for the sake of completeness, other way to implement it could be to extract the ids from bndl like bndl.flop[2].

Line 587 (without this patch), could be:

[15 /* \n_set */, bndl.flop[2], \gate, 0].flop,

bndl array is a larger data structure however I don't know if flop is faster than reverse for this case or how much that is important in this context.

This patch follows implementation semantics used before in that function and ids variable isn't used afterwards (same as bndl). I just thought that to follow current implementation was better in any case.

And, I just saw a possible source of confusion:

Before, line 554:

~id = ids = Array.fill(bndl.size, { server.nextNodeID });

should ~id be reassigned? I don't think so because ids are just synth instantiated in reverse temporal order, ~freq wouldn't change either for example.

The thing is that ids was not a constant in that block, and why is used as nil in line 537 is something I don't want to ask.

I'm just sharing my reasoning about this.

bndl array is a larger data structure however I don't know if flop is faster than reverse for this case or how much that is important in this context.

reversing an array of a couple of numbers will certainly be faster than a flop of the whole structure. If in doubt, you can check, e.g.:

a = (0..32);
b = [2, a, 5];

bench { a.reverse };
bench { b.flop };

should ~id be reassigned?

No, as you say, the order of the variable ids is really only the order for the scheduling timing, not the order of addressing.

The thing is that ids was not a constant in that block, and why is used as nil in line 537 is something I don't want to ask.

I'll answer anyway. It in nil because we don't know how many synths (and therefore how many ids) we'll need before multichannel expansion in

Thanks for the review Julian. My first thought about the variable was it was not initialised, "that's wrong", then I thought it's there because clearly marks the place for the ids defined later as a name instead of raw nil. However, after that I decided for me it would have been better to use nil because the next operations in that array slot makes clear that there go ids. I was a little jocular.

ah, yes! Thanks for the discussion.

Should this be merged into 3.10 as well?