[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-i18n-doc
Subject:    Re: Fuzzy Clock
From:       Chusslove Illich <caslav.ilic () gmx ! net>
Date:       2004-09-28 17:25:30
Message-ID: 200409281925.32586.caslav.ilic () gmx ! net
[Download RAW message or body]

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

> [: Stephan Kulow :]
> You're basically asking for a slow down of all of (translated) KDE. How
> huge that slow down is depends a bit, but I fail to believe a scheme
> interpretation comes at no costs.

Two important things when considering performance are that we are talking 
about GUI display here and that cases where scripting is to be employed 
are rare. When message is not scripted, Scheme interpreter is not run at 
all, and all supporting structure overhead is lost among catalog searches 
and drawing of GUI. When message is scripted, I still couldn't feel any 
holdup in test cases.

I tested performance of scripting system on 1GHz/SDR machine, taking two 
cases. One was scripting case inflection in "Open with" items in e.g. 
Konqueror's context menu. I began to notice some holdup before the menu 
appears only after I had assured that more than thousand string 
comparisons were performed in script file. The other case was scripting of 
"Found %n files" to have fancier plural handling, in bottom of KFind 
window while searching for pattern "*". That is, I wondered if continuous 
fast update of this message will slow down listing of files -- it didn't.

Of special note in these tests is that both scripts also loaded their 
script files in each call, versus one single preloading that I am 
contemplating right now (the only problem are possible name clashes of 
translator defined functions if more than one language + English is 
displayed in GUI).

In the end, if user feels any performance problems, there is a config 
switch to disable script usage (in which case ordinary fallback 
translations are used, like when the script fails). Then, the only 
overhead are class instancing (rather simple one, no virtuals, just one 
bool and pointer data member), one string search and at most two 
additional string copyings.

- -- 
Chusslove Illich (Часлав Илић)
Serbian KDE translation team
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBWZ6MMSGXgigGr3ERAiuyAJ964n3HrBCcwSwLIJXYKomt/QnVnwCdHhE5
cDRZpS0WM6nWyUh3It/TWpI=
=ojGZ
-----END PGP SIGNATURE-----

[prev in list] [next in list] [prev in thread] [next in thread]