​Since the introduction of the new and improved Popup LOV in APEX 19.2, I have been working with our clients to migrate them from various Plugin LOV solutions. During this effort I encountered an issue with setting a default value for a Popup LOV column in an Interactive Grid. Specifically, the return value of the LOV was being set by the defaulting code but the display value was not refreshed to show the newly defaulted value. 

​At one of our clients, we are building an Order Capture application in APEX. This application allows users to enter orders through APEX and then invoices those orders in Oracle e-Business Suite (EBS). Using APEX, we were able to build a custom fit User Interface which significantly increases the number of orders a user can enter in a day. We were also able to leverage standard EBS views and APIs to interact directly with other EBS data (customers, items, purchase orders etc.). This is just one of the many scenarios where APEX can be used to extend standard EBS functionality.
 
In the order capture APEX page, the user selects a Part Number and the system should default a 'Finish' for the selected Part. Nine times out of ten the default finish is correct, so applying the default saves a significant amount of time for the user. Using the new Popup LOV for the Part and Finish fields also means the user can enter an entire order line without having to use their mouse (which they would have to if we used Select Lists).
 
Here is a simplified view of the Interactive Grid where the user selects a Part and Finish.

The problem is when the user selected a Part, the code to set the default Finish was setting the hidden return value OK but the display value was not getting refreshed/changed. What was even more strange was that in about 1 in 10 cases it would refresh the display value.
 
This is how the two columns on the Interactive Grid were configured before we implemented the work around (described later in the post). This screen shot shows you the overall structure of the grid, with the two columns for Part (CUST_ITEM_ID) and Finish (FINISH_CODE).

​The Part field (CUST_ITEM_ID) is setup as a Popup LOV. it has a Dynamic Action which calls PL/SQL code to fetch the default Finish (if there is one). Here is the definition of the Dynamic Action.

The default finish was being set in column FINISH_CODE via the ‘Items to Return’ setting seen in the screen shot below.

​The Finish field is setup as a Cascading LOV, so that when the Parent LOV (CUST_ITEM_ID) is changed, the Finish LOV is also refreshed.

After several attempts at implementing a workaround myself (and lots of Googling), I reached out to the APEX development team. Before doing this, however, I put together a simplified example demonstrating the issue on apex.oracle.com. I can’t stress the importance of this step enough. The APEX development team's time is valuable, I wanted them to focus on finding a solution to the issue, not trying to re-create the issue.
 
At around 10:30pm his time, Joel Kallman received my email and passed it on to Anthony Rayner (based in the UK). When I woke up the next morning, I had an email from Anthony with my sample App updated with a workaround (described below). By 10am the same day, the workaround was implemented in the client's application. If that is not an advertisement for a highly competent, community led development team that cares about their customers then I do not know what is.

So, what did Anthony come up with? It ended up being just a couple of minor changes to what I described above. Firstly, create two new page items P12_FINISH_CODE and P12_FINISH_NAME. These will temporarily store the derived default Finish (return and display value) when the Part field is changed in the grid. These should be hidden fields with ‘Value Protected’ set off.

​Change the ‘Set Default Finish’ Dynamic Action PL/SQL code to set the two new page items with the default Finish return and display values. Make sure to include these in the ‘Items to Return’ setting.

​The next change is reliant on the FINISH_CODE column being assigned a ‘Static ID’ so that it can be referenced more easily from JavaScript.

Finally, add a step to the Dynamic Action with an ‘Action’ of ‘Execute JavaScript Code’. The JavaScript code below sets the display and return value of the FINISH_CODE column in the Interactive Grid (identified using the static ID ‘finishCode’) to the values in the two page items (P12_FINISH_CODE and P12_FINISH_NAME).

​After doing the above, both the display and return values for Finish are set every time the Part number field is changed.

​First of all, I hope this issue gets addressed in the standard APEX product soon. Workarounds like this are great but the additional page items and JavaScript complicate and bloat the page un-necessarily. Apart from that, it also looks hokey.
 
Having said that, I should say that the APEX development team never cease to impress me with their commitment to the customer. I worked for Oracle for nearly 20 years and never encountered this kind of willingness to engage directly with customers from a product development team. Bravo!

​Since APEX 18.1, there has been a new way to call OAuth2 Client Credentials secured web services using APEX_WEB_SERVICE. This new method alleviates the need for developers to store OAuth2 credentials, manage token expiration and also has potential performance benefits. In this post I will describe the new approach and its benefits.

In the OAuth2 client credentials flow you must first call a token web service, passing a client ID and client secret. If these credentials are valid, the token service will then pass back a token. The token can then be used to call certain secured web services covered by the token. The token returned by an OAuth2 client credentials service looks something like the below.

I must confess that I can get stuck in my ways. It often takes a weekend and a few hours buried in the APEX API guide for me to get caught up with the latest and greatest. Case in point, I have been using APEX_WEB_SERVICE for several years and most of my code to call web services secured by Ouath2 Client Credentials looks like this:

1. Fetch token service URL, Client ID and Client Secret from a custom table.
2. Call the OAuth token service to get token using APEX_WEB_SERVICE.
3. Optionally store the token for use again (within its expiration window).
4. Call the main web service using APEX_WEB_SERVICE, passing the token.
5. Repeat as many times as the user calls the web service.

 
Don’t get me wrong, this is still much more convenient than using UTL_HTTP. As a developer, however, I am still having to manage the following:

1. Storing the token end point and the client credentials (most likely not very securely).
   
2. Deciding when the token expires and fetching another token (more likely not bothering to do this and just getting a new token every time I call the main web service).

Of course, there are drawback to this.
 
Storing the Credentials
​I have to build an APEX page to store the token URL and credentials and also some APIs to set and get these values. I then have to build APIs to manage token expiration and decide when to get a new token. Finally, I have to maintain this code and nurse it through upgrades etc. All this was fun the first time but I would much rather have APEX handle this for me so I can focus on delivering business logic.
 
Securing the Credentials
This approach also introduces the obvious security concern which is that I may not be encrypting the credentials when I store them.
 
Performance
Unless you build code to store the expiration time for the token, you are most likely going to call the token service every time you call the main web service. While token services are usually pretty light weight, you are incurring an extra round trip from your database to the endpoint every time you call the main web service. This can add up to a second to every web service call which is noticeable to your end users.

Ever since the APEX development team introduced APEX 18.1 you could sense a shift toward making it easier for developers to talk to the outside world using APEX. This all hinges around the APEX_EXEC package and the drive toward making every APEX component equally as happy sourcing it’s data from a web service as it is from a SQL statement on the local database.
 
Since APEX 18.1 we can make calls to APEX_WEB_SERVICE without having to deal with first calling the associated token service or having to manage token expiration. This is made possible with the addition of 2 parameters p_credential_static_id and p_token_url. These parameters allow you to pass a reference to a set of ‘Web Credentials’ and the token URL to APEX_WEB_SERVICE and it will handle the REST (pun intended).
 
APEX_EXEC For all use cases that fit, I strongly encourage you to look into APEX_EXEC, instead of APEX_WEB_SERVICE. APEX_EXEC takes even more of the load off the developer and has a number of other advantages which I don’t have time to get into in this post. Carsten Czarski wrote a great overview in this post.

​As always, examples speak a thousand words. Before we can reference a set of Web Credentials in our code, we need to define them in APEX. Web Credentials are stored at the Workspace level. Once logged into your APEX Workspace, navigate to App Builder > Workspace Utilities > All Workspace Utilities > Web Credentials.
 
In addition to creating and editing Web Credentials in the APEX UI, you can also set them programmatically using the APEX_CREDENTIAL API (available since APEX 18.2). This API has a number of uses, one of which is to help when migrating new credentials to TEST and PROD instances.

​Note: When you edit Web Credentials APEX does not show you the client secret. It is there, you just can’t see them (for security reasons).

I now want to look back on the three drawbacks from the old five steps approach and see how the two extra parameters above have helped me. The main benefit of course is that we now have just two steps instead of the 5 steps in the original approach:

1. Call the main web service using APEX_WEB_SERVICE, passing Web Credentials Static ID and Token Web Service URL.
2. Repeat as many times as the user calls the web service (APEX handles fetching new tokens etc.).

 
Storing the Credentials
The OAuth2 Client Credentials can now be stored natively in APEX using ‘Web Credentials’. APEX provides a UI to manage these credentials and APIs to set these credentials programmatically.
 
Securing the Credentials
Using Web Credentials, even the Workspace Administrator is not able to view the client secret for a given client. In addition, although there are APIs to set the credentials there are none to get credentials. The only way to use them is via APEX APIs.
 
Performance
​When you pass values for p_credential_static_id and p_token_url you are essentially handing over management of the OAuth2 token to APEX. It will store the expiration for the token, and it will only get a new token when it needs to.
 
All this leaves you as a developer to focus on calling the main web service and parsing the results.

​The APEX development team continues to make it easier for APEX developers to work with REST based web services. This makes it easier and easier for developers to use cloud-based services like SMS, Address Verification, Machine Learning, Object Storage etc. etc. Having easy access to these cloud services allows developers to provide game changing features in their APEX apps.
 
Enhancement Request: It would be great if instead of having to pass p_token_url, we could pass the static id of the remote server plus a suffix for the token service end point. This would prevent us from having to store the URL of the token service in a custom table. Remote Servers are defined under ‘Remote Servers’ in the Workspace Utilities area (right above Web Credentials).

​As you may be aware, Exadata Express is coming to the end of its life. If you like us have found this platform both useful and affordable, then you will be wanting to know what’s next? Where in the cloud can I deploy my RAD (REST, APEX, Database) stack now? In this post I will explore an alternative which we have implemented here at JMJ Cloud.

​We have been using Exadata Express and JMJ Cloud for more than 2 years. During that time, we helped five of our clients implement Exadata Express. Use cases ranged from custom cloud point solutions to ERP Cloud PaaS extensions. I have also blogged about this platform on a number of occasions. Exadata Express offered a highly functional PaaS solution for as little as $175 a month. We (and our customers) were very happy with it. In fact, I thought this product was a sign that Oracle had finally turned the corner to becoming a major cloud player. It provided a very low barrier to entry with more feature rich alternatives ready and waiting when your business grew into them. This is essentially how AWS became so successful. One day you start with half a dozen EC2 servers for a couple of hundred dollars a month and a few years later you are spending a million dollars a month (think Netflix). Oracle on the other hand typically starts by charging a lot and then work their way up from there.

A month or so ago, our Salesperson emailed us… Here is where I need to pause for a small rant. Why on earth, if you are selling automated cloud solutions should I ever have to deal with a Salesperson? I want a server; I know what size server I want, and I know how much I want to pay for it. From my perspective, having a Salesperson in the middle of this is expensive from Oracle’s point of view and extremely frustrating from my point of view. OK, back to the story… Our Salesperson reached out and said we could not renew our Exadata Express contract and oh, by the way, you have a month to get off the platform. After a little discussion (i.e. an hour of my time on the phone) we converted that 1 month into 3 months. We bought ourselves some runway, but we still needed another platform. Enter the Salesperson (again). After another one hour of my time he proceeded to explain the wonders of the Autonomous Database Cloud Service. Don’t get me wrong, this service shows a lot of promise and it is going to be great for our larger customers. At more than $1,200 per month, however, it is not going to fly for us or our smaller customers wanting to run the RAD stack in the cloud. It is nearly 7 times more expensive than Exadata Express. If you wanted to use Autonomous Database with APEX for a Cloud ERP extension and you need a DEV, TEST and PROD instance that is $3,600+ per month instead of $600 / month.
 
To add insult to injury, there is not even a button push migration path from Exadata Express to Autonomous Cloud. This means that even if you have the money to pay for Autonomous Cloud, you are going to have to fork more money to migrate.

I would love nothing more than to say I whole heartedly recommend folks move to the Autonomous Database Cloud Service but that is not where I am going. The answer to our problem was in a service we have dabbled with before which is Amazon’s RDS (relational database service) service along with an EC2 (elastic compute) server running ORDS. I have never really been completely happy recommending this approach for the RAD (REST, APEX, Database) stack before, two things have changed to make this a much more appealing approach.

• You can now perform a full ORDS install on an RDS instance. This is thanks to new functionality in ORDS 19.1 which allows you to run the install as a user other than SYS. This is especially important now that REST services are no longer available in APEX alone.
• AWS have started making the latest version of ORDS available on RDS much sooner than they used to. For example, as of right now, September 2019, you can run ORDS 19.2 and APEX 19.1 on RDS, whereas Exadata Express is running ORDS 3 (ouch) and APEX 18.2.

​The diagram below shows the architecture we are using at JMJ cloud.

Here is a cost breakdown for the above architecture.

There is no perfect solution and there are certainly some things you should consider when moving to AWS.

• Although the database is almost fully managed by AWS, you are on your own with regard to the ORDS server. You need to standup an EC2 server and install ORDS, Tomcat, setup your load balancer etc. This means it is not as hands off as Autonomous Database Cloud Service.
• At time of writing, there is no way to install APEX PSE Bundle Patches to an RDS instance. Depending on the bug, this could be a big deal.
• For license included versions of the Oracle Database, Standard Edition 2 is way more affordable than Enterprise Edition. If you have specific use cases that require Enterprise Edition features, then keep an eye on your costs.
• Although AWS takes care of patching the DB server, you will be responsible for patching the EC2 server and upgrading Java, Tomcat and ORDS.

​I am not saying RDS is a magic bullet that allows you to achieve the hands-off capabilities of the Autonomous Database Cloud Service at a fraction of the cost. If you have the budget and you want absolutely nothing to do with the underlying architecture, then Autonomous Database Cloud Service is for you. If you want lower level access to the database (and / or ORDS) but want to be absolutely sure the database itself is being backed up and patched for you then it offers a great alternative. If you throw in easy access to other AWS services, this soon becomes a very compelling solution.
 
After all, for developers, the health and welfare of the database is really the scariest part of running your own RAD stack. If AWS is essentially taking care of this for you, the rest is reasonably straight forward.