[
https://issues.apache.org/jira/browse/WICKET-3730?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13036710#comment-13036710
]
Martin Grigorov commented on WICKET-3730:
-----------------------------------------
The problem here is that
org.apache.wicket.request.resource.caching.AbstractResourceCachingStrategy.getLastModified(ResourceReference)
uses RequestCycle as holder of the cache for
resourceReference->lastModifiedTime. Thus it is recreated for each request and
helps only in the case when the same resource is decorated several times in one
request. An improvement here is to move the cache on application level and add
logic to ModificationWatcher to clean the entry for each modified resource.
> extremely slow performance in development mode
> ----------------------------------------------
>
> Key: WICKET-3730
> URL: https://issues.apache.org/jira/browse/WICKET-3730
> Project: Wicket
> Issue Type: Bug
> Components: wicket-core
> Affects Versions: 1.5-RC4
> Environment: OS X, java 1.6
> Reporter: Konstantin Ignatyev
> Assignee: Peter Ertl
> Attachments: screenshot-1.jpg, screenshot-visualvm.jpg
>
>
> I just finished migrating application to 1.5 and it became VERY unresponsive
> and slow in development mode. In -deployment- it is OK.
> No changes were made to backend services, only necessary Wicket API use
> changes related to request cycle.
> =======================
> Below is the originals bug: WICKET-3194
> Hello,
> I had a problem with slow loading of pages and response to ajax requests.
> After some debugging I traced the problem to be that wicket constantly tries:
> DEBUG - UrlResourceStream - cannot convert url:
> jar:file:/C:/Users/hok/.m2/repository/org/apache/wicket/wicket/1.5-M3/wicket-1.5-M3.jar!/org/apache/wicket/markup/html/wicket-event.js
> to file (URI is not hierarchical), falling back to the inputstream for
> polling
> DEBUG - ResourceStreamLocator - Attempting to locate resource
> 'org/apache/wicket/markup/html/wicket-event_en_US.js' on path [folders = [],
> webapppaths: []]
> DEBUG - ResourceStreamLocator - Attempting to locate resource
> 'org/apache/wicket/markup/html/wicket-event_en_US.js' using classloader
> sun.misc.Launcher$AppClassLoader@cac268
> and this happens because that by default (or at least I think so) wicket adds
> timestamp on the resources -
> ResourceSettings.setUseTimestampOnResources(true) and every resource is read
> from the jar files on every request. When a resource is in a jar file a
> java.lang.IllegalArgumentException: URI is not hierarchical is thrown in the
> UrlResourceStream constructor and a lot of attempts are made to load the jar
> file through different loaders. In my case this led to a slow response times.
> After disabling timestamp on resources
> (ResourceSettings.setUseTimestampOnResources(false)) the problem disappears
> and the performance is fine. However in the javadoc of
> setUseTimestampOnResources:
> Enabling timestamps on resources will inject the last modification time of
> the resource into the filename (the name will look something like
> 'style-ts1282915831000.css' where the large number is the last modified date
> in milliseconds and '-ts' is a prefix to avoid conflicts with filenames that
> already contain a number before their extension. *
> Since browsers and proxies use the filename of the resource as a cache key
> the changed filename will not hit the cache and the page gets rendered with
> the changed file.
> In this case this useful functionality is lost. Is it possible to have "the
> best of both worlds"? Thanks.
> This issue is raised from the discussion:
> http://apache-wicket.1842946.n4.nabble.com/IResourceSettings-setUseTimestampOnResources-true-and-performance-td3057946.html
> To observe the performance improvement, please change
> getResourceSettings().setUseTimestampOnResources(true);
> to
> getResourceSettings().setUseTimestampOnResources(false);
> in TestApplication. It's most obvious when you press Refresh All link and
> observe the time for the refresh in both cases
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
