[@All] Mojarra Important Perfomance Problems

[cagatay.civici]

Dave, also if you use server side state saving, give number of views in session param a low value, by default I think it is 15, a bit too much.

Or set state saving to client so that session is not used, a bit more cpu though.

[smithh032772]

Hope it will be fixed.

By the way, Thomas (zoigl), I know that you have provided your great work with Stateless JSF to MyFaces. Was it accepted? Will be it a part of MyFaces sub-projects? If not, we can still add stateless JSF to PrimeFaces Extensions as add-on

Per BalusC's Stateless JSF blog, Mojarra 2.1.19(+) now has stateless JSF and should resolve (or is an attempt to) resolve 'Mojarra Important Performance Problems'?

I think it will not be accepted. The code seems to be inefficient - i didn't check this with a profiler. Also the improvement is much lesser then with mojarra.

I would not say that. We (the MyFaces PMC) really appreciate your work, but we still need to do more testing and hacking. As you said, MyFaces is already pretty fast (approx as fast as Wicket, sometimes even faster). But additional speed can always help

There still needs to be a few things to considered with Stateless JSF. E.g. that it doesn't work with pages which have c:if c:forEach etc in it. Basically all pages which doesn't have a static tree. But we keep working on it (slowly but steadily) and are thankful for your contributions!

txs and LieGrue,
strub

@Mark Struberg, does MyFaces 'need' to adapt this 'Stateless JSF' to address any (outstanding, if any) MyFaces important performance problems, since Mojarra just address this very-important performance problem of Mojarra?

[smithh032772]

The ticket mentioned in the original thread was closed, but we made a new ticket. It's a real problem, but you need a really large page in order to really demonstrate it, as shown in the example attached to the following Java Server Faces ticket:

http://java.net/jira/browse/JAVASERVERFACES-2494

The problem is the way that Mojarra finds nodes in the DOM after you submit an AJAX request for a partial page update. If you have a very large HTML tree, say 10k nodes, then every time you do an ajax request, it searches through all 10k nodes, very very inefficiently. They really need to restructure their code using hashmap searching, or something equivalent, instead of crawling through the HTML tree structure.

JAVASERVERFACES-2494 showed up in my email inbox, since I was following this JIRA, and I recognized that someone referenced the following blog.

JSF-Comparison: MyFaces vs. Mojarra

I know, I know, Ronald, you may want to ask Optimus to LOCK this topic. Right?

[kukeltje]

Hahaha, no these kinds of post are fine... It in fact shows that the performance difference for average pages with around 10-100 components on a page is marginal and not worth the effort for migrating...