Benchmarking urllib2 vs. urllib3

urllib3 seems to be a long-abandoned project on PyPI. However, it has some features (like re-using connections, aka HTTP Keep-Alive) that are not present in the Python 2 version of urllib and urllib2. Another package that provides HTTP Keep-Alive is httplib2.

Benchmark results on a single host

Keep-Alive can significantly speed up your scraper or API client if you’re connecting to a single host, or a small set of hosts. This example shows the times spent downloading random pages from a single host, using both urllib2 and urllib3:


urllib2 vs. urllib3 benchmark results urllib2 vs. urllib3 benchmark results

The benchmark script

Here’s a script that will benchmark urllib2 and urllib3 for the domain theoatmeal.con, and write out the results to a CSV files (easy to importy to Google Docs Spreadsheet and generate a nice chart).

If you run it, it will also prent the result summary, something like this:

Starting urllib2/urllib3 benchmark...
 * crawling:
 * crawling:
 * crawling:
 * crawling:
Finishing benchmark, writing results to file `results.cvs`
Total times:
 * urllib2: 183.593553543
 * urllib3: 95.9748189449

As you can see, urllib3 appears to be twice as fast as urllib2.