closing files properly opened with urllib2.urlopen()

I have following code in a python script

  try:
    # send the query request
    sf = urllib2.urlopen(search_query)
    search_soup = BeautifulSoup.BeautifulStoneSoup(sf.read())
    sf.close()
  except Exception, err:
    print("Couldn't get programme information.")
    print(str(err))
    return

I'm concerned because if I encounter an error on sf.read(), then sf.clsoe() is not called. I tried putting sf.close() in a finally block, but if there's an exception on urlopen() then there's no file to close and I encounter an exception in the finally block!

So then I tried

  try:
    with urllib2.urlopen(search_query) as sf:
      search_soup = BeautifulSoup.BeautifulStoneSoup(sf.read())
  except Exception, err:
    print("Couldn't get programme information.")
    print(str(err))
    return

but this raised a invalid syntax error on the with... line. How can I best handle this, I feel stupid!

As commenters have pointed out, I am using Pys60 which is python 2.5.4

I would use contextlib.closing (in combination with from __future__ import with_statement for old Python versions):

from contextlib import closing

with closing(urllib2.urlopen('http://blah')) as sf:
    search_soup = BeautifulSoup.BeautifulStoneSoup(sf.read())

Or, if you want to avoid the with statement:

try:
    sf = None
    sf = urllib2.urlopen('http://blah')
    search_soup = BeautifulSoup.BeautifulStoneSoup(sf.read())
finally:
    if sf:
        sf.close()

Not quite as elegant though.

finally:
    if sf: sf.close()

Why not just try closing sf, and passing if it doesn't exist?

import urllib2
try:
    search_query = 'http://blah'
    sf = urllib2.urlopen(search_query)
    search_soup = BeautifulSoup.BeautifulStoneSoup(sf.read())
except urllib2.URLError, err:
    print(err.reason)
finally:
    try:
        sf.close()
    except NameError: 
        pass

Given that you are trying to use 'with', you should be on Python 2.5, and then this applies too: http://docs.python.org/tutorial/errors.html#defining-clean-up-actions

If urlopen() has an exception, catch it and call the exception's close() function, like this:

try:
    req = urllib2.urlopen(url)
    req.close()
    print 'request {0} ok'.format(url)
except urllib2.HTTPError, e:
    e.close()
    print 'request {0} failed, http code: {1}'.format(url, e.code)
except urllib2.URLError, e:
    print 'request {0} error, error reason: {1}'.format(url, e.reason)

the exception is also a full response object, you can see this issue message: http://bugs.jython.org/issue1544

Looks like the problem runs deeper than I thought - this forum thread indicates urllib2 doesn't implement with until after python 2.6, and possibly not until 3.1

You could create your own generic URL opener:

from contextlib import contextmanager

@contextmanager
def urlopener(inURL):
    """Open a URL and yield the fileHandle then close the connection when leaving the 'with' clause."""
    fileHandle = urllib2.urlopen(inURL)
    try:     yield fileHandle
    finally: fileHandle.close()

Then you could then use your syntax from your original question:

with urlopener(theURL) as sf:
    search_soup = BeautifulSoup.BeautifulSoup(sf.read())

This solution gives you a clean separation of concerns. You get a clean generic urlopener syntax that handles the complexities of properly closing the resource regardless of errors that occur underneath your with clause.

Why not just use multiple try/except blocks?

try:
    # send the query request
    sf = urllib2.urlopen(search_query)
except urllib2.URLError as url_error:
    sys.stderr.write("Error requesting url: %s\n" % (search_query,))
    raise

try:
    search_soup = BeautifulSoup.BeautifulStoneSoup(sf.read())
except Exception, err: # Maybe catch more specific Exceptions here
    sys.stderr.write("Couldn't get programme information from url: %s\n" % (search_query,))
    raise # or return as in your original code
finally:
    sf.close()