Sometimes, when using <h:commandLink>, <h:commandButton> or <f:ajax>, the action, actionListener or listener method associated with the tag are simply not being invoked. Or, the bean properties are not updated with submitted UIInput values.
What are the possible causes and solutions for this?
Introduction
Whenever an UICommand component (<h:commandXxx>, <p:commandXxx>, etc) fails to invoke the associated action method, or an UIInput component (<h:inputXxx>, <p:inputXxxx>, etc) fails to process the submitted values and/or update the model values, and you aren't seeing any googlable exceptions and/or warnings in the server log, also not when you configure an ajax exception handler as per Exception handling in JSF ajax requests, nor when you set below context parameter in web.xml,
<context-param>
<param-name>javax.faces.PROJECT_STAGE</param-name>
<param-value>Development</param-value>
</context-param>
and you are also not seeing any googlable errors and/or warnings in browser's JavaScript console (press F12 in Chrome/Firefox23+/IE9+ to open the web developer toolset and then open the Console tab), then work through below list of possible causes.
Possible causes
UICommand and UIInput components must be placed inside an UIForm component, e.g. <h:form> (and thus not plain HTML <form>), otherwise nothing can be sent to the server. UICommand components must also not have type="button" attribute, otherwise it will be a dead button which is only useful for JavaScript onclick. See also How to send form input values and invoke a method in JSF bean and <h:commandButton> does not initiate a postback.
You cannot nest multiple UIForm components in each other. This is illegal in HTML. The browser behavior is unspecified. Watch out with include files! You can use UIForm components in parallel, but they won't process each other during submit. You should also watch out with "God Form" antipattern; make sure that you don't unintentionally process/validate all other (invisible) inputs in the very same form (e.g. having a hidden dialog with required inputs in the very same form). See also How to use <h:form> in JSF page? Single form? Multiple forms? Nested forms?.
No UIInput value validation/conversion error should have occurred. You can use <h:messages> to show any messages which are not shown by any input-specific <h:message> components. Don't forget to include the id of <h:messages> in the <f:ajax render>, if any, so that it will be updated as well on ajax requests. See also h:messages does not display messages when p:commandButton is pressed.
If UICommand or UIInput components are placed inside an iterating component like <h:dataTable>, <ui:repeat>, etc, then you need to ensure that exactly the same value of the iterating component is been preserved during the apply request values phase of the form submit request. JSF will reiterate over it to find the clicked link/button and submitted input values. Putting the bean in the view scope and/or making sure that you load the data model in #PostConstruct of the bean (and thus not in a getter method!) should fix it. See also How and when should I load the model from database for h:dataTable.
If UICommand or UIInput components are included by a dynamic source such as <ui:include src="#{bean.include}">, then you need to ensure that exactly the same #{bean.include} value is preserved during the view build time of the form submit request. JSF will reexecute it during building the component tree. Putting the bean in the view scope and/or making sure that you load the data model in #PostConstruct of the bean (and thus not in a getter method!) should fix it. See also How to ajax-refresh dynamic include content by navigation menu? (JSF SPA).
The rendered attribute of the component and all of its parents and the test attribute of any parent <c:if>/<c:when> should not evaluate to false during the apply request values phase of the form submit request. JSF will recheck it as part of safeguard against tampered/hacked requests. Storing the variables responsible for the condition in a #ViewScoped bean or making sure that you're properly preinitializing the condition in #PostConstruct of a #RequestScoped bean should fix it. The same applies to the disabled and readonly attributes of the component, which should not evaluate to true during apply request values phase. See also JSF CommandButton action not invoked, Form submit in conditionally rendered component is not processed, h:commandButton is not working once I wrap it in a <h:panelGroup rendered> and Force JSF to process, validate and update readonly/disabled input components anyway
The onclick attribute of the UICommand component and the onsubmit attribute of the UIForm component should not return false or cause a JavaScript error. There should in case of <h:commandLink> or <f:ajax> also be no JS errors visible in the browser's JS console. Usually googling the exact error message will already give you the answer. See also Manually adding / loading jQuery with PrimeFaces results in Uncaught TypeErrors.
If you're using Ajax via JSF 2.x <f:ajax> or e.g. PrimeFaces <p:commandXxx>, make sure that you have a <h:head> in the master template instead of the <head>. Otherwise JSF won't be able to auto-include the necessary JavaScript files which contains the Ajax functions. This would result in a JavaScript error like "mojarra is not defined" or "PrimeFaces is not defined" in browser's JS console. See also h:commandLink actionlistener is not invoked when used with f:ajax and ui:repeat.
If you're using Ajax, and the submitted values end up being null, then make sure that the UIInput and UICommand components of interest are covered by the <f:ajax execute> or e.g. <p:commandXxx process>, otherwise they won't be executed/processed. See also Submitted form values not updated in model when adding <f:ajax> to <h:commandButton> and Understanding PrimeFaces process/update and JSF f:ajax execute/render attributes.
If the submitted values still end up being null, and you're using CDI to manage beans, then make sure that you import the scope annotation from the correct package, else CDI will default to #Dependent which effectively recreates the bean on every single evaluation of the EL expression. See also #SessionScoped bean looses scope and gets recreated all the time, fields become null and What is the default Managed Bean Scope in a JSF 2 application?
If a parent of the <h:form> with the UICommand button is beforehand been rendered/updated by an ajax request coming from another form in the same page, then the first action will always fail in JSF 2.2 or older. The second and subsequent actions will work. This is caused by a bug in view state handling which is reported as JSF spec issue 790 and currently fixed in JSF 2.3. For older JSF versions, you need to explicitly specify the ID of the <h:form> in the render of the <f:ajax>. See also h:commandButton/h:commandLink does not work on first click, works only on second click.
If the <h:form> has enctype="multipart/form-data" set in order to support file uploading, then you need to make sure that you're using at least JSF 2.2, or that the servlet filter who is responsible for parsing multipart/form-data requests is properly configured, otherwise the FacesServlet will end up getting no request parameters at all and thus not be able to apply the request values. How to configure such a filter depends on the file upload component being used. For Tomahawk <t:inputFileUpload>, check this answer and for PrimeFaces <p:fileUpload>, check this answer. Or, if you're actually not uploading a file at all, then remove the attribute altogether.
Make sure that the ActionEvent argument of actionListener is an javax.faces.event.ActionEvent and thus not java.awt.event.ActionEvent, which is what most IDEs suggest as 1st autocomplete option. Having no argument is wrong as well if you use actionListener="#{bean.method}". If you don't want an argument in your method, use actionListener="#{bean.method()}". Or perhaps you actually want to use action instead of actionListener. See also Differences between action and actionListener.
Make sure that no PhaseListener or any EventListener in the request-response chain has changed the JSF lifecycle to skip the invoke action phase by for example calling FacesContext#renderResponse() or FacesContext#responseComplete().
Make sure that no Filter or Servlet in the same request-response chain has blocked the request fo the FacesServlet somehow. For example, login/security filters such as Spring Security. Particularly in ajax requests that would by default end up with no UI feedback at all. See also Spring Security 4 and PrimeFaces 5 AJAX request handling.
If you are using a PrimeFaces <p:dialog> or a <p:overlayPanel>, then make sure that they have their own <h:form>. Because, these components are by default by JavaScript relocated to end of HTML <body>. So, if they were originally sitting inside a <form>, then they would now not anymore sit in a <form>. See also p:commandbutton action doesn't work inside p:dialog
Bug in the framework. For example, RichFaces has a "conversion error" when using a rich:calendar UI element with a defaultLabel attribute (or, in some cases, a rich:placeholder sub-element). This bug prevents the bean method from being invoked when no value is set for the calendar date. Tracing framework bugs can be accomplished by starting with a simple working example and building the page back up until the bug is discovered.
Debugging hints
In case you still stucks, it's time to debug. In the client side, press F12 in webbrowser to open the web developer toolset. Click the Console tab so see the JavaScript conosle. It should be free of any JavaScript errors. Below screenshot is an example from Chrome which demonstrates the case of submitting an <f:ajax> enabled button while not having <h:head> declared (as described in point 7 above).
Click the Network tab to see the HTTP traffic monitor. Submit the form and investigate if the request headers and form data and the response body are as per expectations. Below screenshot is an example from Chrome which demonstrates a successful ajax submit of a simple form with a single <h:inputText> and a single <h:commandButton> with <f:ajax execute="#form" render="#form">.
(warning: when you post screenshots from HTTP request headers like above from a production environment, then make sure you scramble/obfuscate any session cookies in the screenshot to avoid session hijacking attacks!)
In the server side, make sure that server is started in debug mode. Put a debug breakpoint in a method of the JSF component of interest which you expect to be called during processing the form submit. E.g. in case of UICommand component, that would be UICommand#queueEvent() and in case of UIInput component, that would be UIInput#validate(). Just step through the code execution and inspect if the flow and variables are as per expectations. Below screenshot is an example from Eclipse's debugger.
If your h:commandLink is inside a h:dataTable there is another reason why the h:commandLink might not work:
The underlying data-source which is bound to the h:dataTable must also be available in the second JSF-Lifecycle that is triggered when the link is clicked.
So if the underlying data-source is request scoped, the h:commandLink does not work!
While my answer isn't 100% applicable, but most search engines find this as the first hit, I decided to post it nontheless:
If you're using PrimeFaces (or some similar API) p:commandButton or p:commandLink, chances are that you have forgotten to explicitly add process="#this" to your command components.
As the PrimeFaces User's Guide states in section 3.18, the defaults for process and update are both #form, which pretty much opposes the defaults you might expect from plain JSF f:ajax or RichFaces, which are execute="#this" and render="#none" respectively.
Just took me a looong time to find out. (... and I think it's rather unclever to use defaults that are different from JSF!)
I would mention one more thing that concerns Primefaces's p:commandButton!
When you use a p:commandButton for the action that needs to be done on the server, you can not use type="button" because that is for Push buttons which are used to execute custom javascript without causing an ajax/non-ajax request to the server.
For this purpose, you can dispense the type attribute (default value is "submit") or you can explicitly use type="submit".
Hope this will help someone!
Got stuck with this issue myself and found one more cause for this problem.
If you don't have setter methods in your backing bean for the properties used in your *.xhtml , then the action is simply not invoked.
I recently ran into a problem with a UICommand not invoking in a JSF 1.2 application using IBM Extended Faces Components.
I had a command button on a row of a datatable (the extended version, so <hx:datatable>) and the UICommand would not fire from certain rows from the table (the rows that would not fire were the rows greater than the default row display size).
I had a drop-down component for selecting number of rows to display. The value backing this field was in RequestScope. The data backing the table itself was in a sort of ViewScope (in reality, temporarily in SessionScope).
If the row display was increased via the control which value was also bound to the datatable's rows attribute, none of the rows displayed as a result of this change could fire the UICommand when clicked.
Placing this attribute in the same scope as the table data itself fixed the problem.
I think this is alluded to in BalusC #4 above, but not only did the table value need to be View or Session scoped but also the attribute controlling the number of rows to display on that table.
I had this problem as well and only really started to hone in on the root cause after opening up the browser's web console. Until that, I was unable to get any error messages (even with <p:messages>). The web console showed an HTTP 405 status code coming back from the <h:commandButton type="submit" action="#{myBean.submit}">.
In my case, I have a mix of vanilla HttpServlet's providing OAuth authentication via Auth0 and JSF facelets and beans carrying out my application views and business logic.
Once I refactored my web.xml, and removed a middle-man-servlet, it then "magically" worked.
Bottom line, the problem was that the middle-man-servlet was using RequestDispatcher.forward(...) to redirect from the HttpServlet environment to the JSF environment whereas the servlet being called prior to it was redirecting with HttpServletResponse.sendRedirect(...).
Basically, using sendRedirect() allowed the JSF "container" to take control whereas RequestDispatcher.forward() was obviously not.
What I don't know is why the facelet was able to access the bean properties but could not set them, and this clearly screams for doing away with the mix of servlets and JSF, but I hope this helps someone avoid many hours of head-to-table-banging.
I had lots of fun debugging an issue where a <h:commandLink>'s action in richfaces datatable refused to fire. The table used to work at some point but stopped for no apparent reason. I left no stone unturned, only to find out that my rich:datatable was using the wrong rowKeyConverter which returned nulls that richfaces happily used as row keys. This prevented my <h:commandLink> action from getting called.
One more possibility: if the symptom is that the first invocation works, but subsequent ones do not, you may be using PrimeFaces 3.x with JSF 2.2, as detailed here: No ViewState is sent.
I fixed my problem with placing the:
<h:commandButton class="btn btn-danger" value = "Remove" action="#{deleteEmployeeBean.delete}"></h:commandButton>
In:
<h:form>
<h:commandButton class="btn btn-danger" value = "Remove" action="#{deleteEmployeeBean.delete}"></h:commandButton>
</h:form>
This is the solution, which is worked for me.
<p:commandButton id="b1" value="Save" process="userGroupSetupForm"
actionListener="#{userGroupSetupController.saveData()}"
update="growl userGroupList userGroupSetupForm" />
Here, process="userGroupSetupForm" atrribute is mandatory for Ajax call. actionListener is calling a method from #ViewScope Bean. Also updating growl message, Datatable: userGroupList and Form: userGroupSetupForm.
<ui:composition>
<h:form id="form1">
<p:dialog id="dialog1">
<p:commandButton value="Save" action="#{bean.method1}" /> <!--Working-->
</p:dialog>
</h:form>
<h:form id="form2">
<p:dialog id="dialog2">
<p:commandButton value="Save" action="#{bean.method2}" /> <!--Not Working-->
</p:dialog>
</h:form>
</ui:composition>
To solve;
<ui:composition>
<h:form id="form1">
<p:dialog id="dialog1">
<p:commandButton value="Save" action="#{bean.method1}" /> <!-- Working -->
</p:dialog>
<p:dialog id="dialog2">
<p:commandButton value="Save" action="#{bean.method2}" /> <!--Working -->
</p:dialog>
</h:form>
<h:form id="form2">
<!-- .......... -->
</h:form>
</ui:composition>
While I was typing this I figured out what was causing my issue, but I haven't figured out why it was causing an issue, and it seems strange to me. Maybe someone could explain... Here is my situation (question is at the bottom):
I have 2 modal popups.
PopupControlID as follows:
popup1
popup2
calling from server side code:
popup1.Show() 'works fine
popup2.Show() 'works fine
at some point though:
popup1.Show() 'inspecting on server side all properties definitely belong to
'popup1, but when the ajax .js is triggered, popup2
'property values are being used
Everything was set up properly for these controls to run, but something outside of their setup was cauing a problem.
In page_load I was setting the target popup control for popup2 (which was a div) to .Visible = False. When I noticed this code I deleted it because it was unnecessary since I knew the target control was hidden using style="display:none;". After removing this, I ran the code, not expecting this to fix the bug, but the conflict disappeared. I do not understand how this could affect my pop ups in such a manner.
THE QUESTION:
Does anybody know why popup1.Show() would go ahead and use popup2's properties instead of it's own in the ajax script, just because at some point in the code the target control of popup2 was set to .visible = false?
In replicating my issue it seems the answer is as simple as this:
Any time you set a control to .Visible = "False" that is also the PopupControlID of a modal popup extender it will cause an error and the page will not load. I feel kinda stupid for now realizing this up front, a real "duh" moment...
Anyway, pretty basic stuff when I realized what was happening. Here is an article related to this issue:
http://blog.matthewdoyle.net/2008/10/13/visiblefalse-vs-displaynone/
There is a table like this:
<h:panelGroup rendered="#{ImportFromCSVFile.step2Visible}">
<div style="width: 100%; overflow: auto; max-height: 400px;">
<tr:table id="step2ColumnMappings" var="columnMappingEntry" rows="0" rowBandingInterval="1" value="#{ImportFromCSVFile.columnMappings}">
<tr:column>
<tr:outputLabel value="#{columnMappingEntry.columnIndex}"/>
</tr:column>
<tr:column>
<tr:outputLabel value="#{columnMappingEntry.columnValue}"/>
</tr:column>
<tr:column>
<h:selectOneMenu value="#{columnMappingEntry.columnType}" validator="#{ImportFromCSVFile.validateColumnType}" onchange="submit()" valueChangeListener="#{ImportFromCSVFile.columnMappingChanged}">
<f:selectItems value="#{ImportFromCSVFile.columnsToBeMapped}" />
</h:selectOneMenu>
</tr:column>
</tr:table>
</div>
</h:panelGroup>
which renders (properly) as a table with three columns: text in the first two and dropdowns in the last one. dropdowns are also properly initialized based on values in the model.
What I need to do is to be able to run some logic everytime a value in any of the dropdowns changes. I was thinking of using valueChangeListener on selectOneMenu, as you can see in the code, but it doesn't get called. The only thing that comes to mind right now is working with POST parameters when the form is submitted, which would not be optimal.
Do you know how I can get valueChangeListener to work in this context? JSF 1.2. I may be missing an obvious or known solution, but unfortunately I have very very little prior experience with JSF and do not work with it on a regular basis.
Thank you very much in advance!
xmlns:h="http://java.sun.com/jsf/html"
xmlns:tr="http://myfaces.apache.org/trinidad"
The easiest thing to do would be to use <tr:selectOneChoice instead of <h:selectOneChoice. Use it's valueChangeListener in the same way. However, you can't just submit the form via Javascript like that because it is not aware of the JSF life-cycle (not post'ing the right stuff and your code shouldn't interact with that level of JSF internals anyway). If you set <tr:selectOneChoice autoSubmit="true"... this will cause a post to occur when the drop-down menu value changes that calls your value change listener.
Invoking a javascript function like the way you mentioned in the code snippet will not submit your form.To invoke
valuechange
listener you need to submit a request to the backing bean which will check your application level logic. You can try removing the
onClick
attribute and also please put a message as you have placed a
validator
. At times it so happen, that when validation is not succeeded and you have not attached a corresponding message or notification for the same, the desired action won't be performed. So even though your logic might be write but he JSF lifecycle interrupts at
validation phase
and your remaining action is not invoked.
I wonder if anyone else has seen this before:
I have the following ZCML:
<browser:page
for="*"
name="workflow_action"
class=".bika_listing.WorkflowAction"
permission="zope.Public"
/>
When I try to visit siteroot/workflow_action, Plone tells me "This page does not seem to exist…"
The ZCML is being read - if I cause an error in the ZCML, Plone fails to start.
I've tried for="Products.CMFPlone.interfaces.IPloneSiteRoot" with the same result.
The code in WorkflowAction is not being invoked at all - if I place a pdb in __call__, it does not fire.
I have two similar ZCML definitions for different contexts, which work fine and look something like this:
<browser:page
for="bika.lims.interfaces.IAnalysisRequest"
name="workflow_action"
class=".analysisrequest.WorkflowAction"
permission="zope.Public"
/>
If I remove these other views, leaving only one view in the system with name="workflow_action", (the one for IPloneSiteRoot or "*"), the result is the same.
This seems like a very simple view - I don't know what other information I can put here?
Using Plone 4.0.7.
it must be me!?
Most likely your browserview raises an exception in the __init__() method and so the exception is ignored and never displayed, and the result usually is a 404 error.
I have a similar issue when add a Solgema.NavigationPortlet portlet on a Private state folder, then I found the browser:page navTreeitem in navigation.py raised a KeyError, which cause a 404 response for client browser.
Thanks for Giacomo's answer.