last (wtmp) parsing script

[pacofvf]

Hello, I made this script, It counts the ocurrences of the logged users, and gives you how much time a user has been logged in. The output of "last" command is somthing like this:



user1 :13 Tue Mar 1 08:17:08 2011 - down (06:21)
user1 pts/5 :20.0 Tue Mar 1 08:07:09 2011 - Tue Mar 1 08:50:39 2011 (00:43)
user2 :20 Tue Mar 1 08:04:50 2011 - Tue Mar 1 08:56:01 2011 (00:51)
user1 :20 Tue Mar 1 08:04:49 2011 - down (06:34)


and the output of my script is something like this:


Username Session Time(dd/hh/MM)
user1 0 : 15 : 8
user2 0 : 14 : 55


and here is my script:


#! /usr/bin/env python
import re
import subprocess
import operator

p = subprocess.Popen('last', shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
entries = re.split('\n+', p.stdout.read())
entries = [re.split('\\(|\\)| ', entry) for entry in entries][0:-2]
li = []
for entry in entries: #we keep only the username and the session length
if len(entry) > 6:
if len(entry[-5]) > 0:
li.append([entry[0] , entry[-5]])
else:
li.append([entry[0] , entry[-6]])
d = dict() # starts an empty dictionary
for entry in li:#sum all ocurrencies
t = 0
tS = re.split('\+',entry[1])#
u = entry[0]
if len(tS) > 1:
t = t + int(tS[0]) * 60 * 24
tS2 = tS[1]
else:
tS2 = tS[0]
tS = re.split('\:',tS2)
if len(tS) > 1:
t = t + int(tS[1]) + int(tS[0])*60
if u in d:
d[u]= t+d[u]
else:
d[u] = t
sorted_x = sorted(d.iteritems(), key=operator.itemgetter(1), reverse=True)
print 'Username\t\tSession Time(dd/hh/MM)'
for i in sorted_x:
h=int(i[1]/60)
m=int(i[1]%60)
d=int(h/24)
h=int(h%24)
print i[0],'\t\t',d,':',h,':',m


I'm new to python, I think that this could be coded in a more simple and elegant way. I dare you
:O

[LemursDontExist]

This might be a little tidier:


#! /usr/bin/env python

import os
from datetime import timedelta
from collections import defaultdict

with os.popen("last") as f:
data = f.read()

d = defaultdict(timedelta)
for entry in data.split('\n')[:-2]:
username = entry.split(' ')[0]
time_str = entry.split('(')[1].split(')')[0]
if '+' in time_str:
days, time_str = time_str.split('+')
days = int(days)
else:
days = 0
hours, minutes = map(int, time_str.split(":"))
d[username] += timedelta(days=days, hours=hours, minutes=minutes)

max_len = max(map(len, d.keys()) + [8]) + 2

print "%%-%ssSession Time(dd/hh/MM)" % max_len % "Username"
for username in sorted(d.keys()):
t = d[username]
hours, minutes = divmod(t.seconds/60, 60)
print "%%-%ss%%s:%%s:%%s" % max_len % (username, t.days, hours, minutes)

I like to avoid regular expressions where simple string operations will do. The datetime module saves me the trouble of doing a bunch of arithmetic.

I actually liked your use of the operator module and the 'key' argument to sorted - it makes the code feel very functional (That code would come out very naturally in, say, Haskell), but I felt that sorting the keys alone is more pythonic.

At the bottom, the funky string formatting is just to make sure the columns line up - in your original code the columns didn't line up right always. Though, it's worth noting that Python 3 totally eliminates this sort of string formatting, so I wouldn't pay it too much heed - I just haven't gotten around to learning whatever the replacement is.

[pacofvf]

Thanks, I used regex because you can compile them and use it repeatedly, saving a lot of processing capacity and time, but split() doesn't work for compiled objects :S, The datetime module simplifies the time formating and now I see that I should use a dictionary since the beginning, and avoid the part where I use a list and then a dictionary. ;)

[LemursDontExist]

I used regex because you can compile them and use it repeatedly, saving a lot of processing capacity and time

Actually, regexes, even when compiled, involve a lot more overhead than simple string operations. For complex searches, the overhead pays for itself, but for simple operations, the regex will be much slower.

Try the following:


import timeit
t1 = timeit.Timer("regex.split(a)", "import re; a='This is a splitting test'; regex=re.compile(' ')")
t2 = timeit.Timer("a.split(' ')", "a='This is a splitting test'")
t1.repeat()
t2.repeat()

The timeit.repeat() function runs the setup code once, and the main code 1000000 times (and repeats this process 3 times). On my machine the string version takes just a little over half as long as the regex version.

In any case, I feel that string operations tend to be a lot more readable and maintainable which is a lot more important than speed 95% of the time (knowing when you're in the 5% where it matters is of course a fun trick ;)).