Thanks for the quick response. I'll be interested in seeing the results.

I'm working on code now that should address these issues (though some
further decisions will be made based on #2 and #3).

Essentially, we're looking at expanding/collapsing reviews based on the
following logic:

For each review:
    If the user has a pending reply to this review, expand it.
    Else if the user has replied to the review, and there's no further
activity on the review, collapse it.
    Else If the review is newer than the latest change description, AND it's
the latest review from that user, expand it.
    Else if there's activity on the review since the latest change
description and since the last time the user viewed the page, expand it.
    Else, collapse it.

Users can of course manually expand a review.

This should keep the number of visible reviews quite low. Hopefully it'll be
the set of reviews that the user actually wants to see. The review contents
won't be loaded in unless the user does expand the review, so the page
should load a lot faster.

This doesn't solve the issue of one single review with many hundreds of
comments, but I'm hoping that's not a common case, and that would have to be
solved differently.

The the above logic seems broken to someone, please let me know!

Christian


-- 
Christian Hammond - chip...@chipx86.com
Review Board - http://www.review-board.org
VMware, Inc. - http://www.vmware.com


On Fri, Mar 6, 2009 at 4:37 PM, mary <ciaom...@gmail.com> wrote:

>
> answers inline below...
>
> On Mar 6, 3:52 pm, Christian Hammond <chip...@chipx86.com> wrote:
> > A couple more questions. I'm playing around with a couple possible fixes,
> > but need to find out more where the bottleneck is. Clearly it's
> > browser-side, but the question is whether it's the fact that there's a
> lot
> > in the DOM or whether the rendering is the slow part.
> >
> > 1) Is the page still slow once it fully loads?
> >
> Yes, the comment boxes that are used to write comments are slow even
> after the page is fully loaded. It doesn't make sense to me, but this
> is the behavior being seen by most everyone. (This is only the case
> for reviews with many comments.)
>
> In addition to this, some folks are also complaining that the page
> cannot be used until the page is fully loaded - this has not been my
> experience but I wanted to pass it along too.
>
> Also, we are seeing these issues on reviews that have 10-20 different
> file diffs. Now that I mention this I will try to repro with just one
> file diff to see if that makes a difference or not.
>
> And finally, the CPU gets pegged when these pages are loading.
>
> > 2) Can you install the Firebug extension for Firefox and, in the console,
> > type the following:
> >
> >     $(".review .body").hide()
> >
> I will get back to you on this question.
>
> > And see if the page is now faster to interact with? (All the reviews will
> be
> > hidden until you reload, so this is clearly not a fix by itself, but will
> > tell us whether the bottleneck is the DOM or the rendering).
> >
> > 3) Are these comments spread across many reviews? Or does a single review
> > usually have enough comments to cause problems by itself?
>
> Yes, the comments are spread across many reviews.
> I will try to reproduce using a single review to provide more insight
> and get back to you.
>
> >
> > Christian
> >
> > --
> > Christian Hammond - chip...@chipx86.com
> > Review Board -http://www.review-board.org
> > VMware, Inc. -http://www.vmware.com
> >
> > On Fri, Mar 6, 2009 at 3:10 PM, mary <ciaom...@gmail.com> wrote:
> >
> > > Thank-you for making this a priority! I'll keep an eye out for the fix
> > > and grab immediately.
> >
> > > Breaking up the reviews into smaller pieces is not a use case that is
> > > going down well here, but it is known. thanks again!
> >
> > > On Mar 6, 2:54 pm, Christian Hammond <chip...@chipx86.com> wrote:
> > > > We'd have to decide what we're doing to fix this first. Depending on
> what
> > > > that is, it could take a few days to implement, or longer. We can
> make it
> > > a
> > > > priority for beta 1 (the next release), and of course you'd be able
> to
> > > just
> > > > upgrade to a nightly once it's in.
> >
> > > > Short-term, I'd just advise splitting up the changes more, if
> possible.
> > > > Having smaller things to review should mean fewer comments.
> >
> > > > Christian
> >
> > > > --
> > > > Christian Hammond - chip...@chipx86.com
> > > > Review Board -http://www.review-board.org
> > > > VMware, Inc. -http://www.vmware.com
> >
> > > > On Fri, Mar 6, 2009 at 2:51 PM, mary <ciaom...@gmail.com> wrote:
> >
> > > > > Yes, I mean comments not reviews.
> >
> > > > > I'm seeing the same issues on Alpha4 on my test server. It seems to
> me
> > > > > that the loading of the diff fragments across all the comments is
> > > > > causing our problems - loading such a review sometimes crashes the
> > > > > browser (i've seen this on firefox mostly) and in IE the page often
> > > > > shows a script error popup box part way through the load.
> >
> > > > > Changing the page size would help us so much, can you give any
> > > > > indication of a time frame for such a change?
> >
> > > > > We'd benefit from the other suggestions as well, but jsut getting
> > > > > something workable for medium-to-large reviews is our immediate
> > > > > concern.
> >
> > > > > Thanks!
> >
> > > > > On Mar 6, 2:39 pm, Christian Hammond <chip...@chipx86.com> wrote:
> > > > > > I think we should. The thing is that the newest review request is
> at
> > > the
> > > > > > bottom, so it's kinda weird.
> >
> > > > > > Another thing we should look into is auto-collapsing old reviews
> > > (such as
> > > > > > reviews made before the last update to the review request),
> allowing
> > > them
> > > > > to
> > > > > > expand again. This would fetch the collapsed items from the
> server
> > > > > > dynamically.
> >
> > > > > > Scalability of the review request page is certainly something we
> > > should
> > > > > > tackle for 1.0.
> >
> > > > > > As far as using Alpha 2 vs. Alpha 4, if you use Alpha 4 the page
> > > should
> > > > > load
> > > > > > pretty fast, with the diff fragments loading dynamically after.
> Even
> > > with
> > > > > > 150+ comments (do you mean actual comments or reviews, btw?) it
> > > shouldn't
> > > > > > take forever in alpha 4 to display those. Just might take a while
> for
> > > > > those
> > > > > > diff fragments ot finish loading across all comments.
> >
> > > > > > Christian
> >
> > > > > > --
> > > > > > Christian Hammond - chip...@chipx86.com
> > > > > > Review Board -http://www.review-board.org
> > > > > > VMware, Inc. -http://www.vmware.com
> >
> > > > > > On Fri, Mar 6, 2009 at 2:06 PM, David Trowbridge <
> trowb...@gmail.com
> >
> > > > > wrote:
> >
> > > > > > > Perhaps we need to paginate the reviews page in addition to the
> > > diff?
> >
> > > > > > > -David
> >
> > > > > > > On Fri, Mar 6, 2009 at 2:03 PM, mary <ciaom...@gmail.com>
> wrote:
> >
> > > > > > > > Thank you for your reply. The slowness is:
> > > > > > > > 1. page load takes several minutes (the more comments, the
> longer
> > > it
> > > > > > > > takes)
> > > > > > > > 2. typing a comment is very slow on reviews with many
> comments.
> > > > > > > > 3. Scrolling on the review page is painful when many reviews
> > > > > > > > Developers are speculating its due to a huge DOM and say that
> > > > > > > > performance benchmarks seem relative to the document size,
> > > > > complexity,
> > > > > > > > and browser type (Safari works best, then FireFox, then IE.)
> >
> > > > > > > > Can we change the paging size the ReviewBoard uses? That
> would
> > > help
> > > > > us
> > > > > > > > most likely.
> >
> > > > > > > > Further details:
> > > > > > > > yes, we're using memcache. 4G ram.
> >
> > > > > > > > We've been running Alpha2 the past couple weeks. But it seems
> > > > > Alpha1-4
> > > > > > > > also have same issues.
> >
> > > > > > > > On Mar 6, 1:42 pm, Christian Hammond <chip...@chipx86.com>
> > > wrote:
> > > > > > > >> It's important to find out what's causing the slowdown. What
> > > part is
> > > > > > > being
> > > > > > > >> slow? The page itself, or the progressive diffs inside of
> it?
> > > Alpha
> > > > > 1
> > > > > > > >> doesn't have progressive diffs so it will be slower than
> alpha 4
> > > in
> > > > > this
> > > > > > > >> regard.
> >
> > > > > > > >> We have some large review requests like this at VMware too,
> and
> > > > > haven't
> > > > > > > seen
> > > > > > > >> this slowdown.
> >
> > > > > > > >> Are you using memcached on the server?
> >
> > > > > > > >> How much ram do you have on the server?
> >
> > > > > > > >> Christian
> >
> > > > > > > >> --
> > > > > > > >> Christian Hammond - chip...@chipx86.com
> > > > > > > >> Review Board -http://www.review-board.org
> > > > > > > >> VMware, Inc. -http://www.vmware.com
> >
> > > > > > > >> On Fri, Mar 6, 2009 at 10:53 AM, mary <ciaom...@gmail.com>
> > > wrote:
> >
> > > > > > > >> > I was able to roll back to Alpha1 but ReviewBoard
> displayed
> > > the
> > > > > same
> > > > > > > >> > problem... an unresponsive/seriously slow GUI for reviews
> with
> > > > > 150+
> > > > > > > >> > comments. I also upgrade to Alpha4 with the same results.
> >
> > > > > > > >> > Our company cannot use ReviewBoard with this serious
> > > performance
> > > > > > > >> > handicap.
> >
> > > > > > > >> > We've been very happy with ReviewBoard up until now. Is
> anyone
> > > > > looking
> > > > > > > >> > into this perf issue?
> >
> > > > > > > >> > We're going to have to move back to using <cringe>
> CodeStriker
> > > > > soon if
> > > > > > > >> > this isn't addressed.
> >
> > > > > > > >> > On Feb 24, 1:54 pm, mary <ciaom...@gmail.com> wrote:
> > > > > > > >> > > Thanks for the information, I will try it on my test
> server
> > > (and
> > > > > > > yes,
> > > > > > > >> > > I will keep in mind DB changes for future).
> >
> > > > > > > >> > > I've also been having a hard time reproducing the issues
> > > that
> > > > > many
> > > > > > > are
> > > > > > > >> > > reporting but I will continue to try and gather
> information.
> > > The
> > > > > > > >> > > problems are being reported using both IE7 and Firefox 3
> > > (both
> > > > > from
> > > > > > > >> > > Windows and Linux). Everyone is reporting that using
> Firefox
> > > is
> > > > > > > >> > > definitely better than IE7 though, but still very slow.
> I
> > > have
> > > > > > > >> > > reported one bug #906 which I thought was the full
> issue,
> > > but
> > > > > others
> > > > > > > >> > > are reporting the slowness even on reviews with a small
> diff
> > > and
> > > > > > > very
> > > > > > > >> > > few total reviews.
> > > > > > > >> >http://code.google.com/p/reviewboard/issues/detail?id=906
> >
> > > > > > > >> > > It's been reported numerous times that when users open
> up
> > > the
> > > > > review
> > > > > > > >> > > comment pop up box and start to type their comments, the
> > > > > characters
> > > > > > > >> > > typed take forever to show up in the comment field GUI.
> This
> > > is
> > > > > new
> > > > > > > >> > > for us, no one reported this prior to recent alpha
> upgrades
> > > (I
> > > > > think
> > > > > > > >> > > limited to alpha2, although i'm not certain of this.)
> >
> > > > > > > >> > > On Feb 24, 1:39 pm, Christian Hammond <
> chip...@chipx86.com>
> > > > > wrote:
> >
> > > > > > > >> > > > In theory, you should be able to just install the
> > >  1.0alpha1
> > > > > eggs
> > > > > > > of
> > > > > > > >> > both
> > > > > > > >> > > > Review Board and Djblets. This is completely untested
> and
> > > > > > > unsupported,
> > > > > > > >> > > > though. Note that in the future, you'll have a harder
> time
> > > > > with
> > > > > > > this.
> > > > > > > >> > There
> > > > > > > >> > > > were no database schema changes between alpha 1 and 4
> (to
> > > my
> > > > > > > >> > knowledge), but
> > > > > > > >> > > > there may be between 4 and some other version.
> >
> > > > > > > >> > > > I'm going to be committing a fix for interdiffs within
> a
> > > day
> > > > > or
> > > > > > > two.
> > > > > > > >> > You
> > > > > > > >> > > > could wait until then and upgrade to the nightly
> (which
> > > will
> > > > > be
> > > > > > > safe
> > > > > > > >> > enough
> > > > > > > >> > > > for use, as not much has changed since alpha 4).
> >
> > > > > > > >> > > > I don't know what this slow popup dialog issue is.
> I've
> > > heard
> > > > > one
> > > > > > > other
> > > > > > > >> > > > person mention this but I can't reproduce it. The only
> way
> > > > > it'll
> > > > > > > get
> > > > > > > >> > fixed,
> > > > > > > >> > > > though, is if we can gather some debug info and figure
> out
> > > > > what's
> > > > > > > >> > causing
> > > > > > > >> > > > it. Can you tell me what versions of what browsers on
> what
> > > > > > > platforms
> > > > > > > >> > they're
> > > > > > > >> > > > using?
> >
> > > > > > > >> > > > Christian
> >
> > > > > > > >> > > > --
> > > > > > > >> > > > Christian Hammond - chip...@chipx86.com
> > > > > > > >> > > > Review Board -http://www.review-board.org
> > > > > > > >> > > > VMware, Inc. -http://www.vmware.com
> >
> > > > > > > >> > > > On Tue,
> >
> > ...
> >
> > read more ยป
> >
>

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"reviewboard" group.
To post to this group, send email to reviewboard@googlegroups.com
To unsubscribe from this group, send email to 
reviewboard+unsubscr...@googlegroups.com
For more options, visit this group at 
http://groups.google.com/group/reviewboard?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to