Fabian Braennstroem wrote:
Hi,
it's me again :-)
As I mentioned before, I would like to clone lfm to urwid
(at least I try) mainly because of its poor 'speed' on large
directories. Rebecca's 'filechooser' seems a lot faster
handling a lot of files... but testing more and more those
existing urwid examples, I get the feeling that the time I
gain in handling a lot of files I would lose in the
navigation. What I mean is, that going from one line to
another in a listbox takes some time, the navigation in
menus or in radiobutton groups seem pretty slow.
Is it just my feeling or could that be a drawback of urwid?
I haven't spent very much time optimizing Urwid yet. That means there
is still a lot of "low-hanging fruit". A change that would have a very
large impact is to make rendering the screen only render those widgets
that have changed since the last time the screen was rendered. Right
now every visible widget is rendered on every screen update.
There are also some big improvements that could be made by rewriting the
performance-critical sections of code in C.
The patches I have against 0.9.7.1 are all performance related. I think
I might focus on performance for the next release.
For the moment you could try applying the latest patch I posted in the
"Unbearable slowness in UTF-8" thread:
http://article.gmane.org/gmane.comp.lib.urwid/350
Ian
_______________________________________________
Urwid mailing list
[email protected]
http://lists.excess.org/mailman/listinfo/urwid
