无法导入声音文件

Question

I'm using Anaconda and I'm trying to import soundfile/pysoundfile.

I installed the package by running

 conda install -c conda-forge pysoundfile 

and I think it succeeded because when I run

 conda list 

it shows up:

pyopenssl                 17.2.0           py36h5d7bf08_0  
pyparsing                 2.2.0            py36hb281f35_0  
pyqt                      5.6.0            py36he5c6137_6  
pysocks                   1.6.7            py36hfa33cec_1  
pysoundfile               0.10.1           py_0            conda-forge
pytables                  3.4.2            py36hfbd7ab0_2  
pytest                    3.2.1            py36h9963153_1  

To make sure I'm running the "right" python, I've tried running

which python

and I get

 /anaconda3/bin/python

But when I open python and try running

import soundfile 

I get the following:

Traceback (most recent call last):
  File "/anaconda3/lib/python3.6/site-packages/soundfile.py", line 142, in <module>
    raise OSError('sndfile library not found')
OSError: sndfile library not found

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/anaconda3/lib/python3.6/site-packages/soundfile.py", line 163, in <module>
    _path, '_soundfile_data', _libname))
OSError: cannot load library '/anaconda3/lib/python3.6/site-packages/_soundfile_data/libsndfile.dylib': dlopen(/anaconda3/lib/python3.6/site-packages/_soundfile_data/libsndfile.dylib, 2): image not found

Out of curiosity I opened the soundfile.py file with a text editor to see what line 142 looked like, and it looks like this:

try:
    _libname = _find_library('sndfile')
    if _libname is None:
        raise OSError('sndfile library not found')      # <---- line 142
    _snd = _ffi.dlopen(_libname)
except OSError:
    if _sys.platform == 'darwin':
        _libname = 'libsndfile.dylib'
    elif _sys.platform == 'win32':
        from platform import architecture as _architecture
        _libname = 'libsndfile' + _architecture()[0] + '.dll'
    else:
        raise

    # hack for packaging tools like cx_Freeze, which
    # compress all scripts into a zip file
    # which causes __file__ to be inside this zip file

    _path = _os.path.dirname(_os.path.abspath(__file__))

    while not _os.path.isdir(_path):
        _path = _os.path.abspath(_os.path.join(_path, '..'))

    _snd = _ffi.dlopen(_os.path.join(
        _path, '_soundfile_data', _libname))

In the docs, it says "import soundfile" is the correct way to import, but I also tried

import pysoundfile 

and only got

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'pysoundfile'

Is there anyone who knows why this is happening? I'm running this on MacOS btw.

Answer 1

Read the Breaking Changes section here

we changed the import name from import pysoundfile to import soundfile in 0.7.

workaround

pip install SoundFile

Answer 2

• I experienced this error as a result of using the librosa library, which uses SoundFile
  • The OSError seems to occur because SoundFile is an older version.
  • GitHub-librosa: Cannot load when importing librosas #1117
  • librosa seems to install, and should work with pysoundfile , but in my experience, it did not.
    • Remove pysoundfile : conda remove pysoundfile --force-remove
• SoundFile is not available for a conda install.
• In my case, I had already done pip uninstall SoundFile , which then lead to an import error that the package couldn't be found.
• Then I did pip install SoundFile , which installed a newer version and resolved the error.
• This worked on Windows 10 with Anaconda, python 3.8.11 , SoundFile 0.10.3.post1 , librosa 0.8.1