why is python slow?

[Donn Cave]

Quoth Skip Montanaro <skip at pobox.com>:
|
|     Donn> ... I wonder if all of Python's performance problems are really so
|     Donn> deep.  A few weeks ago someone posted some performance
|     Donn> measurements and, among other things, noted that version 2.2 took
|     Donn> twice as long just to start up.  Did that appalling statistic
|     Donn> attract any investigation into the cause, ...
|
| I believe the bulk of the effort in the 2.x series up to this point has
| focused on functionality ahead of performance.  I'm sure there are things
| that can be done.  Why not file a bug report, investigate and solve the
| problem, then grab some of that fame and fortune I alluded in my post
| seeking new developers yesterday?  (I'm not picking on Donn, he just happens
| to be the one who raised this issue and gave me the nice segue.)

I haven't even downloaded 2.2, because that's how I see it too -
that Python development has lately focused on functionality, given
a definition of functionality that doesn't include performance.
In other words, 2.2 is a toy release for people who want to play
with language features.  There's another segue in here back to
another of your posts from yesterday, but I'll leave that alone.

| The group of people who actively beat on the code is still fairly small.  I
| count 38 people currently with checkin privileges.  Of those, I count six
| whose SF usernames I don't even recognize.  There are probably another five
| to ten who don't regularly contribute to either checkins or python-dev
| discussions.  That means the effective pool of developers is really quite
| small.  That, in turn, means we need more eyeballs.  You don't need checkin
| privileges to tackle an open bug or investigate a problem and report its
| cause (a proposed solution would be helpful, but it might go a long ways
| towards fixing a problem if you can narrow down the cause a bit).
|
|     Donn> I believe I recall something similar in 2.0, due to a mistake in
|     Donn> site.py that was discovered during the beta releases.  Not a hard
|     Donn> problem, just one that wouldn't be noticed until someone wondered
|     Donn> why performance went downhill so fast.  We can wave our hands and
|     Donn> say performance is a tough problem, but it would be more
|     Donn> convincing if it looked like it was getting the most minimal
|     Donn> attention.
|
| Agreed.  When can we expect that patch? ;-)

Don't hold your breath.  I might be up for fixing problems in 2.1.1,
but it looks to me like 2.1 doesn't have this particular problem
(anyway, it starts up about as fast as 2.0 for me.)

	Donn Cave, donn at u.washington.edu