Why does writing STDOUT affect my Python LDAP performance?

[Leif H. <le...@pr...>]

Hi,

I wanted to write a very small app to do some simple tests, and I got 
some very, very strange results. I can't understand why the performance 
is so abysmal unless I print on STDOUT... The script is something like:

#!/usr/bin/python2

import ldap

l = ldap.open("localhost")
l.simple_bind_s("", "")

loop = 1000
while loop:
    loop = loop - 1
    print "Going into LDAP, loop is ", loop
    res = l.search_s(
        "dc=propel,dc=com",
        ldap.SCOPE_SUBTREE,
        "cn=Manager"
        )
    print "Done with search"

l.unbind()


(I added the print statements to show my problem). Here are the time 
resusts from two test runs:

269% time test.py
.
.
.
Done with search
2.180u 0.480s 0:06.85 38.8%     0+0k 0+0io 656pf+0w

loki 270% time /tmp/simple.py > /dev/null
115.770u 32.230s 2:31.46 97.7%  0+0k 0+0io 651pf+0w

Without redirecting STDOUT (lots of spam on the display), it takes about 
6 secs, when I do redirect STDOUT to devnull it takes 150 seconds (25x 
slower). This makes no sense to me at all... Also, this simple script 
leaks memory, not sure where yet.

Any ideas where to look? I'm using Python 2.1.2, OpenLDAP 2.0.22 and 
latest Python LDAP source from the CVS server.

Thanks,

-- leif