On Tue, Apr 06, 2004 at 11:30:38AM +0200, Daniele Varrazzo wrote:
> Sorry, i don't know the internals of quixote compiling and traversal,
> but a linear scan on the _q_exports list sounds a bit strange to me.
> Wouldn't it be better to obtain a dict mapping sometimes (but once -
> don't know where)?

I'd be astonished if it made a difference; _q_exports lists are probably
short (<20 items), and on a poky little iBook 'str in list' takes only a
microsecond or two.  Most URLs are probably short, too, so the linear search
will likely cost less than a millisecond.  This time is swamped by network
uncertainties, the time for database queries and I/O, and the computations
done in the application code.

Zope also provides a lesson here; its publishing loop is even more
complicated than Quixote's, yet Zope still performs acceptably.  Computers
are really fast these days, and a few extra Python operations on every
request aren't visible to users.

--amk