Python os和gob：在单个文件与多个文件上执行脚本时，结果不同

Question

I am writing a python shell script that has the option to execute a function (that writes a set of means from a 2D numpy array to a data file) on either a single .wav file or all .wav files in a specified directory.

For example,

> myscript.py --file=audio.wav --data="data.tab"

should write a floating point value to data.tab.

> myscript.py --path="path/with/audio_files" --data="data.tab"

should write a set of floating point values to data.tab.

I have a folder with a bunch of WAV files. When I execute the path on a single file (with the --file option), the result is different than the result when file gets processed with the --path option.

I have implemented the path file 3 ways:

1) Using os.listdir()

for audioFile in os.listdir(options.path):
    if audioFile.endswith('.wav'):
        foo(audioFile)

2) Using os.walk()

for r, d, f in os.walk(options.path):
for audioFile in f:
    if audioFile.endswith('.wav'):
            foo(audioFile)  

3) Using glob.glob()

for audioFile in glob.glob("*.wav"):
    foo(audioFile)

Methods 1 and 2 return the same result. Method 3 returns a different result. All 3 methods return a different result than processing a single file.

What is going differently when I use the os or gob modules?

EDIT: Here is where I process all .wav files in a directory:

for r, d, f in os.walk(options.path): 
    for audioFile in f:
        if audioFile.endswith('.wav'):  
            # Add MFCC 1-12 to data.
            mfcc12(audioFile, sampleRate, data)

This is in mfcc12():

# mfccs is a 2D numpy array. 
# Each column corresponds to one feature of the audiofile
for i in range(mfccs[0].size):
    mfccMean = mfccs[:, i].mean()
    mfccStdDev = mfccs[:, i].std()  
    data.write(str(mfccMean) + '\t' + str(mfccStdDev) + '\t')

I am using YAAFE to extract the features from the audio files.

Answer 1

The glob functions do the Unix-style thing of omitting files whose name starts with . :

>>> os.listdir('.')
['conn.c', 'Makefile', 'conn.o', '.depend', 'conn.c.orig', 'conn']
>>> glob.glob('*')
['conn.c', 'Makefile', 'conn.o', 'conn.c.orig', 'conn']

(note, fnmatch.fnmatch does not do this; the code to skip dot-files is in glob.glob1 ). Presumably you have some dot-files.

Answer 2

I think I solved the problem. The problem was not with my mfcc extraction settings, not the Python code.