Make a usable interface for the link protocol via hylia

[yaxu]

Ref tidal ticket tidalcycles/Tidal#660

and short discussion here Ableton/link#71 (comment)

Lets make Haskell bindings to the lovely https://github.com/falkTX/Hylia

I don't really know what I'm doing, but the current WIP in this repo seems to be basically functional.

alex@teff:~/src/hylia-haskell$ cat examples/test.hs 
import Network.Hylia

import Foreign
import Foreign.Ptr
import Foreign.C.Types

main = alloca $
  \timeinfoPtr -> do hylia <- hylia_create
                     hylia_enable hylia (CBool 1)
                     hylia_process hylia 1 timeinfoPtr
                     timeinfo <- peek timeinfoPtr
                     putStrLn $ show timeinfo
                     hylia_set_beats_per_bar hylia (CDouble 3)
                     hylia_process hylia 1 timeinfoPtr
                     timeinfo <- peek timeinfoPtr
                     putStrLn $ show timeinfo
                     return ()
alex@teff:~/src/hylia-haskell$ cabal v1-install
Warning: The package list for 'hackage.haskell.org' is 64 days old.
Run 'cabal update' to get the latest list of available packages.
Resolving dependencies...
In order, the following will be installed:
hylia-0.1.0.0 (reinstall)
Warning: Note that reinstalls are always dangerous. Continuing anyway...
Starting     hylia-0.1.0.0
Building     hylia-0.1.0.0
Completed    hylia-0.1.0.0
alex@teff:~/src/hylia-haskell$ runhaskell examples/test.hs
TimeInfo {beatsPerBar = 4.0, beatsPerMinute = 120.0, beat = 2.48e-4, phase = 2.48e-4, playing = 0}
TimeInfo {beatsPerBar = 3.0, beatsPerMinute = 120.0, beat = 3.368e-3, phase = 3.368e-3, playing = 0}

[yaxu]

Things to work out:

• How to make a more haskell-ish api?
• Integrate with tidal's scheduler..
• What's that 'frames' parameter to hylia_process
• I hardcoded some output of hsc2hs to get access to the timeinfo struct.. Hopefully that structure won't change..

[yaxu]

Help is welcome!

[Zalastax]

I see that you got started on the 'frames' parameter in falkTX/Hylia#2! I know Haskell, some C++, and would like to help out! Would you like help in falkTX/Hylia#2 or with the haskell-ish API? If so, what help would you need and what initial thoughts do you have, to set me of? Tidal integration should wait until we have those new calls in Hylia, right?

[yaxu]

Hi @Zalastax ! I haven't thought about this in a while. It seemed to me that hylia is really oriented around a particular model for audio apps, and it might be better to start afresh with a subset of link functionality for tidal (as a non-audio app), basically mirroring the link implementation for the supercollider language that I pointed to in that other issue. Is your own motivation to get link support in Tidal, or more generally to widen the scope of hylia to support projects like tidal? Would be happy to work on this in either case.

[Zalastax]

Hi @yaxu, my motivation is to get Link support in Tidal. I'm mainly interested in letting Tidal be a midi clock slave, and I found suggestions that implementing Link support would be a good way to get there. Starting fresh seems smart, since Hylia did not seem like a perfect fit! Would you like to tailor it for Tidal from the start and just add this to the Tidal repository? Or would you prefer to have it as a separate library? I'm ready to dig in, but I want it to fit your vision, so I'd prefer if you set the direction. Cheers!

[yaxu]

@Zalastax I don't have strong feelings about this. For simplicity we could put it in the tidal repo, but try to keep it fairly decoupled so that it could be separated later if there is a demand for that. Does that sound ok?

Currently there is Tempo.hs and Stream.hs taking care of timing. The latter in particular is quite messy and I would be happy to rewrite/refactor that based around Link. Then Link clock sharing would 'just work' by default. That'd be so good.

Tidal calculates events at a fixed rate of 20Hz (by default). So 20 times a second it needs to know what the current beat number and bpm is according to link. It also needs to be able to update link with changes.

There are complications because patterns can change the tempo via the cps parameter. This is taken into account while scheduling, so the tempo can change a lot during those 1/20 second frames. For accuracy and efficiency it should probably only commit the final tempo change back to link. It gets extra confusing because tidal calculates events ahead of time (I think 0.2 seconds ahead by default), so that tempo change should probably be scheduled a little bit in the future.

Anyway, feel free to dig in! I'll be travelling next week so might be a bit slow to respond, but do shout if you have any questions or could do with some help untangling things..

[Zalastax]

Thanks @yaxu, this is a great help to get started! I've managed to compile Tidal and the Link example projects. Link was a bit of a hassle due to me being on Windows. Next, I'll try to add a simple C source file that uses the Link library to the Tidal cabal file, or possibly to a new sub-library. Safe travels!

[fbous]

Hi @Zalastax, @yaxu,

I'd like to help as well! I had a look at Stream.hs last week-end when I was trying to understand the Carabiner implementation so I have a rough idea of what is going on there. Unfortunately I'm busy the next two week-ends but I'll try to follow along and hopefully can start contributing sometime soon. Let me know of your progress!

From what I got from this discussion we're planning to implement Link interaction in the Tidal repository? In that case should we continue this discussion on issue #660?

Best,
Frederik

[yaxu]

Welcome @fbous! Yes good call, lets continue over there.