Suppressing output of module calling outside libra

问题:

I have an annoying problem when using machine learning library PyML. PyML uses libsvm to train the SVM classifier. The problem is that libsvm outputs some text to standard output. But because that is outside of Python I cannot intercept it. I tried using methods described in problem Silence the stdout of a function in Python without trashing sys.stdout and restoring each function call but none of those help.

Is there any way how to do this. Modifying PyML is not an option.

回答1:

Open /dev/null for writing, use os.dup() to copy stdout, and use os.dup2() to copy your open /dev/null to stdout. Use os.dup2() to copy your copied stdout back to the real stdout after.

devnull = open('/dev/null', 'w')
oldstdout_fno = os.dup(sys.stdout.fileno())
os.dup2(devnull.fileno(), 1)
makesomenoise()
os.dup2(oldstdout_fno, 1)

回答2:

Dave Smith gave a wonderful answer to that on his blog. Basically, it wraps Ignacio's answer nicely:

def suppress_stdout():
    with open(os.devnull, "w") as devnull:
        old_stdout = sys.stdout
        sys.stdout = devnull
        try:  
            yield
        finally:
            sys.stdout = old_stdout

Now, you can surround any function that garbles unwanted noise into stdout like this:

print "You can see this"
with suppress_stdout():
    print "You cannot see this"
print "And you can see this again"

For Python 3 you can use:

from contextlib import contextmanager
import os
import sys

@contextmanager
def suppress_stdout():
    with open(os.devnull, "w") as devnull:
        old_stdout = sys.stdout
        sys.stdout = devnull
        try:  
            yield
        finally:
            sys.stdout = old_stdout

回答3:

I had the same problem and fixed it like that:

from cStringIO import StringIO

def wrapped_svm_predict(*args):
    """Run :func:`svm_predict` with no *stdout* output."""
    so, sys.stdout = sys.stdout, StringIO()
    ret = svm_predict(*args)
    sys.stdout = so
    return ret