

Go plop that into Supercollider and shift-enter to hear it.  Neat, huh? 

Okay let's try to translate it with some reformatting...

;; play {
;;   mean(
;;     { |i|
;;       Ringz.ar(
;;         Blip.ar(
;;           LFSaw.ar(i+1/[3,4]) > (LFSaw.ar(i+1/8) + 1) * 25 + 50,
;;           i+[2,3]
;;         ) * LFSaw.ar(i+1/50, i/25),
;;         i+1*99,
;;         0.1
;;       )
;;     } ! 50
;;   ) / 5
;; }

I couldn't find "mean", but I think "mix" does the same thing after a few experiments.  At first, I had no idea what |i| might mean, but I found out (only by looking in the Supercollider book I bought) that { |x| x }.3 creates a vector of [0 1 2].  Happily, that translates to a (for [x (range 3)] x).

One thing that totally screwed me up for a while is that precedence in Supercollider is strict left-to-right (ugh, as if that language wasn't crazy enough!).  So, i+1/50 is really (i+1)/50.

Along the way, I got this message with the default overtone settings

 exception in GraphDef_Recv: exceeded number of interconnect buffers.

A google found the workaround was to increase the wire-buffers, so add

   :sc-args { :max-w-buffers 512 }

to .overtone/config.clj but that still doesn't get it working.  If the for loop exceeds 39 you get

   Exception Invalid ugen tree passed to topological-sort-ugens, maybe you have cycles in the synthdef...
   overtone.sc.synth/topological-sort-ugens (synth.clj:415)

So, I think there is a bug in Overtone--Sam, should I write an issue on this?  The code works fine up to 39, but 40 and higher causes that failure.

Okay, here is the [mostly] final result...it takes a full minute to compile for my i7 MacBook Pro, so be patient...

(defsynth red-frik-130511
  []
  (out 0
       (/ (mix ;; docs suggest this is the same as 'mean'
           (for [i (range 39)] ;; 40 starts fail, see above
             (ringz:ar (* (blip:ar (+ (* (> (lf-saw:ar (/ (+ i 1) [3 4]))
                                            (+ (lf-saw:ar (/ (+ i 1) 8)) 1))
                                         25)
                                      50)
                                   (+ i [2 3]))
                          (lf-saw:ar (/ (+ i 1) 50)
                                     (/ i 25)))
                       (* (+ i 1) 99)
                       0.1)))
          40))) ;; reduce volume to avoid clipping

(red-frik-130511)  

sounds very similar to the equivalent code in supercollider...

play{a=LFSaw;mean({|i|Ringz.ar(Blip.ar(a.ar(i+1/[3,4])>(a.ar(i+1/8)+1)*25+50,i+[2,3])*a.ar(i+1/50,i/25),i+1*99,0.1)}!39)/40}

That was a fun little challenge--go check out @redFrick's tweets.  You can also see my travails in the gist I used here https://gist.github.com/rogerallen/5563902

Cheers,

Roger

[Sam Aaron]

Hey Roger,

great work. I haven't had time to go through it all in detail yet, but a couple of quick points:

* The limit you're running into w.r.t. topological-sort is just a hard limit (placed to detect cycles). This can easily be increased (I'll do this for the next version)
* Taking a minute to compile the synthdef is also a bug - we shouldn't be taking any longer than SuperCollider itself which does it in less than a second.

Feel free to file some GitHub issues...

Sam

---
http://sam.aaron.name

> --
> You received this message because you are subscribed to the Google Groups "Overtone" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to overtone+u...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Sam Aaron]

Also, for those interested, I've attached the GraphViz version of Roger's translation of Frederik's tweet - it's pretty intense!

[Roger Allen]

That's fantastic, Sam.

I've put the issues from this thread into https://github.com/overtone/overtone/issues/231

and I've translated a couple more tweets (MUCH easier after the first one) here

https://github.com/rogerallen/explore_overtone/blob/master/src/explore_overtone/redFrik.clj

and I'll add more to this as I go through his tweets. 

It seems like this project

https://github.com/Engelberg/instaparse (also see http://gigasquidsoftware.com/wordpress/?p=689)

would be generally helpful in making these translations automatically.  I'm not sure I'll get to this, so if anyone else wants to take a run at this--go for it!

And finally (I do run on...) these synths gave me the motivation to figure out how to allow shadertone to monitor taps inside of synthesizers.  On my tree, I have code that allows the "t0" tap from the t synth instance to be sent to the t0 uniform variable  in the t.glsl code via:

  (t/start "examples/t.glsl" :user-data {"t0" (atom {:synth t :tap "t0"})})

I had high hopes that I could make some thing interesting with a redFrik tweet shader, but making something that won't induce epileptic seizures is going to take some time.  I should be able to create a simpler demo to show how to do this, though...

Roger

[Roger Allen]

I pushed this feature out to the shadertone repo[1] and spent some time making a "music video" demo with a redFrick tweet that now comes up when you type "lein run".  I continue to be astonished at how interesting < 140 characters can be musically.  I can't get the dang tune out of my head!

Some screencaps to entice you to try it out.  :)

--Roger

[1] https://github.com/overtone/shadertone