beat function for "step sequencer" style rhythm notation. (Backported from Strudel)

[geikha]

See tidalcycles/strudel#1237

Luckily the fractional implementation done by @yaxu doesn't require any special handling of polyphony, it just works!

beat :: Pattern Time -> Pattern Time -> Pattern a -> Pattern a
beat = patternify2 $ __beat innerJoin

__beat :: (Pattern (Pattern a) -> Pattern a) -> Time -> Time -> Pattern a -> Pattern a
__beat join t d p = join $ (compress (s,e) . pure) <$> p
                      where s = t' / d
                            e  = (t'+1) / d
                            t' = t `mod'` d

[sss-create]

Wow, this looks interesting! Can you tell me what it does? I did not quite get the hang of it in the strudel thread + I couldn’t listen to the audio

[yaxu]

struct (beat "[0,5,12]" "16") $ s "bd" would trigger on the first, sixth and 13th beat of a 16 beat cycle. It's a bit weird that the beats are 'stacked' like this but makes some sense..

[geikha]

@yaxu i think you meant beat "[0,5,12]" "16" $ s "bd", as the function restructures the pattern already

also, i like that we can easily beat "[0,5,<12 11>]" 16 and stuff like that

[geikha]

if anyone wants to try this on the current version of tidal without installing anything you can just run:

__beat :: (Pattern (Pattern a) -> Pattern a) -> Time -> Time -> Pattern a -> Pattern a
__beat join t d p = join $ (compress (s,e) . pure) <$> p
                    where s = t' / d
                          e  = (t'+1) / d
                          t' = t `Data.Fixed.mod'` d
beat :: Pattern Time -> Pattern Time -> Pattern a -> Pattern a
beat = tParam2 $ __beat innerJoin

[yaxu]

@yaxu i think you meant beat "[0,5,12]" "16" $ s "bd", as the function restructures the pattern already

also, i like that we can easily beat "[0,5,<12 11>]" 16 and stuff like that

Ah hmm a binary version might be more flexible, but lets park that thought for now and merge :)

[geikha]

agreed! i was just copying the behav selected in strudel. i also thought about euclid as reference, which always takes in a pattern to "euclidize" heh
noted that if a binary version was made, i'd expect emptyness to be filled with False instead of silence, wonder if there's an easy solution for that without changing the curr implementation that much

[cleary]

I think of this like a flexible Euclidean notation - rather than evenly spreading the events over the increments you just define where you want them to fall.

Its easier than counting out the steps in binary notation too