lsof utility wrapper copied from http://www.remix.net/blog/?p=130

lsof.py

#!/usr/bin/python
 
from subprocess import Popen, PIPE
import re
 
class lsof():
	"""
	lsof wrapper class
 
	An automated wrapper around lsof output that yields
	an array of processes with their file handle information
	in the form of:
 
	results = [
        {
            info: {
                'none': '',
		'user_id': '501',
		'process_id': '22665',
		'process_group_id': '20',
		'parent_pid': '232',
		'login_name': 'nar',
                'command_name': 'mdworker'
	    },
        files: [ 
			...  
		]    
	},
 
	...
 
	"""
 
    LSOF_BIN='/usr/sbin/lsof'
    flagRegex = re.compile("\s+(?P<flag>[a-zA-Z])\s+(?P<desc>.*)$")
    flag_cache = {'': 'none'}
 
    def __init__(self):
	"""
	For initialization, we query lsof using the '-Fh?' flag
	to automatically generate a list of possible key names
	so that later on when we're parsing records we can substitute
	them in the final dict. 
 
	When using the -F option, the key names are only provided as
	single characters, which aren't terribly useful if you don't
	know them by heart.
	"""
 
        cmd = "%s -F?" % (self.LSOF_BIN)
        p = Popen( [ self.LSOF_BIN, "-F?" ] , stdout=PIPE, stderr=PIPE, close_fds=True )
        for line in p.stderr.readlines()[1:]:
            m = self.flagRegex.match(line)
            if m is not None:
                flag = m.group('flag')
                desc = m.group('desc').lower()
                for s in (':', ' as', ' ('):
                    if desc.find(s) > -1:
                        desc = desc[0:desc.find(s)]
                        break
                self.flag_cache[m.group('flag')] = re.sub('[/ ,\t_]+', '_', desc)
 
    def run(self):
	"""
	run lsof and parse the results into a tmp array that gets passed back to the
	caller. 
	"""
 
        tmp = []
        cmd = "%s -PnF0" % (self.LSOF_BIN)
        p = Popen(cmd.split(' '), stdout=PIPE, close_fds=True)
        for line in p.stdout.xreadlines():
            r = dict( [ ( self.flag_cache[x[0:1]], x[1:] ) for x in line.rstrip('\n').split('\0') ] )
            if r.has_key('command_name'):
                tmp.append( { 'info': r, 'files': [] } )
            else:
                tmp[-1]['files'].append(r)
        return tmp
 
if __name__ == '__main__':
    import pprint
    l = lsof()
    for r in l.run():
        info = r['info']
        files = r['files'] 
 
        print "PID(%s) %s" % (info['process_id'], info['command_name'])
        for f in files:
            for k in f.keys():
                print "\t%-20s -> %s" % ( k, f[k] )

Now fixed, tested and working on linux, see the related fork (and the diff)

Changes :

• fixed indentation
• PEP8 compliance
• python3 compatibility using __future__ import
• line 06 : using CamelCase for class name, and new style class
• line 34 : fixed lsof path, using dynamic path resolution from distutils.spawn.find_executable
• line 35 : fixed broken regexp
• line 50 : removed useless cmd variable
• line 61 : replaced flag_cache dict key by previously unused variable flag
• line 63 : renamed r to r2 in run() for clarity
• line 70 : removed useless parenthesis
• line 74 : replaced has_key() by in operator
• line 81 : removed useless pprint import
• line 89 : removed useless .keys() for f

Thanks to the original author Nathan Ramella, see the original post from Nov. 2012 at Internet Archive (original page is defunct)