How to Speed Up Python's urllib2 when doing mu

Question

I am making several http requests to a particular host using python's urllib2 library. Each time a request is made a new tcp and http connection is created which takes a noticeable amount of time. Is there any way to keep the tcp/http connection alive using urllib2?

Answer 1

If you switch to httplib, you will have finer control over the underlying connection.

For example:

import httplib

conn = httplib.HTTPConnection(url)

conn.request('GET', '/foo')
r1 = conn.getresponse()
r1.read()

conn.request('GET', '/bar')
r2 = conn.getresponse()
r2.read()

conn.close()

This would send 2 HTTP GETs on the same underlying TCP connection.

Answer 2

If you need something more automatic than plain httplib, this might help, though it's not threadsafe.

try:
    from http.client import HTTPConnection, HTTPSConnection
except ImportError:
    from httplib import HTTPConnection, HTTPSConnection
import select
connections = {}


def request(method, url, body=None, headers={}, **kwargs):
    scheme, _, host, path = url.split('/', 3)
    h = connections.get((scheme, host))
    if h and select.select([h.sock], [], [], 0)[0]:
        h.close()
        h = None
    if not h:
        Connection = HTTPConnection if scheme == 'http:' else HTTPSConnection
        h = connections[(scheme, host)] = Connection(host, **kwargs)
    h.request(method, '/' + path, body, headers)
    return h.getresponse()


def urlopen(url, data=None, *args, **kwargs):
    resp = request('POST' if data else 'GET', url, data, *args, **kwargs)
    assert resp.status < 400, (resp.status, resp.reason, resp.read())
    return resp

Answer 3

I've used the third-party urllib3 library to good effect in the past. It's designed to complement urllib2 by pooling connections for reuse.

Modified example from the wiki:

>>> from urllib3 import HTTPConnectionPool
>>> # Create a connection pool for a specific host
... http_pool = HTTPConnectionPool('www.google.com')
>>> # simple GET request, for example
... r = http_pool.urlopen('GET', '/')
>>> print r.status, len(r.data)
200 28050
>>> r = http_pool.urlopen('GET', '/search?q=hello+world')
>>> print r.status, len(r.data)
200 79124