Fwd: [squeak-dev] OB performance (Re: Slime)

---------- Forwarded message ----------
From: Andreas Raab <[hidden email]>
Date: Wed, Apr 7, 2010 at 5:25 AM
Subject: [squeak-dev] OB performance (Re: Slime)
To: The general-purpose Squeak developers list <[hidden email]>


On 4/3/2010 2:22 PM, Colin Putney wrote:

On 2010-04-03, at 11:50 AM, Andreas Raab wrote:

On 4/3/2010 11:27 AM, Levente Uzonyi wrote:

On Sat, 3 Apr 2010, Andreas Raab wrote:

Interesting. My encounters with OB mostly originate from Pharo where
it's always been dog-slow so I was never interested in looked deeper.
Any ideas how to install OB without that stuff?

Try this:

Thanks. It's definitely better than I remembered. I'm still noticing a real delay in some areas though, for example when I click on a system category. One thing you can try to test this is to open a regular browser and OB and then just keep scrolling through the system categories with cursor up or cursor down. In the regular browser this entirely smooth; in OB it's "stuttering" and at least I notice the same delay when clicking on the list.

Yeah, this one area where OB is certainly not as fast. That's because OB sorts classes according to inheritance, rather than name. It's just more work. I find the tradeoff worthwhile - it's not that slow, and the extra information imparted by the sort order is useful - but reasonable people may disagree. It wouldn't be hard to make this a preference.

Nope it's not that. A bit of profiling shows the particular problem is caused by some gross inefficiencies when updating. If you file in the attached tweaks, OB scrolls just as smoothly as the standard browser. I'm not sure if the tweaks have side effects though; you might want to verify that (in particular because of the comment in OBColumn>>switchFilter: which indicates some interference although I can't see how that could make a difference if the filter is identical).

Cheers,
 - Andreas