CDI session scoped bean not destroyed results in memory leakage - jsf

I have a question regarding the lifecycle of session scoped CDI beans.
As far as I understand, a session scoped CDI bean is constructed by the container when the session starts and destroyed when the session ends. Before the bean is destroyed the #PreDestroy Method is invoked as described here https://docs.oracle.com/javaee/6/tutorial/doc/gmgkd.html. It also says to release resources in this method.
In a JSF application I build I experience Memory Leak because the bean doesn't seem to be destroyed and hence the #PreDestroy Method is not invoked to free some references for the garbage collector. So I built a simple Application to test the behavior. My experience is that the session bean doesn't get destroyed when the session is over and furthermore it doesn't even get destroyed when the memory space is needed. I cannot believe I am the first to encounter this, but I don't find any information about this behavior..
So my question is: Shouldn't a CDI bean be destroyed - and hence the #PreDestroy Method be invoked - immediately after its context expired? And if not shouldn't it be at least destroyed when the space is needed?
My test Application:
I am not allowed to post a picture, but the outline is the very basic jsf webapp generated by eclipse. I also have the beans.xml file.
Test.java:
package com.test;
import java.io.Serializable;
import java.util.ArrayList;
import javax.annotation.PostConstruct;
import javax.annotation.PreDestroy;
import javax.enterprise.context.SessionScoped;
import javax.inject.Named;
#SessionScoped
#Named
public class Test implements Serializable {
/**
*
*/
private static final long serialVersionUID = 1L;
private String test;
private ArrayList<ComplexType> cps;
private ArrayList<ComplexType> cps_2;
#PostConstruct
public void init() {
System.out.println("test postconstruct..");
test = "Cdi Test";
}
#PreDestroy
public void cleanUp() {
cps = null;
cps_2 = null;
System.out.println("test cleanUp....");
}
public void data_1() {
cps = new ArrayList<ComplexType>();
for(int i = 0; i < 800; i++) {
String[] s = new String[100000];
ComplexType cp = new ComplexType(i, s);
cps.add(cp);
System.out.println(i);
}
System.out.println("data_1");
}
public void free_1() {
cps = null;
System.out.println("free_1");
}
public void data_2() {
cps_2 = new ArrayList<ComplexType>();
for(int i = 0; i < 800; i++) {
String[] s = new String[100000];
ComplexType cp = new ComplexType(i, s);
cps_2.add(cp);
System.out.println(i);
}
System.out.println("data_1");
}
public void free_2() {
cps_2 = null;
System.out.println("free_1");
}
public String getTest() {
return test;
}
public void setTest(String test) {
this.test = test;
}
}
ComplexType.java:
package com.test;
public class ComplexType {
private int id;
private String[] name;
public ComplexType(int id, String[] name) {
this.id = id;
this.name = name;
}
public int getId() {
return id;
}
public void setId(int id) {
this.id = id;
}
public String[] getName() {
return name;
}
public void setName(String[] name) {
this.name = name;
}
}
index.xhtml:
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:h="http://java.sun.com/jsf/html"
xmlns:f="http://java.sun.com/jsf/core"
>
<h:head>
<title>Cdi test </title>
</h:head>
<h:body>
<h:outputText value="#{test.test}"></h:outputText>
<h:form>
<h:commandButton value="cp_1 data" actionListener="#{test.data_1}">
<f:ajax></f:ajax>
</h:commandButton>
<h:commandButton value="cp_1 Free" actionListener="#{test.free_1}">
<f:ajax></f:ajax>
</h:commandButton>
<br></br>
<h:commandButton value="cp_2 data" actionListener="#{test.data_2}">
<f:ajax></f:ajax>
</h:commandButton>
<h:commandButton value="cp_2 Free" actionListener="#{test.free_2}">
<f:ajax></f:ajax>
</h:commandButton>
</h:form>
</h:body>
</html>
I open the index.xhtml page and the #PostConstruct Method gets invoked as expected. The heap space is exceeded when I invoke data_1 and data_2 both without freeing in between. When I free one of the resources in between or I invoke one method twice in a row the heap space is enough, as the garbage collector frees the memory. This works as I would expect it to work.
But when I invoke one data function, close the browser and hence the session, open a new browser and invoke one of the data functions again, then the application stops working as (I guess) the memory space is exceeded. The point is: the first session bean doesn't get destroyed and its #PreDestroy Method not invoked and therefore the ArrayList is still in the memory.
Can someone please explain to me what is going on here? Shouldn't a CDI bean be destroyed by the container as soon its context expires so that references can be set to null and the garbage collector can free resources?
I am using JBoss AS 7.1.1 and its default implementation JSF Mojarra 2.1.

Session beans (regardless CDI or JSF managed) stay alive until some session timeout exceeds (usually 30 minutes by default, dependent on application server), which you can specify in web.xml. Just closing the browser doesn't invalidate session and it wait to be destroyed by servlet container after timeout expiration. So, my assumption, such behaviour is just fine, #PreDestroy method will be invoked later.

The answer of #olexd basically explains what I was getting wrong in my mind, thank you very much! But invalidating the session after a determined period is not an option, so I had to use the comment of #geert3 as well, thank you for that! I am answering my own question to show how I have solved my particular problem in detail here.
What I was wrong about: I thought the session expires as soon as the browser is closed. This is wrong and it makes sense. One may want to close the browser and open it again to work in the same session as before.
For me this behaviour is not appropriate because I want to release resources as soon as the browser gets closed. So the answer is to manually invalidate the session like this:
FacesContext.getCurrentInstance().getExternalContext().invalidateSession();
As soon as this method is called, the #PreDestroy Method is called, exactly as I want it. Now I had to determine when to call this function. I searched for a way to listen to something like a browserclose event. There are the onbeforeunload and onunload events. onunload doesn't seem to work for me in Chrome, but the onbeforeunload does. See also this answer:
https://stackoverflow.com/a/16677225/1566562
So I wrote a hidden button that gets clicked by javascript on beforeunload and invokes an appropriate backingbean method. This works as I would expect it to work. I tested it on Chrome 43.0.2357.65 and IE 11, for now I am content with it. However it doesn't work with onunload, but this is not of concern for me right now.
So my final code likes this:
index.xhtml
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:h="http://java.sun.com/jsf/html"
xmlns:f="http://java.sun.com/jsf/core">
<h:head>
<title>Cdi test</title>
<h:outputScript library="default" name="js/jquery-1.11.3.min.js"
target="head"></h:outputScript>
</h:head>
<h:body>
<h:outputText value="#{test.test}"></h:outputText>
<h:form id="overall">
<h:commandButton value="cp_1 data" actionListener="#{test.data_1}">
<f:ajax></f:ajax>
</h:commandButton>
<h:commandButton value="cp_1 Free" actionListener="#{test.free_1}">
<f:ajax></f:ajax>
</h:commandButton>
<br></br>
<h:commandButton value="cp_2 data" actionListener="#{test.data_2}">
<f:ajax></f:ajax>
</h:commandButton>
<h:commandButton value="cp_2 Free" actionListener="#{test.free_2}">
<f:ajax></f:ajax>
</h:commandButton>
<br></br>
<h:commandButton id="b" style="display:none"
actionListener="#{test.invalidate}"></h:commandButton>
</h:form>
<script type="text/javascript">
$(window).on('beforeunload', function() {
$('#overall\\:b').click();
});
</script>
</h:body>
</html>
Test.java
package com.test;
import java.io.Serializable;
import java.util.ArrayList;
import javax.annotation.PostConstruct;
import javax.annotation.PreDestroy;
import javax.enterprise.context.SessionScoped;
import javax.faces.context.FacesContext;
import javax.inject.Named;
#SessionScoped
#Named
public class Test implements Serializable {
/**
*
*/
private static final long serialVersionUID = 1L;
private String test;
private ArrayList<ComplexType> cps;
private ArrayList<ComplexType> cps_2;
#PostConstruct
public void init() {
System.out.println("test postconstruct..");
test = "Cdi Test";
}
#PreDestroy
public void cleanUp() {
cps = null;
cps_2 = null;
System.out.println("test cleanUp....");
}
public void data_1() {
cps = new ArrayList<ComplexType>();
for (int i = 0; i < 800; i++) {
String[] s = new String[100000];
ComplexType cp = new ComplexType(i, s);
cps.add(cp);
System.out.println(i);
}
System.out.println("data_1");
}
public void free_1() {
cps = null;
System.out.println("free_1");
}
public void data_2() {
cps_2 = new ArrayList<ComplexType>();
for (int i = 0; i < 800; i++) {
String[] s = new String[100000];
ComplexType cp = new ComplexType(i, s);
cps_2.add(cp);
System.out.println(i);
}
System.out.println("data_2");
}
public void free_2() {
cps_2 = null;
System.out.println("free_2");
}
public void invalidate() {
FacesContext.getCurrentInstance().getExternalContext().invalidateSession();
System.out.println("invalidate");
}
public String getTest() {
return test;
}
public void setTest(String test) {
this.test = test;
}
}
Note that I have used JQuery. This works with JBoss AS 7.1.1 and the default Weld implementation as well.
One thing to add: one doesn't have to set all the referenes manually to null. This makes sense as well, as it would be tedious..

Related

Automatically set value of a managed bean variable with JSF

I would like to pass an value to a managed bean under the hood. So I have this managed bean:
#ManagedBean(name = "mbWorkOrderController")
#SessionScoped
public class WorkOrderController {
// more attributes...
private WorkOrder workOrderCurrent;
// more code here...
public WorkOrder getWorkOrderCurrent() {
return workOrderCurrent;
}
public void setWorkOrderCurrent(WorkOrder workOrderCurrent) {
this.workOrderCurrent = workOrderCurrent;
}
}
It holds a parameter workOrderCurrent of the custom type WorkOrder. The class WorkOrder has an attribute applicant of type String.
At the moment I am using a placeholder inside my inputtext to show the user, what he needs to type inside an inputText.
<p:inputText id="applicant"
value="#{mbWorkOrderController.workOrderCurrent.applicant}"
required="true" maxlength="6"
placeholder="#{mbUserController.userLoggedIn.username}" />
What I want to do, is to automatically pass the value of mbUserController.userLoggedIn.username to mbWorkOrderController.workOrderCurrent.applicant and remove the inputText for applicant completely from my form.
I tried to use c:set:
<c:set value="#{mbUserController.userLoggedIn.username}" target="#{mbWorkOrderController}" property="workOrderCurrent.applicant" />
But unfortunatelly I get a javax.servlet.ServletException with the message:
The class 'WorkOrderController' does not have the property 'workOrderCurrent.applicant'.
Does anybody have an advice?
The class 'WorkOrderController' does not have the property 'workOrderCurrent.applicant'.
Your <c:set> syntax is incorrect.
<c:set value="#{mbUserController.userLoggedIn.username}"
target="#{mbWorkOrderController}"
property="workOrderCurrent.applicant" />
You seem to be thinking that the part..
value="#{mbWorkOrderController.workOrderCurrent.applicant}"
..works under the covers as below:
WorkOrderCurrent workOrderCurrent = mbWorkOrderController.getWorkOrderCurrent();
workOrderCurrent.setApplicant(applicant);
mbWorkOrderController.setWorkOrderCurrent(workOrderCurrent);
This isn't true. It works under the covers as below:
mbWorkOrderController.getWorkOrderCurrent().setApplicant(applicant);
The correct <c:set> syntax is therefore as below:
<c:set value="#{mbUserController.userLoggedIn.username}"
target="#{mbWorkOrderController.workOrderCurrent}"
property="applicant" />
That said, all of this isn't the correct solution to the concrete problem you actually tried to solve. You should perform model prepopulating in the model itself. This can be achieved by using #ManagedProperty to reference another bean property and by using #PostConstruct to perform initialization based on it.
#ManagedBean(name = "mbWorkOrderController")
#SessionScoped
public class WorkOrderController {
#ManagedProperty("#{mbUserController.userLoggedIn}")
private User userLoggedIn;
#PostConstruct
public void init() {
workOrderCurrent.setApplicant(userLoggedIn.getUsername());
}
// ...
}
Perhaps you could explain the context a bit more, but here's another solution. If you're navigating from another page, you can pass some identifier of work WorkOrder in the URL, like this http://host:port/context/page.xhtml?workOrderId=1.
Then, you can set the identifier in the managed bean like this:
<h:html>
<f:viewParam name="workOrderId" value="#{mbWorkOrderController.id}"/>
</h:html>
You'll have to add a new property to your bean:
public class WorkOrderController {
private long id;
public long getId() { return id; }
public void setId(long id) { this.id = id; }
// ...
}
And then, after the property has been set by JSF, you can find the work order in a lifecycle event:
<h:html>
<f:viewParam name="workOrderId" value="#{mbWorkOrderController.id}"/>
<f:event type="preRenderView" listener="#{mbWorkOrderController.findWorkOrder()}"/>
</h:html>
public class WorkOrderController {
private long id;
public long getId() { return id; }
public void setId(long id) { this.id = id; }
public void findWorkOrder() {
this.workOrderCurrent = null /* some way of finding the work order */
}
// ...
}
This strategy has the advantage of letting you have bookmarkable URLs.

Value of <h:selectBooleanCheckbox> is not updated according to backing bean value

I'm using JSF2 and Tomcat server. I programmed a simple example in which:
User Selects a faculty from "h:selectOneMenu"
Upon selection, the value of "h:inputText" is changed to "odd" or "even" based on facultyNo
Also, upon selection, the value of "h:selectBooleanCheckBox" is changed to "checked" if facultyNo is even and "not checked" if facultyNo is odd
Everything works fine for "h:inputText". On the other hand, the value of "h:selectBooleanCheckBox" does not change. Why is this happening?
By the way, the use of boolean value inside a HashMap is intentional because the project I'm working on has lots of boolean values in a HashMap. So, replacing the Hashmap with a simple boolean property and using a getter and a setter for it is definitely not a solution for my case.
The code for the xhtml page is below:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:ui="http://java.sun.com/jsf/facelets"
xmlns:h="http://java.sun.com/jsf/html"
xmlns:f="http://java.sun.com/jsf/core">
<head>
<title>Dummy Title</title>
</head>
<body>
<h:form>
<label for="faculty">Faculty</label>
<h:selectOneMenu id="faculty" value="#{test.selectedFaculty}" converter="faccon" valueChangeListener="#{test.facultyChange}" onchange="submit()">
<f:selectItems value="#{start.app.faculties}"/>
</h:selectOneMenu>
<h:selectBooleanCheckbox id="mycheck" value="#{test.x.get(0)}"></h:selectBooleanCheckbox>
<h:outputText value="#{test.res}"></h:outputText>
<h:commandButton value="Save" action="#{test.saveChoices}" />
</h:form>
</body>
</html>
The code for the backing bean is below
import java.io.Serializable;
import java.util.HashMap;
import javax.annotation.PostConstruct;
import javax.faces.bean.ManagedBean;
import javax.faces.bean.RequestScoped;
import javax.faces.event.ValueChangeEvent;
import com.myWork.Application;
import com.myWork.Faculty;
#ManagedBean(name="test")
#RequestScoped
public class TestBean implements Serializable
{
private HashMap<Integer,Boolean> x;
private String res;
private Faculty selectedFaculty;
#PostConstruct
public void init(){
Application app = Application.getInstance();
selectedFaculty = app.getFaculties()[0];
x = new HashMap<Integer, Boolean>();
if (selectedFaculty.getFacultyNo()%2==0)
{
x.put(0, true);
res = "even";
}
else
{
x.put(0, false);
res = "odd";
}
}
public HashMap<Integer,Boolean> getX() {
return x;
}
public void setX(HashMap<Integer,Boolean> x) {
this.x = x;
}
public Faculty getSelectedFaculty() {
return selectedFaculty;
}
public void setSelectedFaculty(Faculty selectedFaculty) {
this.selectedFaculty = selectedFaculty;
}
public String getRes() {
return res;
}
public void setRes(String res) {
this.res = res;
}
public void facultyChange(ValueChangeEvent e){
Faculty fac = (Faculty) e.getNewValue();
if (fac.getFacultyNo()%2==0)
{
x.put(0, true);
res = "even";
}
else
{
x.put(0, false);
res = "odd";
}
}
public String saveChoices(){
return "test";
}
}
Any help is greatly appreciated.
Actually your problem is not related to JSF, it is related to EL. in your case, EL treat you map keys as long not int, so you should change your map to be:
HashMap<Long, Boolean> map = new HashMap<>();
and set the key values as follows :
map.put(0L, false);
map.put(1L, true);
to force it to be auto-boxed to long.
Note: the problem described in detailed in the following question :
EL access a map value by Integer key
In order to send updates to a different component, you can make use of partial page rendering by making Ajax calls to send parameters to Managed bean.
Within <h:selectBooleanCheckBox>
<h:selectOneMenu id="faculty" value="#{test.selectedFaculty}" converter="faccon" valueChangeListener="#{test.facultyChange}">
<f:selectItems value="#{start.app.faculties}"/>
</h:selectOneMenu>
<h:selectBooleanCheckbox id="mycheck" valueChangeListener="#{test.facultyChange}" value="#{test.x.get(0)}">
<f:ajax event="change" execute="#form" render="faculty"/>
</h:selectBooleanCheckbox>
Edit: Removed call to onSubmit() in selectOneMenu. introduced an ajax call in selectBooleanCheckbox for partial page refresh.
Refer full solution at this post.

p:commandbutton update not updating immediately for ScheduledExecutorService

I'm working on a JSF + Primefaces Web App in which I need to make a RESTful GET request every 20 seconds to a server, and display the data returned. This currently works, with the exception of the "every 10 seconds" part. I can click my command button to retrieve and show the data. I tried to implement the repitition with ScheduledExecutorService. Now when I click the commandButton, the backing function runs every 10 seconds (I can see this through System.out.println output), but the table will not update to show data until the button is clicked a second time. Here is my code below:
HTML (This is embedded inside a form)
<p:commandButton id="andon_layout--board0--loadboard0"
value="Load board"
actionListener="#{decryptionBean.loadBoardListen0}"
update="dataPanelGrid" />
<div id="andon_layout--board0--display_div"
class="ui-datatable ui-widget ">
<h:panelGrid id="dataPanelGrid"
columns="#{decryptionBean.displayBoardArray[0].datatableNumberOfCols}"
headerClass="ui-datatable ui-widget-header ">
<f:facet name="header">
<h:outputText
value="#{decryptionBean.displayBoardArray[0].locationName}" />
</f:facet>
<c:forEach var="row"
items="#{decryptionBean.displayBoardArray[0].displayData}">
<c:forEach var="value" items="#{row}">
<div class="ui-dt-c">#{value}</div>
</c:forEach>
</c:forEach>
</h:panelGrid>
</div>
JAVA
#ManagedBean
#ViewScoped
public class DecryptionBean implements Serializable {
...
private Updater left = new Updater(0);
public void loadBoardListen0(ActionEvent event){
left.stopBoardLoading();
left.beginBoardLoading();
}
public final class Updater {
#PreDestroy
public void destroy() {
fScheduler.shutdownNow();
}
private ScheduledExecutorService fScheduler;
//
private long fDelayBetweenRuns = 10;
int boardNumber;
/**
* If invocations might overlap, you can specify more than a single
* thread.
*/
private int NUM_THREADS = 1;
public int isRunning = 0;
private boolean DONT_INTERRUPT_IF_RUNNING = false;
private ScheduledFuture<?> loadBoardFuture;
private class BoardLoaderTask implements Runnable {
private int boardNumber;
public BoardLoaderTask(int boardNumber) {
this.boardNumber = boardNumber;
}
public void run() {
DecryptionBean.this.loadBoard(boardNumber);
}
}
public Updater(int boardNumber){
this.boardNumber = boardNumber;
fScheduler = Executors.newScheduledThreadPool(NUM_THREADS);
}
void beginBoardLoading(){
if(isRunning == 1){
this.stopBoardLoading();
}
isRunning = 1;
Runnable boardLoaderTask = new BoardLoaderTask(this.boardNumber);
loadBoardFuture = fScheduler.scheduleWithFixedDelay(boardLoaderTask,
0, fDelayBetweenRuns, TimeUnit.SECONDS);
}
void stopBoardLoading(){
if(isRunning == 1){
isRunning = 0;
Runnable stopBoard = new StopLoadingTask(loadBoardFuture);
fScheduler.schedule(stopBoard, 0, TimeUnit.SECONDS);
}
isRunning = 0;
}
private class StopLoadingTask implements Runnable {
StopLoadingTask(ScheduledFuture<?> aSchedFuture) {
fSchedFuture = aSchedFuture;
}
private ScheduledFuture<?> fSchedFuture;
#Override
public void run() {
// TODO Auto-generated method stub
fSchedFuture.cancel(DONT_INTERRUPT_IF_RUNNING);
}
}
}
loadBoard(int boardNum) is kind of a large function, so I haven't posted it here, however I will if you guys deem it necessary.
It seems to me that the problem comes from the p:commandbutton update attribute: it seems like the dataPanelGrid is being updated at the start of loadBoardListen0, as opposed to the end. To be clear: The backing function runs on the first click, but only updates the GUI on the second, third, fourth etc. click Any guidance on this one? Thanks in advance!
EDIT 1:
I'm now trying to do this using PrimePush and a socket, but I'm running into trouble there too. I added this to my code:
HTML
<p:socket onMessage="handleMessage" channel="/IPVS" autoConnect="false"/>
...
function handleMessage(data) {
console.log("data received: " + data);
var elm = "andon_layout--board" + data + "--display_div";
elm.style.display="none";
var redrawFix = elm.offsetHeight;
elm.style.display="block";
}
JAVA (at the end of loadBoard(int))
System.out.println("Loading data... no exceptions thrown");
System.out.println("context created");
PushContextFactory.getDefault().getPushContext().push("/IPVS", Integer.toString(displayBoardNum));
System.out.println("Just pushed to /IPVS");
Whenever I attempt to run the code, it simply halts at the PushContextFactory line. The output I get is as follows:
Loading data... no exceptions thrown
context created
And simply no more. This also prevents the loop from running. Note that I am still using a ScheduledExecutorService, and just trying to push some data every time the task completes. Any ideas?
You should try using the Primefaces poll tag. This will fire an ajax event every x seconds. You could use this to call your "getData" method on your backing bean. This will move the logic of the polling into your JSF page, rather than the backing bean.
For example:
<p:poll interval="10" listener="#{decryptionBean.getData}" update="dataPanelGrid" />
This requires you to have a simple "getData" method in your backing bean, without the scheduling code.

Session Scoped Managed Bean constructor being called on each page refresh

I am using a session scoped managed bean for handling login in a Java EE application. After I authenticate the user, the user object is saved in this session bean. However, after I refresh the page, the session bean values are gone.
I was debugging the code and it results that the constructor of the session scoped managed bean is called again on page refresh, therefore initializing the user object with a new user. I guess this is not a normal behavior since it should be preserved on the session shouldn't it?
I am posting some parts of the login managed bean including the parameters and the login method. Basically the enteredEmail and enteredPassword stand for the entered data on the login form. If the authentication succeeds, the loggedIn boolean is turned to true and the logged in user object is stored in the checkedUser variable.
import javax.faces.bean.ManagedBean;
import javax.faces.bean.SessionScoped;
#ManagedBean
#SessionScoped
public class LoginController implements Serializable {
#EJB
private LoginSessionBean loginSessionBean;
#EJB
private LecturerFacade lecturerFacade;
private Lecturer checkedUser;
private String enteredEmail;
private String enteredPassword;
private boolean loggedIn;
/** Creates a new instance of loginController */
public LoginController() {
loggedIn = false;
checkedUser = new Lecturer();
}
public String login(){
RequestContext context = RequestContext.getCurrentInstance();
FacesMessage msg = null;
this.setCheckedUser(lecturerFacade.findLecturerByEmail(enteredEmail));
if(loginSessionBean.checkPassword(checkedUser, enteredPassword))
{
loggedIn = true;
msg = new FacesMessage(FacesMessage.SEVERITY_INFO, "Welcome", checkedUser.getFirstName()+ " " + checkedUser.getLastName());
FacesContext.getCurrentInstance().addMessage(null, msg);
context.addCallbackParam("loggedIn", loggedIn);
}
return "Index";
I am also posting the two EJBs that the above managed bean uses. The lecturerFacade retrieves the user object with the entered email, while the loginSessionBean checks the password.
#Stateless
public class LecturerFacade extends AbstractFacade<Lecturer> {
#PersistenceContext(unitName = "EffectinetWebPU")
private EntityManager em;
Logger logger = Logger.getLogger("MyLog");
FileHandler fh;
protected EntityManager getEntityManager() {
return em;
}
public LecturerFacade() {
super(Lecturer.class);
}
public Lecturer findLecturerByEmail(String email) {
try {
return (Lecturer) this.getEntityManager().createQuery("SELECT l FROM Lecturer l WHERE l.email = :email").setParameter("email", email).getSingleResult();
} catch (NoResultException e) {
System.err.println("Caught NOResultException: "+ e.getMessage());
return null;
} catch (NonUniqueResultException e) {
System.err.println("Caught NonUniqueResultException: "+ e.getMessage());
return null;
} catch (IllegalStateException e) {
System.err.println("Caught IllegalStateException: "+ e.getMessage());
return null;
}
}
_
#Stateless
public class LoginSessionBean {
// Add business logic below. (Right-click in editor and choose
// "Insert Code > Add Business Method")
#PersistenceContext(unitName = "EffectinetWebPU")
private EntityManager em;
protected EntityManager getEntityManager() {
return em;
}
public void setEntityManager(EntityManager em) {
this.em = em;
}
public boolean checkPassword(Lecturer user, final String enteredPassword) {
if (user.getPassword().equals(enteredPassword)) {
return true;
} else {
return false;
}
}
}
Please if someone has any suggestion of what is going wrong, please tell me
Im using glassfish 3.1 as application server and Primefaces as JSF library. Also, I have checked and the imported the sessionScoped annotation from the right package and not from javax.enterprise...
Your problem is thus here:
<p:menuitem value="Logout" ... onclick="#{loginController.logout()}"/>
The onclick attribute should represent a JavaScript handler function which is to be executed in the webbrowser when the enduser clicks the element. Something like
onclick="alert('You have clicked this element!')"
The onclick attribute also accepts a ValueExpression, so you can even let JSF/EL autogenerate its value accordingly:
onclick="#{bean.onclickFunction}"
with
public String getOnclickFunction() {
return "alert('You have clicked this element!')";
}
All the EL is thus evaluated when the page is rendered. In your particular case, the logout() method is called everytime the EL is evaluated and thus you're invalidating the session everytime the page is rendered!
You need to bind it to an attribute which takes a MethodExpression like <h:commandLink action>, <h:commandButton action> and in this particular case <p:menuitem action>.
<p:menuitem value="Logout" ... action="#{loginController.logout()}"/>
This can be understood by understanding basic HTML and JavaScript concepts and keeping in mind that JSF ultimately produces HTML/CSS/JS. Open the JSF page in webbrowser, rightclick and View Source to realize it.
Well I managed to solve it today. This was the problem, although I cannot explain why:
I was using Primefaces 3.2 as JSF library so this was the main menu of the index page.
<h:form>
<p:menubar >
<p:menuitem id="registerLink" value="Register" rendered="#{!loginController.loggedIn}" onclick="registerDialog.show()" />
<p:menuitem id="loginLink" value="Login" rendered="#{!loginController.loggedIn}" onclick="loginDialog.show()" />
<p:submenu label="Units" rendered="true">
<p:menuitem id="addNew" value="Add New" onclick="createUnitDialog.show()" />
<p:menuitem id="myUnits" value="My Units" onclick="" />
</p:submenu>
<p:menuitem id="results" value="Results/Statistics" rendered="#{loginController.loggedIn}" onclick=""/>
<p:menuitem id="profile" value="My Profile" rendered="#{loginController.loggedIn}" onclick=""/>
<p:menuitem id="logout" value="Logout" rendered="#{loginController.loggedIn}" onclick="#{loginController.logout()}"/>
</p:menubar>
</h:form>
After setting breakpoints to the whole code I discovered that the logout() method, which is supposed to destroy the managed bean, was called on every page refresh. I don't know why this happened as it should be called when the logout menuitem was clicked.
However, after changing the onclick="#{loginController.logout()} with action="#{loginController.logout()} the problem was solved.
I checked the documentation of Primefaces but nowhere this behavior was explained

CDI lifecycle of bean, #Inject and params

The view and bean were working until I tried to fix non-standard names, and I've now broken the connection between the two. Oddly, the "back" button has the correct link, but content just doesn't show, nor log. Why doesn't Detail.getComments() execute?
I've been going through the weld docs and trying to better understand #Inject. There seems to be a lifecycle problem which I don't understand, either. If it's not lifecycle, then I cannot even speculate as to why Detail.getComments() never shows in the glassfish logs:
INFO: MessageBean.getModel..
INFO: SingletonNNTP.returning messages..
INFO: MessageBean.getModel..
INFO: SingletonNNTP.returning messages..
INFO: MessageBean.getModel..
INFO: SingletonNNTP.returning messages..
INFO: Detail..
INFO: Detail.getId..null
INFO: Detail.getId..SETTING DEFAULT ID
INFO: Detail.onLoad..2000
INFO: Detail.getId..2000
INFO: Detail.getId..2000
INFO: Detail.setId..2000
INFO: Detail.getId..2019
INFO: ..Detail.setId 2019
INFO: Detail.back..
INFO: Detail.getId..2019
INFO: ..Detail.back 2,018
INFO: Detail.getId..2019
The value 2000 is a default, which only happens when id==null, which it never should. It should pull in that attribute right away. So, I'm not sure whether that's a problem with the scope (I only just now found out that CDI doesn't support #SessionScoped), the lifecycle, or something else. Perhaps I need to use #Inject on that variable?
The view, detail.xhtml:
<?xml version='1.0' encoding='UTF-8' ?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:ui="http://java.sun.com/jsf/facelets"
xmlns:f="http://java.sun.com/jsf/core"
xmlns:h="http://java.sun.com/jsf/html">
<body>
<f:metadata>
<f:viewParam name="id" id="id" value="#{detail.id}" />
</f:metadata>
<ui:composition template="./complexTemplate.xhtml">
<ui:define name="top">
<h:link value="back" outcome="detail" includeViewParams="true">
<f:param name="id" value="#{detail.back}"/>
</h:link>
<ui:define name="content">
<h:outputText value="#{detail.content}" rendered="false"/>
</ui:define>
<ui:define name="bottom">
bottom
</ui:define>
</ui:composition>
</body>
</html>
and the backing bean:
package net.bounceme.dur.nntp;
import java.io.Serializable;
import java.util.logging.Level;
import java.util.logging.Logger;
import javax.annotation.PostConstruct;
import javax.enterprise.context.ConversationScoped;
import javax.inject.Named;
import javax.mail.Message;
#Named
#ConversationScoped
public class Detail implements Serializable {
private static final long serialVersionUID = 1L;
private static final Logger logger = Logger.getLogger(Detail.class.getName());
private static final Level level = Level.INFO;
private String id = null; //should never get default value in getter
private Message message = null;
private SingletonNNTP nntp = SingletonNNTP.INSTANCE;
private String forward = null; //id + 1
private String back = null; //id - 1
private String content = null; //message.content
public Detail() {
logger.log(level, "Detail..");
}
#PostConstruct
private void onLoad() {
logger.log(level, "Detail.onLoad..{0}", getId());
}
public Message getMessage() {
logger.log(level, "Detail.getMessage..");
return message;
}
public void setMessage(Message message) {
logger.log(level, "Detail.setMessage..");
this.message = message;
}
public String getId() {
logger.log(level, "Detail.getId..{0}", id);
if (id == null) {
logger.log(level, "Detail.getId..SETTING DEFAULT ID");
id = String.valueOf(2000);
}
return id;
}
public void setId(String id) throws Exception {
logger.log(level, "Detail.setId..{0}", getId());
this.id = id;
logger.log(level, "..Detail.setId {0}", getId());
}
public String getForward() {
logger.log(level, "Detail.forward..");
int f = Integer.parseInt(getId());
f = f + 1;
logger.log(level, "..Detail.forward {0}", f);
forward = String.valueOf(f);
return forward;
}
public void setForward(String forward) {
this.forward = forward;
}
public String getBack() {
logger.log(level, "Detail.back..");
int b = Integer.parseInt(getId());
b = b - 1;
logger.log(level, "..Detail.back {0}", b);
back = String.valueOf(b);
return back;
}
public void setBack(String back) {
this.back = back;
}
public String getContent() throws Exception {
logger.log(level, "Detail.getContent..{0}", getId());
message = nntp.getMessage(Integer.parseInt(getId()));
content = message.getContent().toString();
return content;
}
public void setContent(String content) {
this.content = content;
}
}
which never seems to have, according to the above logs, Detail.getContent() invoked, despite that being part of the view: <h:outputText value="#{detail.content}" rendered="false"/>
It's odd in that Detail.content() was getting invoked prior to my changing this class to better follow naming conventions. I'm going through some Weld and Oracle Java EE 6 docs, but don't at all mind being directed to a fine manual. The docs I find describing this are invariably using #ManagedBeans, however, which I am not. There seem many gotchas, as described in this answer by #Kawu.
Adding #Inject to the id field causes a deploy error:
init:
deps-module-jar:
deps-ear-jar:
deps-jar:
library-inclusion-in-archive:
library-inclusion-in-manifest:
compile:
compile-jsps:
In-place deployment at /home/thufir/NetBeansProjects/NNTPjsf/build/web
Initializing...
deploy?DEFAULT=/home/thufir/NetBeansProjects/NNTPjsf/build/web&name=NNTPjsf&contextroot=/NNTPjsf&force=true failed on GlassFish Server 3.1.2
Error occurred during deployment: Exception while loading the app : WELD-001408 Unsatisfied dependencies for type [String] with qualifiers [#Default] at injection point [[field] #Inject private net.bounceme.dur.nntp.Detail.id]. Please see server.log for more details.
/home/thufir/NetBeansProjects/NNTPjsf/nbproject/build-impl.xml:749: The module has not been deployed.
See the server log for details.
BUILD FAILED (total time: 9 seconds)
Surely, injecting a String isn't the problem, perhaps it's a bug.
I understand your frustration, and I see that the problem is more your setup / understanding in general. But still, it's pretty hard to find any real questions to answer, maybe you can try to split your problems next time.
Here are some answers:
Why doesn't Detail.getComments() execute?
Hm, maybe because it's not in the bean? I guess that you are refrerring to detail.getContent instead?
which never seems to have, according to the above logs,
Detail.getContent() invoked, despite that being part of the view:
Try rendered = true :-)
#PostConstruct
private void onLoad() {
logger.log(level, "Detail.onLoad..{0}", getId());
}
You've put an awful lot of logic into the getter. Try debugging with the field, not with the getter...
The value 2000 is a default, which only happens when id==null, which it never should.
It looks like private String id = null; is a perfect explanation why id will be null.
Try to keep in mind that modern frameworks like JSF, CDI and Java EE do a lot of stuff behind the scenes, using reflection, proxies and interceptors. Don't rely on classical understanding of when (and how often) a constructor is called, for example.
Again, consider moving your initialisation logic away from the getter. #PostConstruct would be the place that the fathers of the Java EE-spec had chosen for it.
To be honest: Nothing looks extremely wrong, but your code is kind of messy, and extremely hard to understand and to follow.
Try removing all indirections like this one...
int b = Integer.parseInt(getId());
... and everything will look much better.
Oh, and is there a specific reason why you declare a fixed log-level for the whole class? Try something like this
private static final Logger LOG = Logger.getLogger(Some.class);
...
LOG.info("...");
Hope that gives you a start. Feel free to post further questions, preferably a bit shorter and with single, isolated aspects to answer.

Resources