Using diferent VOs for Master Detail Navigation (the Declatative Way)

A user on the OTN forum asked a question how to do a master detail like navigation where the master VO is not equal to the detail VO and no accessors or link is available between the tow VOs.

A use case for this scenario is e.g. you have a read only table as master which holds an attribute which is the foreign key to an other table (the master table has a FK to the detail you like to change). In the sample I’m talking about in this blog I used the HR schema, the employees table as master and the the department as detail. I show how to use the employees as read only table, select an employee to edit the department the employee is assigned to.

Here is the data model of the sample:

Data Model

Data Model


As you can see there are no view links defined which could be used to navigate from the employee to the related department.

I’ll do all this the declarative way, so I don’t use a bean or other Java code. I use a bounded task flow and start with a query panel with the read only employees table. Each row shows the id of the employee, the name and the department id. I add a button to the department id of each row and use this to navigate to the departments edit page. Here you see the running app, the query panel which I used to select employees records and the button which I added to the department is column.

Start Screen

Start Screen

I used a button here because of an error in this version (11.1.2) of jdev which prevent the table from selection the current row when you just hit a link in a row. Frank Nimphius provided a workaround for this here:JDeveloper 11.1.2 : Command Link in Table Column Work Around. A click on the ‘Department’ button for ‘Jannette King’ will navigate to Department ’80’ which is editable

Select a Department from the Table and Edit Department

Select a Department from the Table and Edit Department

The work flow is implemented as shown below:

Work Flow

Work Flow

As you see the whole work is done in bounded task flow which first presents the query panel together with the resulting employees table (read only). The column ‘Department ID’ shows the button I use to navigate to the editable departments page. As there is no view link, it’s not enough to select the employees row to mark it as current row. I have to extract the department id from the selected row and use this to search for the department before showing the departments edit page.

I store the department id in a page flow scope variable named ‘#{pageFlowScope.depKey}’. If you like you can store the value of the department id elsewhere e.g. in the variables iterator of the page binding. To extract and store the value I use a af:setPropertyListener which allows to react on the action of the button and transfer the value to page flow scope variable. Here is the code of the department id column:

                            <af:column sortProperty="#{bindings.EmployeesView1.hints.DepartmentId.name}"
                                       sortable="true"
                                       headerText="#{bindings.EmployeesView1.hints.DepartmentId.label}"
                                       id="resId1c4" width="114">
                                <af:outputText value="#{row.DepartmentId}" id="ot5">
                                    <af:convertNumber groupingUsed="false"
                                                      pattern="#{bindings.EmployeesView1.hints.DepartmentId.format}"/>
                                </af:outputText>
                                <af:commandButton text="ShowDepartment" id="cb2" action="showDep">
                                    <af:setPropertyListener from="#{row.DepartmentId}"
                                                            to="#{pageFlowScope.depKey}"
                                                            type="action"/>
                                </af:commandButton>
                            </af:column>

The button action navigates to he method call ‘SetCurrentRowWithKeyValue’ in the bounded task flow. This method I dragged from the data control palette from the DepatermetnsView1 operation onto the bounded task flow definition page

SetCurrentRowWithKeyValue from DepartmentsView1

SetCurrentRowWithKeyValue from DepartmentsView1

The method searches the department using the the value stored in the page flow scope variable. The dialog below opens automatically when you drop the method on the task flow and lets me enter the key value to search for:

setCurrentRowWithKeyValue  Edit Action Binding

setCurrentRowWithKeyValue Edit Action Binding

Here is the pagedef file for the method call:

PageDef of setCurrentRowWithKeyValue Method

PageDef of setCurrentRowWithKeyValue Method


After the search the current row is set in the DepaertmensView1 and I can navigate to to the edit page. That’s about it.

You can download the sample work space from here Sample Workspace blogmasterdetaildeclarative_v2-zip. You have to rename the file to ‘.zip’ after download!

Validate Data before Export via af:exportCollectionActionListener or af:fileDownloadActionListener

ADF rich faces offer a nice and easy feature to stream data to the client (e.g. Excel) using the af:exportCollectionActionListener or af:fileDownloadActionListener component. Both of the components get the output stream from the response, so the application an add the data.
One problem is that the two components fire before the application has a chance to validate other data on the page or do some other needed work.
To overcome this shortcoming you can implement the trick I’m outlining in this blog. The idea is to use a normal af:commandbutton or af:commandToolbarButton on the page which calls an actionListener in a bean. In there you can validate other page data or do some other stuff in the model, and then queue an event inside the bean method to call another invisible button on the page which has the af:exportCollectionActionListener attached.

As it turned out (thanks Jiandong Xin for pointing it out), there is a problem if the button we queue the action event to has a client behavior tag attached. This is the case for af:exportCollectionActionListener or af:fileDownloadActionListener. These tags don’t work properly if we only queue an event to the parent button. As Jobinesh pointed out in his blog we need to insert a JavaScript function into the page which handles the event queuing from JavaScript and call it from the bean. This way we are able to set the needed parameters (e.g. not to wait for a response).

Now lets implement this: Lets start with a toolbar button which resides on a panel collection holding the table we want to export. The button has a binding to the bean (binding=”#{exportBean.exportCollectionButton}”) to make it easier to call it from the bean. If you don’t like to bind the button to the bean, you can search for the component by name inside the action listener we implement below. The page contains the custom handler (customHandler(event)) which we call from the bean method. This handler then queues the event to activate the button which starts the file download. The connection between the customHander and hte commandToolbarButton is the ID of the commandToolbarButton (“ctb1”). You have to change this if your button has a different ID.

<af:resource type="javascript">
  function customHandler(event){
    var exportCmd = AdfPage.PAGE.findComponentByAbsoluteId("ctb1");
    var actionEvent = new AdfActionEvent(exportCmd);
    actionEvent.forceFullSubmit();
    actionEvent.noResponseExpected();
    actionEvent.queue();
  }
</af:resource>
<af:panelCollection id="pc1">
  <f:facet name="menus"/>
    <f:facet name="toolbar">
      <af:toolbar id="t2" binding="#{exportBean.toolbar}">
        <af:commandToolbarButton text="Export..." id="ctb1"
            binding="#{exportBean.exportCollectionButton}"
            visible="false">
          <af:exportCollectionActionListener type="excelHTML" exportedId="t1"
              title="Export"/>
        </af:commandToolbarButton>
...

The vital part is that the visible property is set to “false”, so that the button will not show up on the page, but is fully functional. Somewhere else on the page (or even on the same toolbar) we have an with an action listener pointing to a bean method (exportBean in the sample). This button is visible and used to initiate the export.

...
<af:commandButton text="export via bean" id="cb8" 
    actionListener="#{exportBean.exportprgActionListener}"/>
...

The exportBean actionListener looks like this

    // toolbar button with export listner attatch bound from page
    private RichCommandToolbarButton exportCollectionButton;
...
    public void exportprgActionListener(ActionEvent actionEvent) {
        // Add event code (validation or other) here...

        // queue the event via a JavaScript inserted into the page
        FacesContext context = FacesContext.getCurrentInstance();
        ExtendedRenderKitService erks =
        Service.getService(context.getRenderKit(), ExtendedRenderKitService.class);
        erks.addScript(context, "customHandler();");
    }

In the action listener above you can validate other data on the page or call service methods in your application module before starting the export, or cancel the export if you like. The export runs as if you click the af:exportCollectionActionListener directly.
You can download a sample workspace BlogFileDownLoadTest.zip
After download remove the ‘.doc’ suffix!

Read Only Filter Table with Sort

Just to help somebody out on OTN I made a small demo app to show how to set up a read only ADF table with filter and sort.
The app uses the default HR schema. The Model project contains a VO EmpRoView which is based on a SQL query (select * from Employees) to get its data.
The view controller project has only one page which shows the resulting ADF table. You can enter e.g. ‘<4000' in the salary filter field and hit 'enter' to see the filter working.
The resulting workspace can be downloaded here ROFTTestApp.zip. After downloading the file you have to remove the ‘.doc’ from the filename!

The workspace is build with JDev 11.1.1.5.0 but works an 11.1.2 too.

JDev: Always Test Your App with ApplicationModule Pooling turned off

In the last couple of weeks I saw a couple of question which mentioned a sporadic misbehavior of the application under different circumstances. In the end they could be answered:

“Your application has not been tested with application module pooling turned off”

Whenever you encounter a sporadic misbehavior of the application under different circumstances, this should ring a bell. Most developers came across a situation like this when programming ADF applications. At some point the application does not react normal or throws exceptions. The errors are not reproducible (most of the times) and you only see them on the production server (never on the developer machine).

The first time I came across this problem it took me about a week to figure it out and solve it. Basically the problem has been private data (stored with the session) which is part of the application module and is stored in the user data hash map. This is not a problem as long as you can guarantee that each user always works with the same application module. This is the case when you test run your application on the developers machine (you are the only user and the application module pool always returns the same application module to the client). On a production server where more then one user uses the application at the same time, the application may be forced to reuse an application module which was formally used by a different user. At this point application module pooling take over.

The general algorithm used is that the application module pool has a number of application module available to use. If more requests arrive the pool generates the additional module until a high water mark is reached. Further requests getting rejected. Currently not used modules are given back to the pool and are available for the next request. The pool tries to return the same application module as long as it’s available to the same session for subsequent requests. If it is not available it uses a currently available module stores the current status of the module into a store (DB or file), clears the module from all information and reconstructs the state from the other session from saved state information. Once the application module is restored you can’t distinguish if it’s a new application module or a reused one. This way your application don’t need an module for each user request, but it shares the available modules between the requests, saving lots of resources.

All this can be read about here 43.2 Introduction to Fusion Web Application State Management.

After this more theoretical prologue, lets do a practical project (workspace for JDev 11.1.2 available, see end of article). To make it as simple as possible, but still useful for anybody running into problems with activation/passivation, we use the HR schema and try to emulate a scenario where a user only sees employee data which depends on a department number. This department number should be set in the application module and be accessible for all queries. In a real world scenario this information is connected to the login of a user and stored in a central place. In the sample we use the user data of the application module to store the number.
The applications start page (I do spare the login part) has a af:query panel to select employees which might be filtered by their last name. As there is no login I added a field to enter the department number which should be used to further filter the result set.The Web UI lokke like

Test app UITest Application UI

As you see there is an input field for the last name, in the bottom era an input field to insert the department number and an other panel to retrieve the currently set department number from the application module.
Lets have a look at the service method to get/set the department number and how it is stored:

    private static final String PRIVATEDATA = "privData";

    public void setPrivateToUserData(String aVal) {
        mLogger.info("Set PrivData:" + aVal);
        if (aVal == null)
            return;

        Session lSession = getSession();
        if (lSession == null) {
            mLogger.warning("getSession returned null!");
            return;
        }
        Hashtable lHashtable = lSession.getUserData();
        if (lHashtable == null) {
            mLogger.warning("getUserData returned null!");
            return;
        }
        lHashtable.put(PRIVATEDATA, aVal);
    }

    public String getPrivateToUserData() {
        Session lSession = getSession();
        if (lSession == null) {
            mLogger.warning("getSession returned null!");
            return null;
        }
        Hashtable lHashtable = lSession.getUserData();
        if (lHashtable == null) {
            mLogger.warning("getUserData returned null!");
            return null;
        }
        String lData = (String) lHashtable.get(PRIVATEDATA);
        mLogger.info("Get PrivData: " + lData);
        return lData;
    }

As you see the string from the UI is stored and retrieved under the key PRIVATEDATA = “privData” in the userData hash map of the application module. This is the place to store data for the current user session (Storing Information About the Current User Session)
The configuration of the application module is the default you see after generating a ‘Fusion Web Application’.

ApplicationModule Default Configuration

ApplicationModule Default Configuration


As you can see ‘Application Module Pooling’ is on. When we run the application, set the department number to e.g. 30, send the data to the AM and hit the search button in the query panel we see
Pooling On

Pooling On


You can hit the edit button in a row to edit the employee and come back to the page and see the application running as expected.
Now, we switch ‘Application Module Pooling’ off and run again:
Application Module Pooling Off

Application Module Pooling Off


Doing the same actions as earlier it look like the application does not see the department number at all:
Application Module Pooling Off

Application Module Pooling Off


Even if you set the department number and directly hit the ‘Get User Data’ button you’ll don’t get the department number back.

The reason for this behavior is that we store the department number in the user data hash map which is NOT passivated when the application module is given back to the pool and given to an other requester. This happens every time you go the the server when am pooling is switched off.
What we need to to is to passivate the session user data together with the other state data stored by the framework and load it back when the AM is requested the next time (when it gets activated again). To do this we have to overwrite two methodes in the ApplicationModuleImpl class.

    @Override
    protected void activateState(Element aElement) {
        super.activateState(aElement);
        mLogger.info("++++++++++ activateState");

        Hashtable lData = getSession().getUserData();
        if (aElement != null) {
            // 1. Search the element for any <PrivData> elements
            NodeList nl = aElement.getElementsByTagName(PRIVATEDATA);
            if (nl != null) {
                // 2. If any found, loop over the nodes found
                for (int i = 0, length = nl.getLength(); i < length; i++) {
                    // 3. Get first child node of the <PrivData> element
                    Node child = nl.item(i).getFirstChild();
                    if (child != null) {
                        // 4. Set the data value to the user data hashmap
                        String lDataString = child.getNodeValue();
                        String[] lSplitkeyval = lDataString.split(";");
                        for (int ii = 0; ii < lSplitkeyval.length; ii++) {
                            mLogger.fine("..."+lSplitkeyval[ii]);
                            String[] lSplit = lSplitkeyval[ii].split("=");
                            lData.put(lSplit[0], lSplit[1]);
                        }
                        break;
                    }
                }
            }
        }
    }


    @Override
    protected void passivateState(Document aDocument, Element aElement) {
        super.passivateState(aDocument, aElement);
        mLogger.info("---------- passivateState");

        // 1. Retrieve the value of the user data to save and build a string representation
        Session lSession = getSession();
        Hashtable lData = lSession.getUserData();
        String lDataString = "";
        Set<String> keyset = lData.keySet();
        if (!keyset.isEmpty()) {
            Iterator<String> keys = keyset.iterator();
            while (keys.hasNext()) {
                String key = keys.next();
                mLogger.fine("..."+key + "=" + lData.get(key));
                lDataString += key + "=" + lData.get(key) + ";";
            }
        }

        // 2. Create an XML element to contain the value
        Node node = aDocument.createElement(PRIVATEDATA);
        // 3. Create an XML text node to represent the value
        Node cNode = aDocument.createTextNode(lDataString);
        // 4. Append the text node as a child of the element
        node.appendChild(cNode);
        // 5. Append the element to the parent element passed in
        aElement.appendChild(node);
    }

The passivateState method gets a Document and an Element as parameter. After calling super() to let the framework do its work, we get the user data hash map and store each key-value pair in a string which is then appended as a node to the element we got as parameter. In a real world application I would use a java to XML serialization tool like XStream which is capable to store more complex data.
The activateState method gets an Element as parameter which we search for the node we save when passivateState was called and restore the user data hash map.

After putting the two methods in the ApplicationModuleImpl class (the one you get when you create the java classes for an application module) the application run OK again, application module pooling still turned off.

When you examine the sample workspace which you can get here BlogActivatePassivateSample_V1.zip (remove the ‘.doc’ extension after downloading the workspace!) you’ll notice, that EmployeesViewImpl class too have the two methods to store private data which is not automatically saved by the framework. Further there are log messages throughout the code to let you follow the action in the log window.

As you see it’s essential to test an application with application module pooling turned off to find activation/passivation errors before the application goes to production.