can you submit a quickstart that reproduces this?
the lock on session only blocks concurrent requests from the browser,
which is usually not a big deal because most users operate one window
at a time.
also we do not get the same results in our wicket-threadtest project.
-igor
On Tue, Nov 10, 2009 at 10:17 AM, Pamir Erdem <pamir.er...@gmail.com> wrote:
> From wicket source code WebRequestCycleProcessor has a lock on session.
> (Look at the source code below).
> From a profiler we can easily observe that %57 of the time is spent on this
> function especially on lock region.
> Is there any way to speed it up this source code ?
>
>
> /**
> * @see
> org.apache.wicket.request.IRequestCycleProcessor#resolve(org.apache.wicket.RequestCycle,
> * org.apache.wicket.request.RequestParameters)
> */
> public IRequestTarget resolve(final RequestCycle requestCycle,
> final RequestParameters requestParameters)
> {
> IRequestCodingStrategy requestCodingStrategy = requestCycle.getProcessor()
> .getRequestCodingStrategy();
>
> final String path = requestParameters.getPath();
> IRequestTarget target = null;
>
> // See whether this request points to a bookmarkable page
> if (requestParameters.getBookmarkablePageClass() != null)
> {
> target = resolveBookmarkablePage(requestCycle, requestParameters);
> }
> // See whether this request points to a rendered page
> else if (requestParameters.getComponentPath() != null)
> {
> // marks whether or not we will be processing this request
> boolean processRequest = true;
> synchronized (requestCycle.getSession())
> {
> // we need to check if this request has been flagged as
> // process-only-if-path-is-active and if so make sure this
> // condition is met
> if (requestParameters.isOnlyProcessIfPathActive())
> {
> // this request has indeed been flagged as
> // process-only-if-path-is-active
>
> Session session = Session.get();
> IPageMap pageMap =
> session.pageMapForName(requestParameters.getPageMapName(),
> false);
> if (pageMap == null)
> {
> // requested pagemap no longer exists - ignore this
> // request
> processRequest = false;
> }
> else if (pageMap instanceof AccessStackPageMap)
> {
> AccessStackPageMap accessStackPageMap = (AccessStackPageMap)pageMap;
> if (accessStackPageMap.getAccessStack().size() > 0)
> {
> final Access access = (Access)accessStackPageMap.getAccessStack()
> .peek();
>
> final int pageId = Integer
> .parseInt(Strings.firstPathComponent(requestParameters
> .getComponentPath(), Component.PATH_SEPARATOR));
>
> if (pageId != access.getId())
> {
> // the page is no longer the active page
> // - ignore this request
> processRequest = false;
> }
> else
> {
> final int version = requestParameters.getVersionNumber();
> if (version != Page.LATEST_VERSION &&
> version != access.getVersion())
> {
> // version is no longer the active version -
> // ignore this request
> processRequest = false;
> }
> }
> }
> }
> else
> {
> // TODO also this should work..
> }
> }
> }
> if (processRequest)
> {
> try
> {
> target = resolveRenderedPage(requestCycle, requestParameters);
> }
> catch (IgnoreAjaxRequestException e)
> {
> target = EmptyAjaxRequestTarget.getInstance();
> }
> }
> else
> {
> throw new PageExpiredException("Request cannot be processed");
> }
> }
> // See whether this request points to a shared resource
> else if (requestParameters.getResourceKey() != null)
> {
> target = resolveSharedResource(requestCycle, requestParameters);
> }
> // See whether this request points to the home page
> else if (Strings.isEmpty(path) || ("/".equals(path)))
> {
> target = resolveHomePageTarget(requestCycle, requestParameters);
> }
>
> // NOTE we are doing the mount check as the last item, so that it will
> // only be executed when everything else fails. This enables URLs like
> // /foo/bar/?wicket:bookmarkablePage=my.Page to be resolved, where
> // is either a valid mount or a non-valid mount. I (Eelco) am not
> // absolutely sure this is a great way to go, but it seems to have been
> // established as the default way of doing things. If we ever want to
> // tighten the algorithm up, it should be combined by going back to
> // unmounted paths so that requests with Wicket parameters like
> // 'bookmarkablePage' are always created and resolved in the same
> // fashion. There is a test for this in UrlMountingTest.
> if (target == null)
> {
> // still null? check for a mount
> target = requestCodingStrategy.targetForRequest(requestParameters);
>
> if (target == null && requestParameters.getComponentPath() != null)
> {
> // If the target is still null and there was a component path
> // then the Page could not be located in the session
> throw new PageExpiredException(
> "Cannot find the rendered page in session [pagemap=" +
> requestParameters.getPageMapName() + ",componentPath=" +
> requestParameters.getComponentPath() + ",versionNumber=" +
> requestParameters.getVersionNumber() + "]");
> }
> }
> else
> {
> // a target was found, but not by looking up a mount. check whether
> // this is allowed
> if (Application.get().getSecuritySettings().getEnforceMounts() &&
> requestCodingStrategy.pathForTarget(target) != null)
> {
> String msg = "Direct access not allowed for mounted targets";
> // the target was mounted, but we got here via another path
> // : deny the request
> log.error(msg + " [request=" + requestCycle.getRequest() + ",target=" +
> target +
> ",session=" + Session.get() + "]");
> throw new
> AbortWithWebErrorCodeException(HttpServletResponse.SC_FORBIDDEN, msg);
> }
> }
>
> // (WICKET-1356) in case no target was found, return null here.
> RequestCycle will deal with it
> // possible letting wicket filter to pass the request down the filter
> chain
> /*
> if (target == null)
> {
> // if we get here, we have no recognized Wicket target, and thus
> // regard this as a external (non-wicket) resource request on
> // this server
> return resolveExternalResource(requestCycle);
> }
> */
>
> return target;
> }
>
>
> --
> Pamir Erdem
>
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
For additional commands, e-mail: users-h...@wicket.apache.org
