New developer introduction

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Wouter van Heyst wrote:
| On Mon, Mar 10, 2008 at 09:08:14PM +0200, Alexander Belchenko wrote:
|
| ...
|
|> And I always can continue to build for myself standalone installer.
|> Because py2exe'd version is noticeable faster for me at startup :-P
|
| That is an interesting side effect I hadn't considered. I wonder if that
| is similar to Martin's experimental blat-everything-to-one-file
| speedups.
|
| Alexander: do you have measurements on how much faster it is?
|
| Wouter van Heyst
|
|

Alexander, any chance you could copy the "profile_imports.py" file over and give
me the results for the standalone version versus the py2exe one?

I'm guessing the big difference is just that sys.path has been dramatically
streamlined by py2exe. So it only has a few possible places to look for files.

I know at our last sprint Andrew Bennetts did some testing about python's odd
search for modules. IIRC he found that from a cold start, a plain 'bzr status'
was taking 7s. Copying the open and stat calls (as reported by strace) took 3s.
(So loading modules was taking 50% of a cold-stat test.) However, when he
streamlined it it still took 2.7s.

Anyway, I'm curious why the standalone is being so much faster, and I wonder if
we can't get some of that from the setuptools eggs. We have to figure out how we
want to handle plugins in that case, but you are already trying to address that
by giving a system-wide plugin install directory, outside of bzrlib/plugins.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFH1YpGJdeBCYSNAAMRAk6mAJoCti2yzrk1yQ+T/MTfFM6umTuwGgCfV0Ei
W0b//RyWjp/1CRLW4nUE450=
=pJKe
-----END PGP SIGNATURE-----