[jbossseam-issues] [JBoss JIRA] Commented: (JBSEAM-2023) short-circuit page actions on first navigation event

[Dan Allen (JIRA)]

    [ http://jira.jboss.com/jira/browse/JBSEAM-2023?page=comments#action_12379797 ] 
            
Dan Allen commented on JBSEAM-2023:
-----------------------------------

After writing a 250 line unit test, I definitely understand how these page actions currently work and even a stronger feeling about why it should be changed.

Assume that you have a sequence of page actions configured for a view-id. If the first page action results in a navigation event that changes the viewId, then the next action that executes is going to look for navigation rules for this new view-id, rather than the original. This means that while actions are executing, the navigation rule matches are jumping all over the place.

Quick example:

	<page view-id="/action-test99a.xhtml">
		<action execute="#{testActions.nonNullActionA}" />
		<action execute="#{testActions.nonNullActionB}" />
		
		<navigation from-action="#{testActions.nonNullActionA}">
			<rule if-outcome="outcomeA">
				<render view-id="/action-test99b.xhtml" />
			</rule>
		</navigation>

               <navigation from-action="#{testActions.nonNullActionB}">
			<rule if-outcome="outcomeB">
				<render view-id="/pageC.xhtml" />
			</rule>
		</navigation>
	</page>
	
	<page view-id="/action-test99b.xhtml">
		<action execute="#{testActions.nonNullActionC}" />

		<navigation from-action="#{testActions.nonNullActionB}">
			<rule if-outcome="outcomeB">
				<render view-id="/pageB.xhtml" />
			</rule>
		</navigation>

When you request /action-test99a.xhtml, here is what happens:

1. nonNullActionA() executes
2. viewId changes to /action-test99b.xhtml (a new UIViewRoot is created as well)
3. nonNullActionB() executes
4. viewId changes to /pageB.xhtml (a new UIViewRoot is created as well)
5. /pageB.xhtml is rendered

What you may find strange is that /pageC.xhtml is not rendered (even though it appears at first glance like it would match the second action) and nonNullActionC() is never called, even though navigation rules are being taken from that page node.

With my proposed changes, here is what would actually happen

1. nonNullActionA() executes
2. viewId changes to /action-test99b.xhtml (a new UIViewRoot is created as well)
3. /action-test99b.xhtml is rendered

> short-circuit page actions on first navigation event
> ----------------------------------------------------
>
>                 Key: JBSEAM-2023
>                 URL: http://jira.jboss.com/jira/browse/JBSEAM-2023
>             Project: JBoss Seam
>          Issue Type: Feature Request
>          Components: Core
>    Affects Versions: 2.0.0.CR1
>            Reporter: Dan Allen
>         Assigned To: Dan Allen
>            Priority: Minor
>             Fix For: 2.0.x
>
>         Attachments: JBSEAM-2023-v1.txt
>
>   Original Estimate: 2 hours
>  Remaining Estimate: 2 hours
>
> Page actions currently have one severe quirk. If multiple page actions are configured for a view-id pattern, they all execute without checking whether or not a navigation has occurred (a navigation match was found). What is even more strange is that even when a navigation match is found and the JSF NavigationHandler has processed the navigation, the actions just keep on executing. I cannot imagine that invoking the navigation handler multiple times was anticipated behavior. Even if it is, the developer is still going to be massively confused which navigation event is going to actually stick.
> Here is the current logic:
>       for ( Action action: getActions() )
>       {
>          if ( action.isExecutable() )
>          {
>             String outcome = action.getOutcome();
>             String fromAction = outcome;
>             if (outcome==null)
>             {
>                fromAction = action.getMethodExpression().getExpressionString();
>                result = true;
>                outcome = Pages.toString( action.getMethodExpression().invoke() );
>                Pages.handleOutcome(facesContext, outcome, fromAction);
>             }
>             else
>             {
>                Pages.handleOutcome(facesContext, outcome, fromAction);
>             }
>          }
>       }
> I think that the short-circuit that needs to take place is that if the response is marked as complete (which indicates a redirect navigation) or if the UIViewRoot changed (which indicates a non-redirect navigation), then the actions should stop executing.
> We should definitely include a test case if we make this change. Its too important to screw up.
> If there is a desire to maintain the previous behavior, then I believe a flag should be added on the page element like so
> <page short-circuit-actions="false">

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira