Quo vadis ADF?

Last week I attended DOAG Konferenz & Ausstellung in Nürnberg Germany. The DOAG (Deutsche ORACLE-Anwendergruppe e.V.) is the biggest German Oracle user group. The conference covers all Oracle products and technologies, way too much to name them all.

As my personal center of gravity is middle-ware and here ADF and the surrounding technologies, I attended lot’s of sessions about middle-ware, cloud, ADF, MAF and JET. The big picture of Oracle becoming a cloud company is getting clearer.

The way developers currently are working on premise with their products migrating to the cloud is getting clearer. There where about 4-5 sessions which gave explicit advice when to use which technology and what problems might arise mixing them. I’ll cover the main three here.

Frank Nimphius started with a session ‘The Future of Application Development Welcome to your new Job’ where he summarized areas of future of application development as

  • “Server-less” deployment
  • [Micro] [Cloud] Services
  • Mobile centric
  • API first
  • Multi channel
  • Artificial Intelligence
  • Cloud Native Development
  • JavaScript
Future Application Development Summary 1

Future Application Development Summary 1

Future Application Development Summary 2

Future Application Development Summary 2

and defined different job roles around this like

  • Citizen (Low Code) Developer
  • Mobile Developer
  • Service Developer
  • Architect
  • Line of Business Manager

Each role using different technologies to fulfill the tasks. This should open spaces for new and old developers

Mobile Job Roles

Mobile Job Roles

Duncan Mills tackled the bear from a different perspective. In his session ‘Standing at Crossroads’ (Oracle ADF and Oracle JET) he pointed out the differences between ADF and JET

Oracle ADF Oracle JET
Support 5 + 3 + unlimited, no backport limitations Major release every 6 month, backports only to previous version
API are stable No guarantee of API stability
Could or on premis Cloud
Metadata focused Code focused
Full stack solution Client only solution
Has to „own“ the page Can be used „anywhere“

However, there are things both have in common, as Duncan states:

“Don’t assume the you have to go to JET to look ‘modern'”

“Don’t assume that JET will automatically be more perfomant”

There are more things you have to take into account before making a decision between ADF and JET like

  • Transaction and Services: here you have to check if your services and data model can support a stateless model. Same for your UI which handles the interaction with the user. One thing to note too is that using JET will produce less client – sever traffic.
  • Need to shape the services for the convenience of the UI: paging data, pre-computation, attribute reduction and mega endpoints

If you plan to mix ADF and JET there are a couple of things which should make you think twice:

  1. No session sharing between ADF and JET
  2. ADF and JET can’t use the same cache
  3. No shared transaction
  4. Separate timeouts
  5. geometry management
  6. Drag & drop not possible between ADF and JET
  7. Different maintenance and different libraries
  8. Different popup’s and glasspane

Summary is that there are plenty of reasons not to mix ADF and JET. If you want to mix ADF and JET in a project you should stick to module level and not mix them on one page.


The decision for ADF or  JET should take these points into account.

Shay Shmeltzer attended the German Oracle (ADF) Developer Community meeting on the DOAG and we ask him to talk about this topic ‘The Future of Developer Frameworks’.

shay2_doag1 Shay started by giving a main difference between ADF and JET:

“ADF is a framework, JET is a toolkit”

meaning that ADF allows development in all tires (MVC) whereas JET is only a client technology. Using JET you still have to have a back-end which generates the needed REST services. Here ADF comes into the picture again.

“ADF hides the complexity of the technology from the developer” 

True, building a REST service from an exiting ADFbc model is very easy and allow shaping the service too. Besides ORDS (Oracle REST Data Service, a tooling which allows to develop modern REST interfaces for relational data in the Oracle Database ) this is the easiest way I know.

During the Q&A of his talk we specifically ask him how Oracle sees the future of ADF as some rumors are that ADF is dead. Shay answered (loud and clear):

“ADF isn’t dead!”

Oracle is using ADF heavily in the SaaS products and will going on to do so. There are areas where building UI with JET is preferred (not in SaaS), but here the points mentioned by Duncan Mills are always considered.

My personal opinion is that ADF is alive will be used in the future, but there are options now which allow developers to choose different technologies in certain areas. Using ADF in the model layer and working with relational data bases, create REST or SOAP services with ease is a big plus. For the UI there are use cases where JET will be used, but ADF has its share too.

Undo Reorder of Columns in af:table

A question on OTN about how to undo a reorder of columns in an af:table can be undone. In this blog I show how to undo such a reorder to show the columns of an af:table in their natural order.
The natural order is defined when you create the table. You can move the attributes in the create dialog or delete attributes you don’t want to see in the UI from the table.

In the image above we see the dialog after we drop a VO as table onto a page. To change is order of the columns in the table you can use the arrows on the right (in the red rectangle). Once you save the table you can reorder the columns in the property editor of the af:table.


The order of the columns you see in the dialog or the property editor is what is called default order of the columns. This default order can be different than the order of the attributes in the query the VO is based on.
The page we drop the af:table on is very simple. It is build from a quick layout and has a header for the page title and a panelCollection which holds the table.


We can reorder the columns in the UI by dragging a column and dropping it at a different location.

The question now is how to undo this manual reorder without refreshing the browser window.

To understand how this is implemented, we need to look how the the reorder is done in the first place. A table is build from one or more columns. Each of the columns describes the data to be shown in the column, the header to show and the display index which is the order of the columns shown in the UI. If the display index is less then zero (e.g. -1) the default order is used. Any other positive number is used to show the columns in ascending order of these display index.
To undo any reorder of the columns is an af:table we simply have to get to each column and set it’s display index to -1.

public class UndoColumnReorderBean {
    private static ADFLogger _logger = ADFLogger.createADFLogger(UndoColumnReorderBean.class);
    private RichTable table;

    public UndoColumnReorderBean() {

    public void undoColunmReorder(ActionEvent actionEvent) {
        _logger.info("Undo reorder...");
        // get the tables child components
        List<UIComponent> children = this.table.getChildren();
        for (UIComponent comp : children) {
            // check if the child is a column
            if (comp instanceof RichColumn) {
                RichColumn col = (RichColumn) comp;
                // if hte display index is greater 0 set it to -1
                if (col.getDisplayIndex() >= 0) {
                    _logger.info("...unset column "+col);
        _logger.info("... done!");

    public void setTable(RichTable table) {
        this.table = table;

    public RichTable getTable() {
        return table;

The bean above has a method undoColumnReorder which is an action event Listener triggered by clicking the ‘Undo Column Reorder’ button. This method uses the af:table component which is bound to the bean as property. It iterates over the child components of the table, checking if the child is a RichColumn (or af:column in the UI) and if yes sets its display index to -1;
To show the change in the UI, we have to ppr the table by adding the button as partial Trigger to the table


After clicking the button in the ui the table again looks like


so the default order of the columns is shown again.

You can download the application from GitHub BlogUndoColumnReorder. The sample is build using JDev but you can do the same with any other JDev version 11g or 12c you use. It uses the HR DB schema.

Reset Table Filter when Navigating to Page

This blog is a continuation of an older blog about how to reset the filters of an af:table component from a bean (How to reset a filter on an af:table the 12c way). In the older blog I described the technique to reset the filters defined in the FilterableQueryDescriptor of a filterable af:table.

Now users on OTN JDev & ADF space ask for a small variation of the use case. The filter should reset whenever a navigation takes place to the page which holds the af:table. No button should be clicked to reset the filter values.

As the original technique can still be used, I don’t go into detail about how to do this. It’s described in the other blog for JDev versions 12c. The same technique can be applied to 11g but different Java code has to be used (see How to reset a filter on an af:table). I changed the sample application, which you can download (see link at the end of the blog), so that the query panel with the af:table has an additional button to navigate to a different page.

Run through

After starting the application we see the page with an empty table as no search was done. Clicking hte search button will give us


The ‘Navigate’ button simply navigate to another view which holds twu buttons which let you navigate back to the original page.


The ‘back without clear filter’ just navigates back to the page, whereas the ‘back with clear filter’ navigates to a method in the task-flow which prepares the af:table for reset. This is the bounded task flow:


The EmpQueryPanel holds the af:query with the result table as shown in the first image. The view is marked as default activity in the task flow. When you first run the application (page RTFQPTest.jsf) the task flow is added as region to the page showing the query panel with the result table.

When you hit the search button on the page the table shows all employees. Now we can filter the results like ‘FirstName’ contain ‘s’ and ‘LastName’ contains ‘k’


Now if we hit the ‘Navigate’ button we go to the page shown in image 2 with the two buttons. If we click on hte ‘back without clear filter’ we come back to the page as shown above. The filter values are still present!

If we click on the ‘back with clear filter’ we see


so the filter values are cleared. So, how is it done?


In the original sample we had a button which we used to trigger a method which get the FilterableQueryDescriptor from the table. This descriptor holds the filter values which are cleared by looping over all ConjunctionCriterion which are the filter values. Here is the full method for 12c

 * method to reset filter attributes on an af:table
 * @param actionEvent event which triggers the method
 public void resetTableFilter12c(ActionEvent actionEvent) {
   FilterableQueryDescriptor queryDescriptor = (FilterableQueryDescriptor) getEmpTable().getFilterModel();
   if (queryDescriptor != null &amp;&amp; queryDescriptor.getFilterConjunctionCriterion() != null) {
     logger.info("Filter found...");
     ConjunctionCriterion cc = queryDescriptor.getFilterConjunctionCriterion();
     List&lt;Criterion&gt; lc = cc.getCriterionList();
     if (!lc.isEmpty()){
       logger.info("...iterating criterions...");
     for (Criterion c : lc) {
       if (c instanceof AttributeCriterion) {
         AttributeCriterion ac = (AttributeCriterion) c;
         Object object = ac.getValue();
         logger.info("...found " + ac.getAttribute().getName() + " value: " + object);
         if (object != null) {
getEmpTable().queueEvent(new QueryEvent(getEmpTable(), queryDescriptor));

public void setEmpTable(RichTable empTable) {
 this.empTable = empTable;

public RichTable getEmpTable() {
 return empTable;

A look into the log after clicking hte ‘back with clear flter’ shows


We see that the for loop caught all filters and resetted every filter to null.

The interesting part is how we triggered the call of the method resetTableFilter12c. As there is no button or other action event involved we use a trick. We add a method to the ‘ShortDesc’ property of the af:table which points to a bean method


Now, whenever the af:table is rendered it goes to the bean method asking for the test for hte short description. We use the call of this method as trigger to reset the filters. As this method is called multiple times during the JSF lifecycle, we need some kind of flag which tells us that the reset operation is done already. Otherwise we will spende lots of time calling the reset method without need.

public void setShortDescription(String shortDescritopn) {
logger.info("Set ShortDescription called");
this.shortDescription = shortDescritopn;

public String getShortDescription() {
logger.info("get ShortDescription called");
AdfFacesContext adfFacesCtx = AdfFacesContext.getCurrentInstance();

// get the PageFlowScope Params
Map<String, Object> scopePageFlowScopeVar = adfFacesCtx.getPageFlowScope();
Boolean reset = (Boolean) scopePageFlowScopeVar.getOrDefault("resetFilter", Boolean.FALSE);
boolean flip = reset.booleanValue();
if (flip) {
logger.info("ResetTable Filter!");
scopePageFlowScopeVar.put("resetFilter", Boolean.FALSE);
logger.info("Unset filter reset flag!");

return shortDescription;

As there are cases where the short description is ask for which we don’t want to use as triggers to clear the filters, we need another flag which we can check. For this we set a flag in the pageFlowScope of hte bounded task flow named ‘resetFilter’.  in the method we get the pageFlowScope and read the flag (lines 8-13). Only when the flag is set to true in the pageFlowScope we call theresetTableFilter12c method (line 14-19) and reset the flag to false.

The only thing left to do is to set the flag in the pageFlowScope when we liek the filters to get cleared when navigating to the page. For this we use the method action ‘resetTableFilter’ which is defined in the task flow. This method action points to a bean method


which puts the flag ‘resetFilter’ with a value of ‘Boolean.TRUE’ into the pageFlowScope:

public void setRestFlag() {
AdfFacesContext adfFacesCtx = AdfFacesContext.getCurrentInstance();
// get the PageFlowScope Params
Map<String, Object> scopePageFlowScopeVar = adfFacesCtx.getPageFlowScope();
scopePageFlowScopeVar.put("resetFilter", Boolean.TRUE);
logger.info("Set filter reset flag!");


You can download the sample application from GitHub:  BlogResetTableFilter12c

The sample uses JDev and the HR DB schema.