Sample program inspired by the todo in kriskowal/xbin

samp.py

#!/usr/bin/env python
"""usage: %prog [-n N]

Returns a random uniform sample of the lines on stdin using a technique
called reservoir sampling [VITTER '85]. Preserves input ordering."""

import sys
from optparse import OptionParser
p = OptionParser(usage = __doc__)
p.add_option(
    '-n', default = 1, type = int,
    help = 'number of lines to sample (default 1)')
(options, args) = p.parse_args()
if options.n < 1:
    p.error('sample size must be positive')

import random
sample = []
for i, line in enumerate(sys.stdin):
    if i < options.n:
        sample.append((i, line))
    else:
        r = random.randrange(i+1)
        if r < options.n:
            sample[r] = (i, line)

sys.stdout.write(''.join(l for i,l in sorted(sample)))