Re: [E-devel] Elementary table regression

[Carsten H. (T. R. <ra...@ra...>]

On Fri, 9 Oct 2009 16:59:37 -0300 Iván Briano (Sachiel) <sac...@gm...>
said:

> On Fri, Oct 9, 2009 at 4:46 PM, Nicolas Aguirre
> <agu...@gm...> wrote:
> > 2009/10/9 Iván Briano (Sachiel) <sac...@gm...>:
> >>
> >> It's not a problem with table or list, but evas_box and evas_table inherit
> >> from evas_object_smart_clipped, which uses a smaller clipper than the
> >> smart objects in elm were. Increasing the size of the clipper "fixes" it,
> >> but you will run into the same problem if you keep growing the table.
> >>
> >> I just commited a bigger clipper, managed to get 1000 columns in the
> >> table just fine, albeit very slow to scroll, and that's already an insanely
> >> big number for elm_table.
> >>
> >
> > It's usable with SVN < 42781, you can scroll 3x2000 items without lag
> > in test_photo. but it's very very very slow with recent SVN. But as I
> > said in my previous mail, table creation is much faster with
> > evas_table.
> >
> > Test with 3x2000 items :
> > SVN < 42781 : Table creation take 38.868 seconds, scroll is smotth.
> > SVN r42984 : Table creation take 'only' 1.329 seconds. you can't
> > scroll at least on my core2 duo.
> >
> 
> I can't test that now, but I *could* scroll the table with a huge ammount
> of objects, it was just noticeably slower.
> I'll look into this with more detail during the weekend.
> 
> > I agree that it's not very smart to add so many items in a table and
> > show all of them in the same time. But IIR you can't use elm_genlist
> > in horizontal mode. And I use a lot of elm_box in enna. And with these
> > last commits, it's very slow :/
> > But I find it unfortunate to have changed something that was really
> > fast and reliable to evas_box that seems much slower, maybe we should
> > try to fix evas_box before switching ?
> >
> 
> Both evas_box and evas_table were there for some time already and no one
> touched them, it's much more likely that they'll get fixed now that elm moved
> to them.

actually the move to evas_box/table broke a slew of things. index, scrolling in
photocam and scrolling in several other tests. one reason is that the smart
table/box would calc their min size on every add - thus its alweasy available
as "current". evas table/box are faster on inserts because they dont do this..
but thsi emans the min size is not known until some time later (or u explicitly
call calc) i suspect this has triggered a small mountain of issues. i added
something to foce a canvas-wide calc BEFORE any window show because one effect
was windows coming up "1x1" pixel THEN resizing to be bigger as min size wasn't
known until after first render. i added to edje the ability to listen to size
hint changes and auto-adjust the swallowed element. this reduced some swallows
in the code as its not needed anymore, but i suspect this has given rise to
other problems. i suggest you try every test and look at them in detail. but
something bad happened (tm).

-- 
------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler)    ra...@ra...