Optimized alternative to urllib.quote

[kgriffs]

See if we can do better than ~2 us (and ~4 us on Py33)

[queertypes]

Here's some benchmarks: http://paste.openstack.org/show/46739/

[queertypes]

Things are especially slow in python 3, but are reasonably snappy under pypy. The first thing that comes to mind here is calling out to cffi using a C module for quoting and unquoting URLs.

I found python-percentcoding on PyPI, but it doesn't work on Python 3 and it might not work on pypy, either.

[kgriffs]

Thanks for taking a look! Good data points.

[syllog1sm]

I remember looking at the urllib.quote function a long time ago, because I hit the unicode bug in it.

Check out the implementation: https://gist.github.com/syllog1sm/6656171

It's easy to see that this could be made more efficient, just by reimplementing it in a more straight-forward way.

Looking at the falcon source, it seems this is only used in utils.percent_escape, right? So the set of reserved characters is fixed. That makes things much simpler.

If the simple Python way isn't good enough, a Cython array would be very efficient.

[syllog1sm]

Oh, also, it seems you want urllib.parse.quote_from_bytes for Python 3?

http://docs.python.org/3.4/library/urllib.parse.html

[kgriffs]

Cool, thanks for the tips! I ended up referencing py2 and py3 implementations of urllib and then wrote a lean-and-mean url encoder for Falcon.