Fix PSS algorithm to ensure c:if and ui:include src="#{...}" related use cases 
will work without rely on   
org.apache.myfaces.REFRESH_TRANSIENT_BUILD_ON_PSS_PRESERVE_STATE
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------

                 Key: MYFACES-3329
                 URL: https://issues.apache.org/jira/browse/MYFACES-3329
             Project: MyFaces Core
          Issue Type: Improvement
          Components: JSR-314
            Reporter: Leonardo Uribe
            Assignee: Leonardo Uribe


There is a well known issue related to use c:if and ui:include src="#{...}". It 
has been described in different ways for a long time. For example see:

MYFACES-3271 ui:include tag handler evaluates its bindings to soon during JSF 
lifecycle

There is a lot of related issues but to keep things simple, PSS algorithm is 
based on the following conditions:

1. Each time the view is build by first time or when it is restored, the same 
initial state will be retrieved.
2. It is possible to calculate a delta state, based on that initial state.

In practice, there are some use cases that breaks that:

a. c:if tag
b. c:choose, c:when and c:otherwise
c. ui:include src="#{...}" (usually known as "dynamic include")
d. ui:decorate template="#{...}"
e. c:forEach 

The reason is they change the component tree dynamically. For example a c:if 
tag could be bound to a ValueExpression that in one moment could be true and 
then false, so the "initial state" calculated by PSS algorithm through a 
vdl.buildView() will be different each time, breaking condition 1.

It is true there exists the workaround with 
org.apache.myfaces.REFRESH_TRANSIENT_BUILD_ON_PSS_PRESERVE_STATE, which mark 
some parts of the tree to be restored fully, so it does not matter the initial 
state, at the end just we restore such parts fully and everything works as 
facelets 1.1.x. But even so, it is a workaround, and we need a better solution 
to that.

In practice, points a., b., c. and d. are relevant. The case e. related to 
c:forEach is not because the iterated data has 

usually a life longer than the view (session, conversation or view scope), so 
we can exclude this case. 

The solution to this problem is store the result of the evaluated expressions 
within the view. It is the only that will work 

as good as org.apache.myfaces.REFRESH_TRANSIENT_BUILD_ON_PSS_PRESERVE_STATE. 

To do that we need to generate an unique id per tag and make components and 
tags inside them generate unique tags too but the 

same ids should be generated each time the view is build no matter the 
conditions. For example:

<h:outputText value="-A-"/>
<c:if test="#{condition}">
<h:outputText value="-B-"/>
</c:if>
<h:outputText value="-C-"/>

-C- should have the same id each time, no matter what happen inside c:if. Since 
PSS algorithm uses clientIds to store the "delta" state of the components, 
ensure that is CRITICAL, otherwise the state will get mixed or lost.

Additionally, we must restore that "facelet state" before the view is built to 
use it when the initial state is derived.


--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to