Eclipse Community Forums: JavaServer Faces » 2.0.1RC1 JSF semantics validator performance

Help

Home

Home » Eclipse Projects » JavaServer Faces » 2.0.1RC1 JSF semantics validator performance

Show: Today's Messages :: Show Polls :: Message Navigator

2.0.1RC1 JSF semantics validator performance [message #475076]

Tue, 11 September 2007 09:11

Vadim Dmitriev

Messages: 74
Registered: July 2009

Member

Hi!

Just installed WTP 2.0.1RC1 and noticed, that JSF Semantics validator
performance degraded about 15-20 times. Is such a performance drop to be
expected in FR of 2.0.1?
Searched bugzilla and haven't found any fixed/opened bugs related to jsf
validator mechanism. Sorry, if I panic too early, but I consider RC's as
bug fixing stage when no new features are added. And since jsf validator
was rather fast in previous 2.0.1 builds, I think it is something
intentionally changed :)

I used eclipse 3.3.1 (last stable M for Win32), sun's jdk 1.5.0_12.

Report message to a moderator

Re: 2.0.1RC1 JSF semantics validator performance [message #475078 is a reply to message #475076]

Tue, 11 September 2007 13:48

Yury Kats

Messages: 104
Registered: July 2009

Senior Member

On 9/11/2007 5:11 AM, Vadim Dmitriev wrote:
> Hi!
>
> Just installed WTP 2.0.1RC1 and noticed, that JSF Semantics validator
> performance degraded about 15-20 times. Is such a performance drop to be
> expected in FR of 2.0.1?
> Searched bugzilla and haven't found any fixed/opened bugs related to jsf
> validator mechanism. Sorry, if I panic too early, but I consider RC's as
> bug fixing stage when no new features are added. And since jsf validator
> was rather fast in previous 2.0.1 builds, I think it is something
> intentionally changed :)

I don't if there were any changes recently, but was it ever fast, really?
Maybe you were testing with different JSPs? The more complex the JSP,
the slower it gets, especially when there are many EL expressions.

This bug has been open for a while now, and just deferred to 3.0:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=163890

Report message to a moderator

Re: 2.0.1RC1 JSF semantics validator performance [message #475080 is a reply to message #475078]

Tue, 11 September 2007 16:42

Vadim Dmitriev

Messages: 74
Registered: July 2009

Member

Yury Kats wrote:
> On 9/11/2007 5:11 AM, Vadim Dmitriev wrote:
>
> I don't if there were any changes recently, but was it ever fast, really?
> Maybe you were testing with different JSPs? The more complex the JSP,
> the slower it gets, especially when there are many EL expressions.
>
> This bug has been open for a while now, and just deferred to 3.0:
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=163890

Thanks for reply, Yury.

I was testing on project, that is almost complete and now it is in
testing stage so JSPs are mainly stay unchanged for a long time. It
consist of relatively 250 JSPs and with even 3 weeks-old WTP2.0.1 build
validated in 4 minutes topmost (the same for 2.0 release). With 2.0.1RC1
estimated completion time is about 40 or so minutes. It looks like
validation algorithm changed and i don't have nor time neither i have
will to check validation sources myself to see whether it is true :)
Well, maybe it worth to wait for RC2.

Best regards.

Report message to a moderator

Re: 2.0.1RC1 JSF semantics validator performance [message #475081 is a reply to message #475080]

Tue, 11 September 2007 17:54

Cameron Bateman

Messages: 481
Registered: July 2009

Senior Member

Vadim Dmitriev wrote:
> validated in 4 minutes topmost (the same for 2.0 release). With 2.0.1RC1
> estimated completion time is about 40 or so minutes. It looks like
> validation algorithm changed and i don't have nor time neither i have
> will to check validation sources myself to see whether it is true :)
> Well, maybe it worth to wait for RC2.

I am seeing significant degradation on the variable resolver stress test:

===================================================
Report for performance test: Stress Variable Resolver
Number of iterations: 100000
===================================================
Max: 2,469,061,850, Max Index: 0
Min: 2,793,092, Min Index: 71,326
Avg: 3,793,538, StdDev: 8,415,462.297, StdDev Ignore Max/Min: 3,169,308.624
===================================================

Compared with:

===================================================
Report for performance test: Stress Variable Resolver
Number of iterations: 100000
===================================================
Max: 2,812,605,233, Max Index: 0
Min: 156,444, Min Index: 77,049
Avg: 257,688, StdDev: 8,915,112.194, StdDev Ignore Max/Min: 621,480.648
===================================================

In the 2.0 pre-release baseline using Matthias' type info caching
mechanism. The property resolver may have degraded slightly too, but it
looks like by only about 10% which could be within the natural variance of
the test.

If this is a serious issue for you in 2.0 please escalate the bug:

https://bugs.eclipse.org/bugs/show_bug.cgi?id=163890

by ramping the severity. Also, if you could run the StressTest test case
(use -Xmx256M -XX:PermSize=256M) in org.eclipse.jst.jsf.validation.el and
paste your console output to the bug, that would help too.

--Cam

Report message to a moderator

Re: 2.0.1RC1 JSF semantics validator performance [message #475083 is a reply to message #475081]

Wed, 12 September 2007 20:18

Cameron Bateman

Messages: 481
Registered: July 2009

Senior Member

I think I have found the source of degradation in performance for
variables. The JSPModelProcessor that discovers and caches variables
declared by tags in a page is being disposed too early which causes page
re-parses for each variable lookup. See updates to
https://bugs.eclipse.org/bugs/show_bug.cgi?id=163890 for more details.

--Cam

Report message to a moderator

Re: 2.0.1RC1 JSF semantics validator performance [message #475085 is a reply to message #475083]

Sun, 16 September 2007 16:40

Vadim Dmitriev

Messages: 74
Registered: July 2009

Member

Cameron Bateman wrote:
> I think I have found the source of degradation in performance for
> variables. The JSPModelProcessor that discovers and caches variables
> declared by tags in a page is being disposed too early which causes page
> re-parses for each variable lookup. See updates to
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=163890 for more details.
>
>
> --Cam
>

Thanks for your effort, Cameron.
It's a relief to know that jsf validator performance is still
work-in-progress. I'll post my subjective opinion about acceptable
validation speed limits in related bugzilla entry.

P.S. I looked into running stress test you mentioned, but haven't found
one with provided name. Plus, I don't have enough experience with JUnit
(and running JUnits for eclipse plugins). Sorry.

Report message to a moderator