Hi Pablo,
Consider creating a custom profile for dojo. This makes a significant
improvement to the page load time as it reduces the number of http
requests (read on first though)
http://cwiki.apache.org/S2WIKI/creating-a-custom-dojo-profile-for-struts-20x.html
The reload problem you're encountering is that you're either
accumulating too many objects in memory or too many nodes in the DOM.
Use Firebug to make sure the previous versions of the tabs are not still
in the DOM somewhere. Next, you'll have to ensure the widgets objects
Dojo created are released from memory before creating them again after
the reload (I've never done this...not sure how to ask dojo to dispose
of them), and finally you'll have to ensure any listeners attached to
DOM nodes inside your tabs are released (this causes major memory leaks
in IE if they're not). All of this has to be controlled using the Dojo
API, not the struts 2 tags.
The dojo struts2 tags in 2.1 include an attribute to control whether
dojo should parse the content for widgets. This makes a significant
performance improvement rendering tabs that contain a lot of markup.
This feature is available in Struts 2.0 version of dojo, but not as an
attribute of the Struts2 tag. Again, you'll have to dig into the
Dojo/widget API to enable the feature.
When I reached that point of complexity (and slowness) I stopped using
Dojo and switched to another library. When you're using the Dojo API
the benefit of the struts2 tags is completely lost. Most people love
JQuery. I use YUI. Dojo1.x looks good too, but I was too burnt to try
it. Migrating from the struts2 tags may be a significant effort if you
have a lot of notifiers in there, but JQuery and YUI are such a pleasure
to use compared to the Struts2 tags so I think you'll find conversion a
positive experience without regret.
Good luck!
regards,
Jeromy Evans
Pablo Vázquez Blázquez wrote:
Just a bit more information. I have profiled (using firebug) the
action (always the same action with the same content) with and without
the <s:tabbedpanel> tag and here are the results:
Without:
---
1727.355ms
1424.947ms
1748.429ms
1523.518ms
1539.649ms
1606.369ms
1657.058ms
1599.767ms
1632.157ms
1564.907ms
With:
----
2294.719ms
2513.386ms
3398.823ms
4414.39ms
5716.896ms
7383.391ms
10188.268ms
12604.345ms
15876.5ms
20576.395ms
As you can see, executing the action with the tag multiplies by 10 the
response time!, while when not using the tag, response times are more
or less equilibrated.
Jeromy Evans escribió:
Pablo Vázquez Blázquez wrote:
Any idea about why am I getting increasing response times each time
I access to the page?
I have already a great response time the first time I access to the
page (which is also a problem), but my main problem is that after 6
or 7 accesses it in unmanageable...
Thanks!
I don't recall the 0.4 dojo tabbed panel being particularly slow to
render, other than the extreme initial page load while Dojo makes too
many http requests.
I can't think of any reason why it would get slower after a genuine
page reload with the same content (if it's reloading via ajax
though, then there's a good reason). The reload problem is not
evident in the showcase: it's consistently unbearably slow for each
reload at planetstruts.
Dojo parses your html for widget markup, so it there's a lot of
mark-up inside the tabs it's going to be slooowww (such as tables).
Whenever I had a dojo problem I created a simplified page of html
with just the one widget on it to isolate the cause. You're putting
yourself through unnecessary torture by not using a newer client-side
library though.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
