Cycling 74 » Topic: Bjorklund algorithm

Bjorklund algorithm

withakay — Fri, 14 Oct 2011 10:20:56 +0000

Hi all,

I implemented Bjorklunds algorithm to Javascript for use in Max and thought I would share.

http://gist.github.com/1286731

I know there exists some other implementations for various languages around the web but I found that most were broken for more than 16 steps so I fairly closely ported the original c implementation.

Hopefully someone will find it useful

Re: Bjorklund algorithm

roger.carruthers — Fri, 14 Oct 2011 12:49:11 +0000

Thanks – I can’t look at it right now, because the network police have banned me from using Max at work (really!), but thought you might be interested in this post, if you hadn’t already seen it:

http://www.cycling74.com/forums/topic.php?id=16979

There was also a js Max implementation – I have the patch here, but can’t open it of course, and I can’t find it on the forum, though I think it may have been by Axiom-Crux ?
cheers
Roger

Re: Bjorklund algorithm

Timo Rozendal — Fri, 14 Oct 2011 13:09:14 +0000

haha, I also made an implementation of this bjorklund algorithm (in java) and I plan to share that later too, but you beat me to it :-)

Re: Bjorklund algorithm

withakay — Fri, 14 Oct 2011 14:00:22 +0000

@roger.carruthers Thanks for the linkage. That ruby code looks very similar to a lot of the other implementations that I found to not work with more than 16 steps (although for 16 or less steps they work fine and are more efficient).

A few of the options are listed here http://createdigitalmusic.com/2011/03/circles-and-euclidian-rhythms-off-the-grid-a-few-music-makers-that-go-round-and-round/

most look to be based on the same logical implementation (across different languages), but whatever was used as the base implementation that has subsequently been ported does not follow Bjorklunds original c version. That is, as far as I can tell, I am by no means an algorithm expert or particularly at maths! But as I said I have noticed funny results when trying to create patterns greater that 16 steps with all but my version.

@Timo Rozendal

If you have some java code please make it available, I was possibly going to make an mxj to see if it performs much better, although I am hoping that the new Javascript engine in Max 6 might solve that issue for me!

Re: Bjorklund algorithm

Marcos — Sat, 20 Oct 2012 03:31:18 +0000

Downloaded this as the other implementations have errors in them…How do you use this in max? I saved the file in the folder I’m working in and used [js bjorklund.js] … According to the source it’s one function (bjorklund) that takes 2 arguments (steps, pulses) and returns an array…I’ve tried sending a message “bjorklund 8 3″ to the object but nothing comes out the outlet. Any ideas?

Re: Bjorklund algorithm

Marcos — Sat, 20 Oct 2012 03:48:33 +0000

Nevermind… I had to:

1) initiate the object such as [js bjorklund 8 3]

2) edit the .js and at the end instead of returning the value send it to outlet 0

3) call the function as: “bjorklund 8 3″

Re: Bjorklund algorithm

cudnylon — Wed, 06 Mar 2013 06:29:11 +0000

I’m having trouble getting this to work in max and Marcos instructions leave me confused. new to the js object and javascript coding in general.

Re: Bjorklund algorithm

Anthony Palomba — Wed, 06 Mar 2013 20:27:12 +0000

I updated the bjorklund example and made it a little more user friendly…

Re: Bjorklund algorithm

roger.carruthers — Wed, 06 Mar 2013 20:39:04 +0000

I’m getting ‘js:no function generatePattern’ in the Max window,
Cheers
Roger

Re: Bjorklund algorithm

Anthony Palomba — Wed, 06 Mar 2013 20:58:38 +0000

Are you loading the bjorklund.js I up loaded? generatePattern is definitely there…

Re: Bjorklund algorithm

roger.carruthers — Wed, 06 Mar 2013 21:33:04 +0000

Woops, yes the old one was still in my search path!
Cheers
Roger

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 05:13:18 +0000

thanks for the remake!

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 05:41:06 +0000

weird, the fix either breaks the code or there is something wrong with the original
the code’s output is offset for certain pulse values. for “generatePattern 16 5″ i get
[ . x . . x . . x . . x . . x . . ] when the article says the expected result is
[ x . . x . . x . . x . . x . . . ] “generatePattern 8 5″ causes the same offset.
weirdly, the starting example of “generatePattern 8 3″ works but “generatePattern 16 3″ does not
i went back to withakay’s original code
replaced only “return pattern.reverse();” on line 63 with “outlet(0, pattern.reverse());”
the error remained the same. going to fiddle around with it and see what’s happening.

i’ve used the java code from http://doc.gold.ac.uk/~ma801dp/blog/?p=40 in the past
might go back to that for now if a few test values come back correct.

edit: jurgen schmitz’s m4l device works while using this code because the js ends with
———————————-
build(level);

pattern = pattern.reverse();

// make sure pattern starts with a ’1′
for(s = 0; pattern[s] != 1; s++);
pattern = rotate(pattern, -s);

return pattern;
———————————-
but was designed to work specifically with his device, as just copying out the original portion gives “js: bjorklundFIX2.js: Javascript ReferenceError: rotate is not defined, line 75″
the fiddling continues….

Re: Bjorklund algorithm

bertrandfraysse — Thu, 07 Mar 2013 17:54:48 +0000

jurgen’s idea about rotating the list until it starts with a 1 is great, simple and efficient.
isn’t it strange the rotate function doesn’t work in js ? there should be a way…

Re: Bjorklund algorithm

Anthony Palomba — Thu, 07 Mar 2013 18:21:14 +0000

Actually, I have been looking a little more closely at the output. It seems to be doing
exactly what I would expect. Given the following sequence of input, you get the following
output…

8 beats 3 pulses
output: 1 0 0 1 0 0 1 0

16 beats 3 pulses
output: 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0

8 beats 4 pulses
output: 1 0 1 0 1 0 1 0

So what this is really doing, is taking a pattern of beats and evenly distributing
another pattern within it.

I have updated the helper patch to allow for easier control of pattern lengths
and note division. By adding two separate clicks you can really hear the difference.

– Pasted Max Patch, click to expand. –

Copy all of the following text.Then, in Max, select New From Clipboard.

----------begin_max5_patcher----------
2413.3oc4bszjihiD9rqeEJHlCyNgWGnG7ZOsO9CLGla8LQEXrZWTEFH.4p6
omX5e6qd.XvFDxtswTUcnrcIDnL+TloxLUJ9qGVXsN6qzRKv+B7IvhE+0CKV
HaRzvhp+eg0tvuFkDVJ6lUT1tczTl0R00XzuxjsiWAJoIzHFHMiQAaheMtLN
Kste4grnmhS29XAuKpwC65txyYI.QHq3eAcrWYK9uU1f+n5tVue85Dpn6vpV
h2HGsr0O+OcH0O7OmkxRC2I6n0+oHNLo4JEaWKZU7nUsjteWbZBkU19oJd.k
weSMRn18MaOqty1hF+6GdP7wRCQq8brZuNLvFqXZID33o.D6VPPaFFMNCmGV
vamQKdjlFVgc+3rNtlXXzcURKVD9bKvZo3C+C+D5d32XTpUCendZr+LmVc+w
BgHfk7OvebAXaJ8KbT4DAQN2WjAHof+cXDK9UJ.pA+QdAJH2SNO.gAxoiAl.
vdiOAzAjQ5A47BZIWWJjwUTZSTHGIQg8CV4X6SP3kRAXPaxp6DDbHTdcX51q
J79KemSFHGMXpmiDCcjPIQOf5dcATCAkx3so7A3hfk06YLsl0PAsXejsV1mX
oS0b.wCGjZ.7qDIlGxEQIwQu7c.bEmZFW3.K4AWesniyYJb.u6BG7gXMsv.6
MXLQv5N5EOv2K68nQLZe4RO6nkkgaomH9rklRKBYzeMjwYlTvOAA+DRCPBU5
WAtqrqrPBwRvD5O.XBmDa2PDTRW1h0yuTkyakvGTp7w80HH.534Mp7m86O4u
QAoJAKiAID7cHH0uI97vW.9.73fmmcieK5AOn8cY8+KCUFJxG9Rd0Vu.qogL
PtxFlNbh3IsUgTA93ddA93O+i6YHrZC8yg6SXf3Rfutvh7TgFVEdDTuTjy7G
OFRiB7e4BL+ur8oLybaPgG9Z8ax04tnTIr5zPNRdfyRJDoBR3iPbB8UZgLw.
GdVKrByya07hV2h.GeNS9fbW1zTbppI6llJnU4af2pupQ9ryC0WrAHkq.5Js
R4RjRUD+V3HedbaRVzKzMse7VY4zz3z1K424xUR0O1O.185eNLhN3M26j0Bq
sEwaxREDQm6Tzb8vIr8pTUZyLxdjFl2yMWx4i8kqCKD3VksFT8EYYYIcuTyT
EWNOLMdG2XGKVQrH6lGZ7t7BobPqARsj2SkQEYIIcdTpq7ZOWYCe1Lh9k3Mr
m5n0JtBu6w40SAVMXzl3szRV21XgaK61Rmzc0VLqsBbm1ORQVp0T+3Fx5tqz
5dcbZ0e1LsbraVsebcTLs6dgS0BGWST1iZqwhQjEVrkxjKTYU2CowrJUlKDW
TDwX.SUrwXCQFxfHCbPjwd.99ZxrlHD33bNx.nqsLPkyeS.XzIgHChFpXyrc
abL7zLibx5XWxzuIfxUj46DNwHh8PQ5NFzs3S39NhDCu3vHwWnG6FXAqAi0v
z3MtRna2jEbjW2i.59JmpsQMRbXk219Zvbayw797f57vTshpSGPVxWMXLvzy
6X0Wr2wIa4Dux8emBlc7k+XrTtCX9lYMzCeHLYzX1CbbmR3DMpYzkSHrFFEs
e2nBoUnJ9vllTEXsNT04BQU7UWHU40VO.5A24ShSGx6UIdHtd+HcY19hn5Qp
xoSPWTg6YMKNsI3lOcPrC.6zwmh2roavDJ+uKEK8roIB4QEKNWRV3OfYzryI
c7tQyHCoYx7gjsOGXddHZHVq4slnQOZV8Sy8LgbWoYivY+4CM6XHM2it5rml
6YB4dQylZoy8ZYpqZwx5z7YUF9Jcyi7gf6QvigLVQLOzT05ksyA4OZx3FOYe
mQFp5kHzL3CMnaSxVGlTk00l6smDg8vAP7JkBaQB+.qeNq3kj8oaFeygD68A
ONLeWsoz29Vt2+1WY1WCSG3TGFJ+K0l9ODOSBdivy7oaPrFdVUyOU0RkJp6A
4Y+aIOizV8KKa6y80AYJ3iDkabCNd8AojHbztSNDzYhN3Kon9NMb4aQY48sD
PJ6oQwkpMCEqsrwHva4NbgzhKW0sN92sJneINcyuaoqjo7jA155zDWqG971v
X7a2MLFsBTxBKXfvzM7ekkCXOQArhvzx7rBl18PVgRGR.Hzmbd3Fxa9CbiVO
h0kYsOVSt26v0tZs7dyKvPV11sITcbjCtd+TLigHWbg3bV688Urb3zWCkXXU
0S1jMVhtReC8dnJaD61LPValr8E5jNbqpUBYQiTgO9ZWNAZfZtfNhxRTkcvm
.hBbBF3BW16ufsFm4pgUFcWdF3mWmu6enALcHHUppkdw4izBiteDgw3nWJ05
Kr+JjSCBpOB.n+GPDbeZLSKBh8Wc..IPs.H4CH.JJkiRCJlNnGThf5UhC9Hh
fgEZAPX092KSbAVahKfnOf3WAU3KNHZeQA+xfWCS1S0BnApkSfGJd+piBh4t
l+dFmGxqPBfnywGDd0Q98P7z5WH9MjegeNIi+704phJyAsYeG8q1ZhCKSbo6
y4xvevh2eTfh3W4RhJQriAR3OlfDO5RiwHa+2iXznGUl5LKTCR5crv18C3oI
BCgs1yC8HD1+8G.st9f.nwtspZoammNWY5bvCfSMdn1.Qhha+2pSI3JNAjGx
5cAra9VibVG97Z8GmVrt9CfN59bzNt5mjLixeqJKeP6fl0ywAdZQG+aI5DLh
RR0WMFdO8G80qazgPWsEbXaz3GB8CKveeR2K8aaBi99nGYbrhkHN0e1u2JZm
muYp8sO76hXJTeYvaIg5C3FR+I55M2Ag2zi4s9iSZ0I5V4IFTarLj2PgxLfP
zykGJihUOWN5Vk5IwEerVk66hjygyAng0ji3bA1Pc8.BC33yoEyhbf6VmrpY
jiqlop4kSqhoNUmz.kvjvLSqNcR4KcZoKchDhozi3UNynziqyjQOh2AQGpZw
AnGIQCmF7ww.7A6NcyWtlJ+.mYzyzfODjIyWjoa9xD8c7zoe4Zj7LdxnGDzT
6OSi7Lz1D7ANuvmiH5Somcwaxy3gFTVe.hka4EOvXYd5Z8eGFnqmFfsob.zb
N.gHs4fffaJGXzbv7REl3Oczy7ZEfY1BjGMQLvr0Dt.ouANXQlP7wDG9fS2B
RDST1EkR8rhdPSG8LqrEhLQ6B6OsziyHziXeQmHsKI8PF0Zn6zBP3QIH7zRP
nQcOYhEgF0fn8DNkYxJpnIDfLw+GzDhOjYF9XaJ8LQgTgMIDuICdLY8TG6YE
4bwtGpRW4Qu8yDTxQu0yN5Md1wusy3Os+9g+OVpZzCB
-----------end_max5_patcher-----------

Re: Bjorklund algorithm

bertrandfraysse — Thu, 07 Mar 2013 18:44:40 +0000

when using more than one generator at the same time, it’s nice to have a synchronisation point, which is the start of each loop.
if every loop start with 1, it feels more musical, and rotating the list doesn’t change the evenly distribution.
it’s a kind of normalisation.

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 18:52:10 +0000

bertrandfraysse, not so fast. it seems jurgen’s code is flawed as well.
for example E(5, 8)
[ . x x . x x . x ] = bjorklund code output
[ x x . x x . x . ] = jurgen’s rotate fix (from ADesso ERG v2.3)
[ x . x x . x x . ] = expected result (from Toussaint’s paper)

but for an example like E(5, 16) the pattern from the m4l device is correct.
so for some pulse values the array needs to be rotated positively, others negatively.

Re: Bjorklund algorithm

bertrandfraysse — Thu, 07 Mar 2013 19:01:42 +0000

maybe it feels more natural to have 1 0 as a beginning than 1 1.
so we need to make our algorithm to try and find the right rotation to find 0 after 1 if it can…
just an idea…

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 19:28:15 +0000

i feel like i’m overcomplicating this, especially as i’m only planning to use the algorithm for generating quick drum loops to be refined later. going to stick to the current code and put a zl.rot at the end to adjust the pattern as i like. thanks again for the help, anthony.

Re: Bjorklund algorithm

David — Thu, 07 Mar 2013 19:50:00 +0000

Try this code,that I wrote it for my euklid-sequencer m4l-device http://www.maxforlive.com/library/device.php?id=993 . The outcoming list always starts with a 1. I’m not an experienced js-coder, so there are probably smarter ways to compute this list, but it works ;-).

inlets = 1;
outlets = 1;

function bjorklund(pulses, steps)
{
	if (steps >= pulses && steps > 1)
	{
		var pauses = steps - pulses;
		var rythm = new Array();
		var colSizes = new Array();

		//build arrays
		for (var i = 0; i < pulses; i++)
		{
			rythm.push(1);
			colSizes.push(1);
		}
		for (var i = 0; i < pauses; i++)
		{
			rythm.push(0);
			colSizes.push(1);
		}

		//actual algorithm
		while ( !((pauses==1 && pulses==1) || (pauses==0 || pulses==0)) )
		{
			var counter = 0;

			if (pauses>pulses)
			{
				counter = pulses;
				pauses -= pulses;
			}
			else
			{
				counter = pauses;
				pulses -= pauses;
			}

			var writePosition = 0;

			var level = colSizes[colSizes.length - 1];

			for (var i = 0; i < counter; i++)
			{
				writePosition += colSizes[i];

				var readPosition = rythm.length - level;
				var elementsToMove = rythm.splice(readPosition, level);

				for (var j = 0; j < elementsToMove.length; j++)
				{
					rythm.splice(writePosition, 0, elementsToMove[j]);
					writePosition ++;
				}
				colSizes[i] += level;
				colSizes.pop();
			}
		}
		outlet(0, rythm);
	}
}

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 21:12:13 +0000

sorry david but your results are coming back out of order.

example (3, 8)
[ x . . x . x . . ] = your code
[ x . . x . . x . ] = Bjorklund code & Toussaint’s paper

example (5, 8)
[ x . x x . x . x ] = your code
[ . x x . x x . x ] = bjorklund code
[ x . x x . x x . ] = Toussaint’s paper

same for most other values i tried. wish i could learn js code quicker so i could offer something useful other than criticism. sorry.

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 23:01:07 +0000

edit: SEE BELOW

Re: Bjorklund algorithm

cudnylon — Thu, 07 Mar 2013 23:32:29 +0000

well i’ve almost cracked it. this new code accounts for the list rotation (either positively or negatively).

so far the only example error i’ve found is E(5,13)
[ x . x . . x . x . . x . . ] = my code
[ x . . x . x . . x . x . . ] = toussaint’s paper

if anyone finds another let me know. not sure how to go about correcting that but for my use it won’t be an issue.

save the new bjorklund.js file i’ve attached and use the below patch for testing

– Pasted Max Patch, click to expand. –

Re: Bjorklund algorithm

David — Fri, 08 Mar 2013 04:49:06 +0000

Yes, I think you are overcomplicating this a bit ;-). The output of all three algorithms are valid as they evenly distribute the events over the number of steps. They are all just rotations of each other, and I would always include some kind of rotation-parameter in a device anyway.
For beats I just find it much easier, if the list always starts with a 1 and not a 0, which is the reason why I wrote my code.

Re: Bjorklund algorithm

Timo Rozendal — Fri, 08 Mar 2013 09:42:05 +0000

and here is my quick and dirty (string manipulating) way in java

– Pasted Max Patch, click to expand. –

I like how with verbose on you see the process in the max window :-)
I should have posted this a year ago, but I forgot, these recent messages in this forum reminded me of this