为什么此python代码挂在import / compile上，但可以在shell中运行？

Question

I'm trying to use python to sftp a file, and the code works great in the interactive shell -- even pasting it in all at once.

When I try to import the file (just to compile it), the code hangs with no exceptions or obvious errors.

How do I get the code to compile, or does someone have working code that accomplishes sftp by some other method?

This code hangs right at the ssh.connect() statement:

""" ProblemDemo.py
    Chopped down from the paramiko demo file.

    This code works in the shell but hangs when I try to import it!
"""
from time           import sleep
import os

import paramiko


sOutputFilename     = "redacted.htm"  #-- The payload file

hostname    = "redacted.com"
####-- WARNING!  Embedded passwords!  Remove ASAP.
sUsername   = "redacted"
sPassword   = "redacted"
sTargetDir  = "redacted"

#-- Get host key, if we know one.
hostkeytype = None
hostkey     = None
host_keys   = {}
try:
    host_keys = paramiko.util.load_host_keys(os.path.expanduser('~/.ssh/known_hosts'))
except IOError:
    try:
        # try ~/ssh/ too, because windows can't have a folder named ~/.ssh/
        host_keys = paramiko.util.load_host_keys(os.path.expanduser('~/ssh/known_hosts'))
    except IOError:
        print '*** Unable to open host keys file'
        host_keys = {}

if host_keys.has_key(hostname):
    hostkeytype = host_keys[hostname].keys()[0]
    hostkey     = host_keys[hostname][hostkeytype]
    print 'Using host key of type %s' % hostkeytype


ssh     = paramiko.Transport((hostname, 22))

ssh.connect(username=sUsername, password=sPassword, hostkey=hostkey)

sftp    = paramiko.SFTPClient.from_transport(ssh)

sftp.chdir (sTargetDir)

sftp.put (sOutputFilename, sOutputFilename)

ssh.close()

Answer 1

That's indeed a bad idea to execute this kind of code at import time, although I am not sure why it hangs - it may be that import mechanism does something strange which interacts badly with paramiko (thread related issues maybe ?). Anyway, the usual solution is to implement the functionality into a function:

def my_expensive_function(args):
    pass

if __name__ == '__main__':
    import sys
    my_expensive_functions(sys.args)

This way, just importing the module will not do anything, but running the script will execute the function with the given arguments at command line.

Answer 2

This may not be a direct reason why, but rarely do you ever want to have "functionality" executed upon import. Normally you should define a class or function that you then call like this:

import mymodule
mymodule.run()

The "global" code that you run in an import typically should be limited to imports, variable definitions, function and class definitions, and the like...

Answer 3

Weirdness aside, I was just using import to compile the code. Turning the script into a function seems like an unnecessary complication for this kind of application.

Searched for alternate means to compile and found:

import py_compile
py_compile.compile("ProblemDemo.py")

This generated a pyc file that works as intended. So the lesson learned is that import is not a robust way to compile python scripts.