The design for the OpenNTF home page has been with us for a little while now and has served us pretty well. It looks good and covers the bases it needs to. However, it's getting a little long in the tooth and, more importantly, doesn't cover some capabilities that we're thinking of adding.

While we could potentially expand the current one, this provides a good opportunity for a clean start. I had actually started taking a swing at this a year and a half ago, taking the tack that I'd make a webapp and deploy it using the Domino Open Liberty Runtime. While that approach would put all technologies on the table, it'd certainly be weirder to future maintainers than an app inside an NSF (at least for now).

So I decided in the past few weeks to pick the project back up and move it into an NSF via the XPages Jakarta EE Support project. I can't say for sure whether I'll actually complete the project, but it'll regardless be a good exercise and has proven to be an excellent way to find needed features to implement.

I figure it'll also be useful to keep something of a travelogue here as I go, making posts periodically about what I've implemented recently.

The UI Toolkit

The original form of this project used MVC and JSP for the UI layer. Now that I was working in an NSF, I could readily use XPages, but for now I've decided to stick with the MVC approach. While it will make me have to solve some problems I wouldn't necessarily have to solve otherwise (like file uploads), it remains an extremely-pleasant way to write applications. I am also not constrained to this: since the vast majority of the logic is in Java beans and controller classes, switching the UI front-end would not be onerous. Also, I could theoretically mix JSP, JSF, XPages, and static HTML together in the app if I end up so inclined.

In the original app (as in this blog), I made use of WebJars to bring in JavaScript dependencies, namely Hotwire Turbo to speed up in-site navigation and use Turbo Frames. Since the NSF app in Designer doesn't have the Maven dependency mechanism the original app did, I just ended up copying the contents of the JAR into WebContent. That gave me a new itch to scratch, though: I'd love to be able to have META-INF/resources files in classpath JARs picked up by the runtime and made available, lowering the number of design elements present in the NSF.

The Data Backend

The primary benefit of this project so far has been forcing me to flesh out the Jakarta NoSQL driver in the JEE support project. I had kind of known hypothetically what features would be useful, but the best way to do this kind of thing is often to work with the tool until you hit a specific problem, and then solve that. So far, it's forced me to:

• Implement the view support in my previous post
• Add attachment support for documents, since we'll need to upload and download project releases
• Improve handling of rich text and MIME, though this also has more room to grow
• Switched the returned Streams from the driver to be lazy loading, meaning that not all documents/entries have to be read if the calling code stops reading the results partway through
• Added the ability to use custom property types with readers/writers defined in the NSF

Together, these improvements have let me have almost no lotus.domino code in the app. The only parts left are a bean for formatting Notes-style names (which I may want to make a framework service anyway) and a bean for providing access to the various associated databases used by the app. Not too shabby! The app is still tied to Domino by way of using the Domino-specific extensions to JNoSQL, but the programming model is significantly better and the amount of app code was reduced dramatically.

Next Steps

There's a bunch of work to be done. The bulk of it is just implementing things that the current XPages app does: actually uploading projects, all the stuff like discussion lists, and so forth. I'll also want to move the server-side component of the small "IP Tools" suite I use for IP management stuff in here. Currently, that's implemented as Wink-based JAX-RS resources inside an OSGi bundle, but it'll make sense to move it here to keep things consolidated and to make use of the much-better platform capabilities.

As I mentioned above, I can't guarantee that I'll actually finish this project - it's all side work, after all - but it's been useful so far, and it's a further demonstration of how thoroughly pleasant the programming model of the JEE support project is.