wc.py

import sys

si = sys.stdin

CHUNK_SIZE = 100000
n = 0

while True:
    b = si.read(CHUNK_SIZE)
    if not b:
        break
    n += b.count('\n')
    if len(b) < CHUNK_SIZE:
        break

print n

This is about as fast as the C version of wc -l:

$ time yes | head -10000000 | python wc.py
10000000
yes  0.76s user 0.01s system 58% cpu 1.319 total
head -10000000  1.29s user 0.01s system 98% cpu 1.318 total
python wc.py  0.04s user 0.02s system 4% cpu 1.319 total

$ time yes | head -10000000 | wc -l
 10000000
yes  0.76s user 0.01s system 59% cpu 1.294 total
head -10000000  1.28s user 0.01s system 99% cpu 1.293 total
wc -l  0.02s user 0.01s system 2% cpu 1.292 total

very cool. let's speed out the ol' streaming version then! feel free to file issues there