UGen.buildSynthDef is nil in {}.plot (this makes LocalBuf and {}.plot incompatible)

[jleben]

[Issue migrated from SourceForge | ID: 2612612 | Submitted by 'danstowell']
[http://sourceforge.net/support/tracker.php?aid=2612612]

{IFFT(FFT(LocalBuf(1024), PinkNoise.ar))}.play // works

{IFFT(FFT(LocalBuf(1024), PinkNoise.ar))}.plot(1) // doesn't - buildSynthDef is nil

[Sciss]

apologies for polluting multiple issues with automatic mentions from unrelated commits; I accidentally pushed non-overlapping ScalaCollider branches to my fork of SuperCollider (a wrongly set-url remote). I deleted those branches again, but seems the mentions are still in the ticket system, now pointing to orphaned commits.

[jamshark70]

Reopening this issue: it is still a problem (not "completed" in 2012); neither was it was fixed by the commit merged in August 2020 (which commit is totally irrelevant to this issue).

Plot appears to be evaluating the function twice -- I put in a debugging statement in LocalBuf *new1 and got:

(
a = {
	var sig = SinOsc.ar(440);
	var fft = FFT(LocalBuf(2048), sig);
	IFFT(fft);
}.plot;
)

buildsynthdef: SynthDef:-1699245723  -- a-ha! It *is* running fine at first
-> a Plotter
buildsynthdef: nil  -- and then it isn't

So something spurious seems to be going on in the plot process.

[telephon]

yes, it seems to be called twice.

(
a = {
	UGen.buildSynthDef.postln; // nil
	0.0
}.plot;
)

a SynthDef
-> a Plotter
nil

// plot calls asBuffer
// which works.
(
a = {
	UGen.buildSynthDef.postln; // a SynthDef
	0.0
}.asBuffer;
)

[telephon]

So the problem is that the function is called a second time in

supercollider/SCClassLibrary/Common/GUI/PlusGUI/Math/PlotView.sc

Line 1319 in 50d9927

frameDur = if(this.value.rate == \control) {

Will need some care how best to inform the call about the rate of the UGen in the function.

As a first thought, it could be either:

• more information passed through an extra argument (a dict whose values can be set in asBuffer).
• just convert the output into audio rate if it isn't.

Any opinions?

[dyfer]

Thanks @telephon
I'm not sure which approach is better - the proposal in #6007 looks a bit like a workaround to me (this is not a criticism of the implementation, but rather of the difficulty in solving this gracefully).

At the risk of pushing the discussion in a wrong direction: on some occasions I felt that a Buffer (both the language object and the server implementation) should have a settable sample rate parameter. At some point I was recording quite a bit of data at control rate and the inability to set the sample rate properly always resulted in some inelegant assumptions/workarounds (e.g. when loading and playing back the data).

If we could set Buffer's sample rate, then we could extract that data from the passed buffer object in the action = {} of Function -plot and do frameDur = buf.sampleRate.reciprocal;

[telephon]

At the risk of pushing the discussion in a wrong direction: on some occasions I felt that a Buffer (both the language object and the server implementation) should have a settable sample rate parameter.

Yes, that was actually my thought as well when I read your comment. It is definitely a useful bit of information.

Going further down that road:

If a buffer has an optional sample rate, then this value should probably be accessible from the server side as well? Somehow DiskIn has such an internal info (should be somewhere here:

supercollider/server/plugins/DiskIO_UGens.cpp

Line 280 in 5c8e944

SndBuf* bufr = World_GetNRTBuf(unit->mWorld, (int)fbufnum);

).

[JordanHendersonMusic]

I can't reproduce this.

Using the initial example, play makes a noise, and plot draws noise.

{IFFT(FFT(LocalBuf(1024), PinkNoise.ar))}.play 
{IFFT(FFT(LocalBuf(1024), PinkNoise.ar))}.plot(1) 

(
a = {
	UGen.buildSynthDef.debug('building synthdef'); // nil
	0.0
}.plot;
)

prints

building synthdef: a SynthDef
-> a Plotter

@jamshark70 since you reopened this, would you mind testing again?