Casper Leerink

I just had a general question if someone ever tried to use gen for fast data manipulation outside of the audio/video domain. It would be great if Cycling 74 could add a feature so you can access dict or table objects in Gen. I was trying to make this work with the Data object and peek and poke, but it seems to be hard to apply multiple methods to the same gen instance, for example if you want different methods to get and set one index in the Data object it always both sets and gets the data.

Hi everyone!
I just had a general question if someone ever tried to use gen for fast data manipulation outside of the audio/video domain. It

gen-without-~-for-data-manipulation

the data object works so much like buffer, you could also just use the regular MSP object 'peek~' outside of gen entirely(see helpfile for 'peek~' - it can work outside the audio/video domain in the event domain)

non-audio gen can then just be used for processing in-between get and set(if you are looking for everything all in one gen, it then becomes more like the mxj or js objects, so you might try just using one of those instead - depends on what you want to do, though, if you post a patch people can be more specific with advice).

the data object works so much like buffer, you could also just use the regular MSP object 'peek~' outside of gen entirely(see helpfile for 'peek~' - it can work outside the audio/video domain in the event domain)

non-audio gen can then just be used for processing in-between get and set(if you are looking for everything all in one gen, it then becomes more like the mxj or js objects, so you might try just using one of those instead - depends on what you want to do, though, if you post a patch people can be more specific with advice).


Thank you for your reply! I will look into peek~. I am just looking for a faster and more reliable way for data processing than js. I am using the node.script now to process data that is in the form of: list of lists of amplitude and frequency pairs. I am analyzing audio that creates 32 amplitude/frequency pairs for each frame (amount of time between each frame can be varied). I am putting those into a list. I then getting amplitude/frequency pairs pair by pair and find the next by calculating the difference in amplitude for each pair in the two frames and choosing the frame with the least amount of difference. In node script without the difference calculation function it looks like this:

//get index, frequency and amplitude from part of current frame 

if (currentPair < currentFrame.length - 1) { // this also takes account for start since currentPair is never less than empty list length

currentFrame = frames[0]; //begin with first frame

frames.splice(currentFrameIdx, 1); // remove the used frame from the list, so it isnt used again

const deviation = util.ampDeviations(currentFrame, frames); //this returns the new frame with the least difference.

Max.post(deviation.deviation); //post the difference to max

currentFrame = frames[currentFrameIdx]; //set new frame

1. I have to use a separate language outside of max, which gives bugs that I don't know how to solve, for example when I save my patch as a m4l device Ableton crashes. I think the patch would be more stable if I can do all the data processing within Max itself.

2. With Gen I thought I can make this process less CPU intensive. Long lists of frames of 10000 can become problematic at some point. (although node.script is a big improvement over the js object)

Anyway, I might try some method with using the dict object instead, I just wish there was some better way to work with data directly in max without having to learn C or C++.

//get index, frequency and amplitude from part of current frame 
Max.addHandler("getNextPair", () => {
 if (frames.length > 0) {
 if (currentPair < currentFrame.length - 1) { // this also takes account for start since currentPair is never less than empty list length
 currentPair += 1;
        } else {
 if (currentFrame.length === 0) {
 currentFrame = frames[0]; //begin with first frame
            }
 frames.splice(currentFrameIdx, 1); // remove the used frame from the list, so it isnt used again
 const deviation = util.ampDeviations(currentFrame, frames); //this returns the new frame with the least difference.
 currentFrameIdx = deviation.idx;
 Max.post(deviation.deviation); //post the difference to max
 currentFrame = frames[currentFrameIdx]; //set new frame
 currentPair = 0; //reset pair to index 0
        }
 outputPair(currentPair, currentFrame);
    }
});


I have two problems with this method:
1. I have to use a separate language outside of max, which gives bugs that I don't know how to solve, for example when I save my patch as a m4l device Ableton crashes. I think the patch would be more stable if I can do all the data processing within Max itself.
2. With Gen I thought I can make this process less CPU intensive. Long lists of frames of 10000 can become problematic at some point. (although node.script is a big improvement over the js object)

Anyway, I might try some method with using the dict object instead, I just wish there was some better way to work with data directly in max without having to learn C or C++.

Thanks! Casper


Hi Casper, I will be releasing a new version of Scheme for Max in the next few weeks with direct buffer and table writing, in case that is of interest. It works much like the JS object, except runs S7 scheme-lisp. It might be helpful if by "reliable" you mean the timing problems that one gets from JS running always in the UI thread. In my benchmarks so far it's about 20% faster than JS, but more importantly, runs in whatever thread you trigger it in, so can run in the high priority thread. It is also open source, so if you are comfortable recompiling, you can add a new scheme function that runs as compiled C pretty easily to allow one to script high performing functions from a high level language. The scenario you describe was one of the motivations for making it actually, so if you're interested, let me know and I can help you get going, it would be great to have someone testing the buffer functions in real world situations.

https://iainctduncan.github.io/scheme-for-max-docs/

Hi Casper, I will be releasing a new version of Scheme for Max in the next few weeks with direct buffer and table writing, in case that is of interest. It works much like the JS object, except runs S7 scheme-lisp. It might be helpful if by "reliable" you mean the timing problems that one gets from JS running always in the UI thread. In my benchmarks so far it's about 20% faster than JS, but more importantly, runs in whatever thread you trigger it in, so can run in the high priority thread. It is also open source, so if you are comfortable recompiling, you can add a new scheme function that runs as compiled C pretty easily to allow one to script high performing functions from a high level language. The scenario you describe was one of the motivations for making it actually, so if you're interested, let me know and I can help you get going, it would be great to have someone testing the buffer functions in real world situations.
https://iainctduncan.github.io/scheme-for-max-docs/

Gen (without ~) for data manipulation