Problems running JDeveloper 12c

In the last couple of weeks, I get more and more reports of problems running JDeveloper 12.2.1.x. (to be exact, and

The problems reported are

  • properties editor not working
  • JDeveloper hangs during start
  • showing wireframe instead of page design
  • problems to configure the JDBC connection
  • problems compiling expression on attributes (not 100% verified that this is JDK problem)
  • problems migrating projects created with earlier versions of JDeveloper
  • problems with the groovy script engine
  • deadlocks within JDeveloper when editing multiple java files

to name some. The problems are not ADF related but IDE related. It turned out that they only could be reproduced if the used JDK to run JDeveloper on was newer than JDK 1.8.0_101. All problems are not reproducible when running JDeveloper with JDK 1.8.0_101.

Currently, there is a bug pending (Bug 26766333) with for some but not all mentioned issues. At the moment of writing this, there is no patch available.

My recommendation is to install JDK 1.8.0_101 and run JDeveloper using this JDK. You can do this by

  1. installing JDK 1.8.0_101 on your machine. The download for this old version is hard to find in the WWW. To make it easier, you can find it on this page: Java SE 8 Archive Downloads
  2. change the product.conf file you’ll find in your .jdeveloper folder inside your home folder. Open the file and set the SetJavaHome property
  3. to be on the safe side, you can recreate the integrated WLS to make it use JDK 1.8.0_101 too. If you know our way in the jungle of script files which are used to start the embedded WLS, you can change those files directly. As there typically are not many changes made on the integrated WLS, I find it easier to delete the integrated WLS and create it again. It’ll pick up the JDK JDeveloper is running on automatically and use it to run the WLS too.

You don’t need to update or change the JDK your standalone server is running on. To my knowledge, the problems are only IDE related, so they don’t affect the running application.

If you find a problem, which is related to using a JDK newer than 1.8.0_101, feel free to leave a comment on this post. I’ll add them to the list for reference.


JDev 12c: Debug Application Module Tester (BC4JTester) Problems

When you develop ADF Web Application you often use the ADF ApplicationModule Tester (BC4J Tester) to quickly test your business components data model and your self-written code in any EntityObject, ViewObject or ApplicationModule. For more information about how to do this look at JDeveloper & ADF: Use the Application Module Tester (BC4J Tester) to Test all your BusinessLogic.

Users who use one of the latest JDeveloper versions and newer may have noticed, that the BC4J Tester application starts without an error, but doesn’t show the dialog. I run into this a couple of times lately and decided to dig into this problem. On the Oracle Development Spaces, I saw some threads about this too.

The reason for this behavior is that any EO, VO or other methods in the application module have an error, which can’t be found during compile time.

Use case

To show the effect, we start with a simple Workspace and a model project which only has one ViewObject in the Application Module’s data model

We implement a small use case where we want to see the total salary of all rows retrieved by the query behind the VO. Without any added where clause we get the total salary of all employees. If we add a filter e.g. by DepartmentId=90 we only get the total salary of all employees of department 90. Here are some images of the final running model in the BC4J Tester


OK, so how do implement this use case?

We do this by adding a transient attribute to the EmployeesView and use a SQL default expression to do the calculation


In the image below we see the definition of the transient attribute in the ViewObject


This should do the trick. However, when we try to test this in the BC4J Tester we get

In the log window, but no dialog where we see the application module. We don’t get any hint about what went wrong. The tester is up and running, but we don’t see anything.

Shay Shmeltzer mentioned in one of the ODC threads, that the reason for this is that there is an error in the application module (ViewObject, EntityObject or AM method). As the only thing we added is the SQL statement for the transient attribute, it’s clear that the statement must have an error. It’s simply a missing ‘s’ character, as the DB table we use is named ‘Employees’ and not ‘Employee’. So the correct statement is


This will solve this problem and the BC4J Tester will start up and show (see the images above). But what if we added more things to multiple objects?

How to find the error then?

Older versions of JDev, the BC4J Tester did show an error message which showed the error and made solving the problem easy. Here is an image of the same application running using JDev


I did not manage to get the same output using JDev or newer, but you can get the same message in the message window.

For this, you need to start the BC4J Tester with the java option


The option is added in the model projects ‘Run/Debug’ option in the project’s properties

Whenever you start the BC4J Tester and don’t get any dialog, you can assume that there is an error in the application module. To find out what the problem is, add the java option to the model project and you get the detailed information in the log window.

Train Stop Status Handling

A question on the Oracle Developers Community was about how to handle a train stops visited status.

Use Case

The use case behind this was that a train can be used as a workflow visualization. A normal user starts the train, but at one point a manager has to approve something. This approval is one or more stops on the same train. If the manager picks up the workflow he should automatically start with the approval stop. There is no need for him to see the data accumulated in the stops before.

The use case has multiple challenges:

  1. Securing train stops for different user roles
  2. Allow starting the train from any stop
  3. Handling the state of the train stops

The first two challenges are handler by All Aboard, 97. How-to defer train-stop navigation for custom form validation or other developer interaction, and 82. How to programmatically navigate ADF trains.

The missing part is how to handle the train stops ‘visited’ state (see image above). If you start the train directly with ‘Stop 3’ you get this state


To implement this use case, we use a simple UI. It contains an input field, a button and the train which is added to the page as a region.

In the input field names label 1 you can enter the stop where the train should start. If no number is given, the train starts with the first stop. We use this input field to mimic the different starting stop for different users. This is the page when we start the application:

This is the page when we start the final application:

You can navigate between the train stops by using the ‘Back’ and ‘Next’ button, or by clicking the next stop in the train bar. As the stops are set to sequential, you can’t directly click on the 4th stop. You have to go through the stops 1 to 3 first.

Enter a number between 1 and 5 into the input field and tab out of the field will set the parameter for the train task flow and restart the task flow. The navigation is done via a router in the task flow. In the image below the stop number 3 is set as the starting stop for the train

And as you see the stops 1 and 2 are looking like they have visited before.


To show how to implement this we start with a simple bounded task flow which builds the train

The start builds a router which we use to navigate to the stop where we want to start the train. The starting stop is passed as parameter to the task flow

In the router, which is marked as default activity, the parameter is used to execute the navigation

The Magic

If you look at the train stop properties in the properties inspector you’ll notice, that there is no property for the visited state

This option is not available in the UI. Oracle has missed or deliberately missed to make this property accessible via the properties. If you dig into the implementation of the train task flow (see the articles provided at the begin of the blog), you’ll see how to access the train and its stops by code:

ViewPortContext currentViewPortCtx = controllerContext.getCurrentViewPort();
TaskFlowContext taskFlowCtx = currentViewPortCtx.getTaskFlowContext();
TaskFlowTrainModel taskFlowTrainModel = taskFlowCtx.getTaskFlowTrainModel();
// get the stop from the map
TaskFlowTrainStopModel currentStop = taskFlowTrainModel.getCurrentStop();

The TaskFlowTrainStopModel doesn’t provide any access to the visited state. If you look at the class definition you’ll notice, that it’s only an interface

which doesn’t provide access to the visited property. Setting a breakpoint in the debugger we can inspect an instance of this interface

and we get the class implementing the interface as:


This class has the visited property we are looking for.


Now we can implement a method which we call before a train stop gets rendered and which sets the visited property of all previous stops to true.



However, it’s the class we need to get to the property. You have to understand, that the usage of the class has its risks, but that it’s not forbidden. The risk is that Oracle can change or delete the class without notifying you beforehand. So, in later versions, your code might break.

The method checks the task flow parameter if it’s null to set to a number less or equal to 0. In this case, the method returns an empty string. We do this check to avoid that the method does it’s work every time we navigate the train. It should be done only once when the train starts.

If the check finds a positive number, it sets the task flow parameter to zero (line 37).

It then gets the task flow information from the Context (lines 39-43). In line 50 we acquire the current stop before we loop over all previous stops and set their visited property to true (lines 53-59).

The missing part is how to call this method when a train stop is rendered. For this, we use a technique called Lazy Initalizing Beans. The trick is to use a hidden af:outputText and set e.g. the value property of the component to a bean property.

When the page or fragment is rendered, the method getInitStatus() in the bean is called. This is exactly the method shown above. We add this hidden af:outputText to each train stop before the af:train component.


You can download the sample from GitHub BlogTrainStopStatus. The sample is build using JDev and doesn’t need a DB connection. You can use the same technique in other JDeveloper versions.

Query and Filter an af:listView

Most of the time we use tables to show tabular data to users. However, JDev and ADF allow for other components like the af:listView to be used to show such data to the user in a more modern way.

The image above shows the normal display of data when an af:query is used together with a table to show the result.

A more fancy, modern look we get if we use a af:listView to show the results as this allows us to style the data

Use case 1

We like to use an af:query to search for employees and show the result in a styled af:listView.

Implementation 1

This is pretty easy as we only have to use an af:listView as the result component of the af:query

And to exchange the af:table with an af:listView. Or you build the page by first dropping an af:query onto the page (without table) and then add the af:listView

Then you get the wizard to layout the list

This will give you a basic layout which can be styles in JDev as

The final result is

which looks more modern. One thing the af:table give you out of the box is the second use case.

User Case 2

We like the af:listView to be able to be filter the result like the af:table can.

Implementation of second use case

Easy you think? Well, the af:listView component doesn’t provide any filter out of the box. There isn’t even a filterModel like there is for an af:table.

So, how do we get this implemented. The idea is to use a af:table component but only use the filter provided by the af:table. The remaining parts like table data, possible scroll bars and status bar or scrollbars we remove.

We start by dragging the EmployeesView1 from the data control onto the page again.

And drop it after the closing af:panelHeader and before the af:listView as ‘ADF Table’

In the image you see that I have removed some available columns. Before we go to hide the part of the table we don’t need, we make the table work together with the af:query and the af:listView. When we use the af:query the table shows the right detail (auto PPR triggers the refresh of the table). However, if you have queried for the ‘Purchasing’ department and then enter an ‘s’ into the ‘First Name’ filter field of the table and hit enter, you get

As you see, the table shows the right result (2 rows) but the listView still shows all employees of the Purchasing department.

To make it work, we need to add a partialTrigger to the listView which points to the table. This way each time the table changes the listView will too.

Save all changes and refresh the page. Now if you enter a value into a filter field and hit enter, the listView will update too.

After the page works we have to get rid of the data below the header of the table. This is easy to accomplish by styling the table. We only need the filter field and the header below the filter fields so that we know which field filters which data. Simply set the maxHeight of the table to the exact height of the the two components. You can use your browser’s developer tools (F12) to measure the height. In my sample it’s 65px. So, setting the tables inlinestyle to

max-height: 65px;

will hide everything below the filter and the header

If you like you can create a skin and create a style class and use this style class instead of setting the max-height directly to the inlineStyle of the table. A nice addon is that the table header sorting is working too for the listView.

You can download the sample from gitHub BlogFilterListView. The sample is build using JDev and uses the HR DB schema. The principle can be used in other JDev versions too.