| 1 |
Author : TGL |
| 2 |
Purpose : |
| 3 |
While profiling a script that reads CONTENTS files using |
| 4 |
'dblink.getcontents()', i've seen that it was spending quite some time |
| 5 |
in ~400k calls to 'os.path.normpath()' (one per referenced element). I |
| 6 |
think there is nothing to normalize here (paths in CONTENTS files are |
| 7 |
already normal by construction) and thus this calls could be avoided. |
| 8 |
|
| 9 |
The one-line patch i will attach does this. It's effect is easy to |
| 10 |
check, for instance with equery: |
| 11 |
|
| 12 |
Before: |
| 13 |
% time equery belongs /bin/bash > /dev/null |
| 14 |
real 0m13.583s |
| 15 |
user 0m13.215s |
| 16 |
sys 0m0.249s |
| 17 |
|
| 18 |
After: |
| 19 |
% time equery belongs /bin/bash > /dev/null |
| 20 |
real 0m6.526s |
| 21 |
user 0m6.218s |
| 22 |
sys 0m0.246s |
| 23 |
|
| 24 |
A quick testing of normpath in python shows it acts how I think it does, |
| 25 |
which means i think this is patch is good to put in, it's short simple, |
| 26 |
and shouldn't break anything. |
| 27 |
|
| 28 |
I kind of see Brian's point in being defensive in filenames though |
| 29 |
(os.path.normpath("///////var/db/pkg/fex") being invalid in a CONTENTS |
| 30 |
file but still usable by most(?) calls in python, which will just figure |
| 31 |
it all out anyway. |
| 32 |
|
| 33 |
If portage errored on invalid paths in CONTENTS file, I would see |
| 34 |
leaving it there, but if it doesn't help anything ( because it doesn't |
| 35 |
hurt anything either, right? :) ) then why do the work? |
| 36 |
|
| 37 |
*has ranted too long :P* |
| 38 |
|
| 39 |
-Alec Warner (antarus) |
| 40 |
-- |
| 41 |
gentoo-portage-dev@g.o mailing list |
Archives