Python parallel http requests using multiprocessing

parhttp.py

#!/usr/bin/env python

from multiprocessing import Process, Pool
import time
import urllib2

def millis():
  return int(round(time.time() * 1000))

def http_get(url):
  start_time = millis()
  result = {"url": url, "data": urllib2.urlopen(url, timeout=5).read()[:100]}
  print url + " took " + str(millis() - start_time) + " ms"
  return result
  
urls = ['http://www.google.com/', 'https://foursquare.com/', 'http://www.yahoo.com/', 'http://www.bing.com/', "https://www.yelp.com/"]

pool = Pool(processes=5)

start_time = millis()
results = pool.map(http_get, urls)

print "\nTotal took " + str(millis() - start_time) + " ms\n"

for result in results:
  print result
  
  

why is this "tread safe"?

Because of the GIL. You don't have to worry about observability side effects in python (at the cost of performance, but then again you're using python)