Routing the output of ctcsound to /dev/null

General
#1

Hi,

I’m using the ctcsound API to control the rendering process with performKsmps(). I’m analysing every k and using the analysis to train a Pytorch model.

I would like to send the output of the ctcsound rendering to /dev/null. I basically want to achieve what –logfile=null does, but as it says in the docs, this option only works when calling Csound from the command line. Is there a way to achieve this through the ctcsound API?

I’ve tried to set sys.stdout and sys.stderr to /dev/null within a context, as suggested here. That works for pure Python code, but not for the parts that make calls to Csound.

Routing all the Python logs to /dev/null is not an option since I want to be able to log other parts of my program.

Any ideas?

#2

Hi @ulrikah I’m not overly familiar with the Phyton bindings, but can’t you use csoundSetOption() to specify what command line flags you wish to put in place?

#3

Unfortunately, that doesn’t seem to work. I’ve tried with some variations of the same option, and all seem to give me a CSOUND_SUCCESS. However, even illegal options seem to give the same success code.

The result is that Csound renders, but with all the standard logging.

Here is the code for ref:

import sys
import ctcsound

csd_text = '''
  <CsoundSynthesizer>
  <CsOptions>
    -odac
  </CsOptions>
  <CsInstruments>
  instr 1
    out(linen(oscili(p4,p5),0.1,p3,0.1))
  endin
  </CsInstruments>
  <CsScore>
  i1 0 5 1000 440
  </CsScore>
  </CsoundSynthesizer>'''

cs = ctcsound.Csound()
options = [
    "--logfile=null",
    "-O null",
    "-Onull",
    "illegal_option"
]
for option in options:
    if cs.setOption(option) != 0:
        raise Exception("Illegal option", option)
result = cs.compileCsdText(csd_text)
result = cs.start()
while True:
    result = cs.performKsmps()
    if result != 0:
        break
result = cs.cleanup()
cs.reset()
del cs
sys.exit(result)
#4

Can you create your own message callback and then simply ignore everything that gets piped to it from Csound?

I prefer to use the message buffer, where you can pull and pop messages from it at your leisure.

Again, I’ve no idea how these are implemented in ctcsound…

#5

That seems to have removed most of the output. Thanks!

However, there is still some meta information that gets printed when I use messages:

0dBFS level = 32768.0
--Csound version 6.15 (double samples) Nov 20 2020
[commit: 18c2c7897425f462b9a7743cee157cb410c88198]
libsndfile-1.0.28
0dBFS level = 32768.0
--Csound version 6.15 (double samples) Nov 20 2020
[commit: 18c2c7897425f462b9a7743cee157cb410c88198]
libsndfile-1.0.28
end of score.		   overall amps:      0.0
	   overall samples out of range:        0
0 errors in performance
Elapsed time at end of performance: real: 0.000s, CPU: 0.000s
#6

Hmm, I’m not sure how to get rid of that. I think we might need to ping @stevenyi!

#7

I’m not too familiar with this part of the codebase, but I seem to remember similar issues discussed on the mailing list some time back. If there’s some example code that uses setMessageCallback() I can do a debug session to investigate.

#8

Actually, it seems like setMessageCallback() is deprecated in the latest version of ctcsound.py.

Right now, instead of creating a callback, I’m simply deleting the content of the buffer when I’m done rendering.

#9

just to let you know that i asked françois pinot (the author of
ctcsound) about this because he is not on this forum. his reply was:

I’m afraid it’s not a ctcsound problem. You can try this in a python
console:

import ctcsound
cs = ctcsound.Csound()
0dBFS level = 32768.0
–Csound version 6.15 (double samples) Oct 29 2020
[commit: 89d14ffa12fcac632181b309dc5e4f3cf7b73085]
libsndfile-1.0.28

del cs
end of score. overall amps: 0.0
overall samples out of range: 0
0 errors in performance
Elapsed time at end of performance: real: 3.063s, CPU: 0.001s

We see that python systematically echoes what’s coming back from
ctcsound. For me, it’s a python problem…

best -
joachim

1 Like
#10

Yes there’s a bit of a chicken and egg issue here with Csound printing right at instantiation. I had to look again and in the C API there is this function:

PUBLIC void csoundSetDefaultMessageCallback(
           void (*csoundMessageCallback)(CSOUND *csound,
                                         int attr,
                                         const char *format,
                                         va_list args))
{
    if (csoundMessageCallback) {
      msgcallback_ = csoundMessageCallback;
    } else {
      msgcallback_ = csoundDefaultMessageCallback;
    }
}

That sets the default message callback. This is also used by the logging system. It is part of the C API but is not available in ctcsound.py. I wonder if this could be added to ctcsound to accomodate messages that happen during csound creation and deletion.

1 Like
#11

ctypes does not provide a clear way of dealing with varargs functions and I have actually no much time to study the problem. But I did add setDefaultMessageCallback to ctcsound.py; it’s not fully implemented but it will allow you to define a noMessage callback like this:

import ctcsound
import sys

def noMessage(csound, attr, flags, *args):
    pass

csd_text = '''
  <CsoundSynthesizer>
  <CsOptions>
    -odac
  </CsOptions>
  <CsInstruments>
  instr 1
    out(linen(oscili(p4,p5),0.1,p3,0.1))
  endin
  </CsInstruments>
  <CsScore>
  i1 0 5 1000 440
  </CsScore>
  </CsoundSynthesizer>'''

ctcsound.setDefaultMessageCallback(noMessage)
cs = ctcsound.Csound()

result = cs.compileCsdText(csd_text)
result = cs.start()
while True:
    result = cs.performKsmps()
    if result != 0:
        break
result = cs.cleanup()
cs.reset()
del cs
sys.exit(result)

I hope this will help you and I’ll come back to this when I have more time…

You can download the new ctcsound.py file directly from github (https://github.com/csound/csound/blob/develop/interfaces/ctcsound.py

Regards

François

1 Like