Re: [orca-list] heads up/please test

[Joanmarie Diggs <joanmarie diggs gmail com>]

Hi Ash.

On Tue, 2010-04-27 at 07:48 +0100, ashley wrote:
> Just out of curiosity, will this make orca's performance faster?

Not in and off itself. In other words, the changes that were committed
to master and referred to in the message you quote have no performance
impact whatsoever. (I did some performance testing to be sure there was
no negative impact.)

> From my message:

> > Today I was looking at some performance-related issues in Orca. One of
> > them is that we regularly spend 0.01 to 0.25 seconds in updateBraille().
> > We do this automatically, whether the user has braille and/or braille
> > monitor support enabled or disabled.

That means that we're spending quite a bit of time doing braille-related
things. If you do not have a braille display (many users don't), your
performance is suffering because Orca is doing braille-related things
that it shouldn't be doing. We need to stop Orca from doing that. But
there was a snag:

> > Unfortunately, it's not as simple as adding a check in updateBraille()
> > which returns immediately if braille support is disabled. Doing so winds
> > up breaking things elsewhere. We need to add checks pretty much anywhere
> > there is a braille-related call. And that, it turns out, is all over the
> > place throughout Orca

Breakage is bad. Having a bazillion direct calls to the methods in
braille.py is bad. Scripts should be using braille generators instead of
cobbling their own strings together in most cases. (That's a different
bug I filed yesterday.) When it is not possible for scripts to do the
ideal thing, they should use convenience methods from a central location
(i.e. the default script). That way, I don't have to add a bazillion "Is
braille enabled?" checks, but rather only a handful would be needed.

What I checked in is the bulk of the work needed to create those
convenience methods and cause scripts to use them. Once I finish that
task, the next task will be to add in those handful of checks to only
perform braille-related tasks if braille is actually being used. And
*that* is what should improve performance, *if* you happen to not be a
braille user.

After that, I'll look for other areas where I can improve performance
and make changes there.

Hope this clarifies things. Take care.
--joanie