Hello,

I've just seen this amazing tutorial on Tables and Arrays.

But I was left with a doubt relative to tables / arrays.

In the example he sequences a small collection of notes, and mentions that logic could be used to build a sequencer in PD.

But in his example, he has an array with only one note at a time, as in a monophonic instrument.

What if one needs to build a sequencer for a polyphonic instrument? How to store two notes at a time, or chords ? How to have stored the velocity for each note?

Are there bi-dimensional arrays in PD... or multi-dimensional?

Or can we store objects, like instances of a Class, inside arrays? That way each object could be a "event" with multiple sub-events within, and the respective information for each.

Or are Lists used for this purpose ?

easiest way is just to use multiple tables

for monophonic sequences you could store the velocity as decimal places (add vel/256 to the note values for example), and 'decode' it with [wrap] (and * 256 (or any nr > 128)) when playing back the sequence.

if you have a fixed (or max) nr of voices, you could use a table of the size 'nr of voices' * 'nr of steps', multiply the counter for reading the table with 'nr of voices', and get the individual voices with a 'until-loop'.

tables might become huge for longer sequences and high time precision, so it might be a better idea to store this kind of sequences in lists or data structures, though you will not have random access to the individual notes with that.

for a small nr of complex events the method you suggested should be a good idea. store the 'addresses' of the events in a simple table, and have a table, message,whatever for all the details of each event.

By using tables to store sequence information I think you'll always reach a point where you have to limit the number of voices you can use. Either using multiple tables or by adding an offset per voice like mnb suggested.

But really it only makes sense to limit the number of voices in the synthesis engine, why shouldn't I be able to send a huge cluster of notes to a monophonic synth? It should be up to the synth to decide how many voices to use and which notes to choose.

Using a list approach would mean you could store a theoretically unlimited number of notes per steps. For example maybe having a list in the following format:

[stepIndex noteIndex noteValue noteVelocity (

Then a chord might look something like this:

[0 0 60 100 1 64 100 2 67 100 (

I might have missed something as thing was just off the top of my head but it should be fairly trivial to create a list based sequencer. The only downside is that lists are not really as efficient as tables. That being said, if you have a limited number of sequence notes per step I would use tables.

yes the writeable number of notes will be 82 I think, like a piano. Then in the synth I will limit the polyphony to 16. But technically it must be possible to add 82 simultaneous notes to the sequencer visually, even though the polyphony will be limited in the end.

It may help to clarify that my intent is to build a sequencer, like a piano-roll. A simple piano roll! I wonder why there isn't one out there yet. Then I will have another app using the patch and representing visually whatever contents of the sequencer the patch has stored, or is running.

Therefore it is a must to be able to the edit any notes stored at any place in time and to add notes at any place in time! And it should allow infinite length for the composition, like in piano rolls such as the ones of Ableton Live's MIDI clips.

Martin Brinkmann said:

"tables might become huge for longer sequences and high time precision, so it might be a better idea to store this kind of sequences in lists or data structures, though you will not have random access to the individual notes with that."

That sounds like I'm screwed! Tables lose precision and lists and data structures don't provide me access to individual notes?

Does that mean it's impossible to store the contents of a piano roll sequencer in PD?..

@antoniobrandao said:

That sounds like I'm screwed! Tables lose precision and lists and data structures don't provide me access to individual notes?

Does that mean it's impossible to store the contents of a piano roll sequencer in PD?..

no. tables seem to be not ideal for the things you want to do, though a table with a reasonanble time-resolution of, lets say 10 ms, 60 min maximum lenght, per voice, seems manageable compared to audio samples. but probably not the most elegant solution.

with datastructures, and lists, qlist and textfile you can of course access individual notes, but not randomly. if you want to access a note at a given position, you have to seek through (traverse) all the notes from the beginning (or at least the current position). this is not so much a pd, but a general problem. a possible solution would be some kind of hashing, or a hierarchical form of storing the data like a list of bars which contain a list of 1/16 ranges, which contain a list of notes, so that seeking a position becomes a lot quicker. i dont know if someone has already done this in pd. and maybe you do not need this, and a qlist or datastructure where you can move from one event to the next is sufficient.

it looks like johannes kreidler has built a decent piano-roll (with some extras and algorithmic composiotion features) editor in pd. though i have not seen the patch, only the video: