Category Archives: Architektur

Maven and the magic of profiles

Maven and the magic of profiles

POI4XPages has a module called poi.assembly. This module builds the final ZIP file which can be uploaded to the OpenNTF website. One step in the assembly is that the module downloads the source code direct from github. This step takes a lot of time, so each mvn clean build takes a lot of time, because the assembly line is also executed. Fortunately Maven has a concept called profiles. Maven let you define different profiles. In the POI4XPages case, I’ve built two different profiles, which contains different set of modules:

200Each profile has its own id tag. The default_no_assembly profile is activated by default. The <activation> markup let you activate also profile by defining environment variables.

I’ve now to different run configuration in my eclipse client. To run the default profile, I use the following:


But if I need the assembly profile, I use the following configuration:


Merry christmas!

Leave a comment

Posted by on December 24, 2014 in Architektur, Java, Maven, OpenNTF, POI4XPages, XPages



Maven and XPages Plugin – From Scratch / Part III – The Testsuite

Part I has covered the building of the parent pom.xml, Part II was about the plugin and now in Part III we talk about the test suite. It’s fair to say that this part seems not to be very interesting, but in this Part we talk also about the why.

Why do I make the org.openntf.base.jaxb.dxl plugin?

The core purpose of the jaxb.dxl plugin is to convert a DXL export of a document, or of some design elements into an object based representation. Imagine this: You have a document with 2 rich text fields, one is a traditional rich text field and the other is mime based. For some reason (like transforming the content of the rich text field to a POI document or PDF document) there is an interest in the structure of this rich text field. A simple export as DXL and conversion to this java objects which are cascaded helps a lot.

Or you have this database, and you want to scan the content of all design element for a specific @Formula….. Export the database to DXL and convert it with jaxb to java object tree. And the scan the tree and analyse.

But why not start with the implementation, that sound better than testing!

Agree. Implementing is always better than writing test for the sake of writing tests. But after reading the book “Test Driven Developement” by Kent Beck, my understanding of test has changed. Tests explains what I want to do. Read the following test:


This test covers my user story “Reading a dxl-document.xml and convert it to a java object structure”. To make this test work, I’ve to writte a lot of implementation stuff, so stay tuned.

The next Maven module -> org.openntf.base.jaxb.dxl.testsuite

Again with the eclipse client:


select as packaging type “eclipse-test-plugin”:

152And we have a new project.

And now repeat the following steps, like you have done for the plugin in Part II

  1. New file called MANIFEST.MF into the folder META-INF
  2. New file called

Open the MANIFEST.MF File and go to the MANIFEST.MF Tab and enter the following stuff:

153The directive “Fragment-Host” makes this plugin to a fragment of org.openntf.base.jaxb.dxl. This is the best behavior to separate test and implementation in a plugin.

Open the File and go to the tab. Enter the following stuff:

154Now is a good moment to do a right-click on the project go to Plugin-Tools and execute “Update Classpath”

Open the MANIFEST.MF again and add the following dependency. This is needed that we can start writing our test class.


Let’s write the first Test:


We have also imported the dxl-db-test.xml and the dxl-document-test.xml. This documents were produced by a nice simple DXL Export.

To make the test compile (STEP 2 in TDD), we have to import the XSD File for domino_9_0_1 and convert it to classes. We use for this the JAXB import

JAXB Import of the domino xsd


Start with menu “New” and select the JAXB Classes from Schema Wizard


Select the target project for the import (yes our plugin)


If you do not have the domino_9_0_1.xsd in a project in your workspace, you can import it from your Notes/Data directory


Please specify carefully the package, where the classes should be produced. Existing classes will be overwritten (note to myself, move the DXLActivator Class to an other package or a level up)164

Select finish. The import will now generate a lot of classes.

You can now resolve the import of the Document object (yes its a …jaxb.dxl.Document object)


Time to finish the compilation problems in the test class.

Create the ConvertorFactor in the plugin (see the package that I select? It’s outside of the dxl package, because the content of this package was generated before).


We build also a dummy implementation of “convert2DocumentFromStream”. 167

Only one step, and we can test (with executing our runtime configuration)

We have to include the *.xml Files in our binary build. We open for this step the MANIFEST.MF and add the 2 files


Lets build the project by executing org.openntf.base clean install169

Oh yes the build fails! This is total ok at this point, because our test fails, and a failing test should stop our build!

Now its time to have some fun and finalize the plugin… The next Part will cover how to build the feature and the update site

If you wanna follow the project, please watch this repository on

Leave a comment

Posted by on December 23, 2014 in Architektur, Java, Maven, XPages


Tags: ,

A word about Maven

When I try to explain what Maven is, then I’m using often the following picture:

Maven is a factory building with a brilliant factory manager

Imagine this: There is this empty factory building and you walk with your construction plan to the office of the factory manager. You enter the room and the factory manager takes your construction plan. He read it and then he runs to tool storage. In a short time he has all the tools to build your construction taken from the storage and build a production line.

According to your plan, he has also hired the Q & A people, which let first run some unit tests directly in the production line and also some integration tests. At the end of the whole production line is also an assembly line to build the kit that is ready to ship.

… at this point you say. “mvn clean install” and the factory manager builds the software.

The other day, you come back to the factory building with a new construction plan. And the factory manager “cleans” the building and setup a new production line for you. If he finds some tools not in the storage, he orders the tools directly from *Maven Central*. But with the same passion and quality as the day before the manager tries to build the software.

What Maven can do for you depends on your construction plan and the tools that are available

Back from the analogy:

  • factory building -> maven
  • factory manager -> mvn
  • construction plan -> pom.xml
  • tool storage -> maven repository
  • tools -> maven plugins

Your construction plan contains all instruction about what builder should be used, how unit testing should work, and also how the assembly should run.

And it is not some kind of magic to build your own builder. I’ve built the headless designer integration, also a builder to build complete XPages Applications. It’s only hard work, nothing more and nothing less (and also a lot of fun 🙂 )

Leave a comment

Posted by on December 23, 2014 in Architektur, Maven



Ready for todays webinar about OpenNTF Essentials?

I hope that you have already dowloaded OpenNTF Essentials and at least installed on your development computer. If not, it’s a 5 minute job as you can see in this youtube video (starting 12:00).

In Todays webinar, Mark will cover some cool stuff about debugging and Nathan takes a deep dive into the OpenNTF Domino API. My part will be about some cool @nnotations. Imagine this:

Your are writing a call center application and you have designed the “Call” object as a java object according the java bean specification. With the ObjectData Source from the ExtLib are you know ready to edit each field of this java object, but now you have to save….. and read …. and delete ….. the object from the database.

I will show you in todays webinar how you can do this with writing only “3 or 4” lines of code 🙂

I hope you will join us, today 10:30 EST, register now!

Leave a comment

Posted by on December 12, 2013 in Architektur, Design Pattern, Domino, Java, OpenNTF, XPT


The “ninja-style” programming model by WebGate

First I’ve to excuse that we are so selfish to call our programming model ninja-style. It was happened based on the fact, that we have programmed ninjas in internal programming course. But the term ninja-style was established and if you gave something a name, it’s very hard to change it (Maybe you have seen Monster.Inc by Pixar, then you know what I mean).

Our intention is to make programming for XPages as much fun as possible.  And fun means in the case of programming: Having success, with less of effort and stress.

But the ninja-style has also to do with leaving the comfort zone and go out from this protected workshop in which the most of the domino developers where living. XPages requires a complete new set of skills. On the front end part are you faced with: HTML5, JavaScript, CSS, DOJO (and that’s only the beginning) and of course SSJS for the binding part. On the backend is Java the most required skills, that you should have. But you have also to learn about data sources, controls, data binding and something called JSF life-cycle.

Definitely a lot of topics to cover and the most of us are not geniuses, specially universal genius. It’s a matter of fact that it is easier to split and separate the topics. Because of this we have introduced the N-Tier architecture to our programming model. We have separated the front end stuff, from the back-end stuff.

The front end covers all the presentation (Presentation Layer)

Typically our front end developers are brilliant in arranging all the controls in the XPages and they are mostly virtuous in JavaScript, CSS and Dojo. They know about the “beans” as far as they have to know how to use them as API to the model and the business logic. They don’t care, how object are loaded and stored, they don’t care about how processes and logic are executed, as long all is working correct.

The backend covers all the model and business logic (Business logic Layer)

The back-end guys do all the brilliant stuff with the model of the data and all the processes. Mostly they are Java Cracks or on the way to it. They care about good backend performance and have read the book “design pattern” by the gang of for. No interface and connection to any backend is to complex for them, but do not let them do any HTML stuff.

But where are the storage guys?

At this point, some of the core features of the XPT cames to action. The DSS of the XPT solves all the storage stuff.

But keep this: Having success with XPages has to be about simplification. Our first step was to separate front end from back-end.

To be continued >>>



Tags: , ,

<xpt:….> – The XPages Tookit – Beta1 released

We have released an early beta of the XPages Toolkit, called beta1 on OpenNTF. A short time before I was going to the hospital for a heart revision task :). But the XPages Toolkit need some explanations about the why and what.


The XPages Toolkit (in short XPT) is a extendable Extension to the XPages Runtime and the Extension Library. The XPT contains a strong core for a programming model that we have introduced at WebGate as the ninja-style. I will blog later this week about this programming model and I will also make some snippets to this available.

But what contains the XPT:

  • The core (for the ninja-style / hope this name is not protected)
  • RSS Datasource and UIControl -> a ready to use UI-Element so display RSS Feeds
  • Read and write Access to properties files
  • Some missing OneUI Elements (like the aboutbox, the welcomebox and the tipbox)
  • And the XPages Agents with ready to use UI Elements and a replacment for the agents manager

And it is extendable. We have designed the XPT to make it easy to plugin other solutions. I will cover this topic also in a separate blog post. But in short, I think there are some many good things available under apache licence. Or think about the charting in dojox, its one of the things that I miss as a good UI Control. I know that there is a great OpenNTF project available from Julian Buss (YouAtNotes xCharting), but why not integrate the code into a plugin?

All the best

PS: Heart revision task was successfully completed and I have a good recovery time.

1 Comment

Posted by on September 23, 2013 in Architektur, OpenNTF, XPages, XPT


Tags: , ,

Accessing Beans from Java Code in XPages (learned by Reading the SourceCode of the ExtensionLibrary)

Hello Java and XPages Fans

During our work for the myWebGate 2.0 OpenNTF Edition i was faced with a very simple Problem. I had to access a public method in one of our managed beans. This managed beans are designed to act as session facade. So the bean name is directory bean and the scope is session. The ClassName is based on our internal guidlines DirectorySessionFasade.

In this class, is a method called getMySelf() witch gives me some information about my current user in context of the myWebGate Framework. The powerfull thing about managed beans is, that the object exists only one time baed on the scope, in this case one time each session.

For our activtystream implementation I have to acces this getMySelf() methode, because all my relations are stored in the returning object. A short google search doesnt help, but I rember that I’ve ready a pattern in the ExtensionLibrary, that maby can help. An voilaz there it was, written by Phillip Riand, I found some thing that I changed to this for my situation:

public static final String BEAN_NAME = “directoryBean”; //$NON-NLS-1$
public static DirectorySessionFacade get(FacesContext context) {
        DirectorySessionFacade bean = (DirectorySessionFacade)context.getApplication().getVariableResolver().resolveVariable(context, BEAN_NAME);
        return bean;
    public static DirectorySessionFacade get() {
        return get(FacesContext.getCurrentInstance());

This simple codesnipped gives you the capability to execute DirectorySessionFacade.get().getMySelf(). The DirectorySessionFacade.get() Code calls the get( FacesContext context() and there the inmemory Variable directoryBean can be accessed.

What a simple construct to access the same instance in java as in the SSJS and XPAGES. Thanks to IBM for sharing the ExtensionLibrary as OpenSource.


myWebGate Building Process – Dojo und Templates

Das Benutzerverzeichnis ist ein Kernteil des myWebGate Framworks. Um dem Benutzer ein besonderes Erlebnis bei der Bedienung zu bieten, ist das Benutzerverzeichnis als “OnePage” Applikation designet. D.h. wir präsentieren alle Inhalte dynamisch. Die Daten werden via REST API direkt vom Server gelesen und dann on the Fly gerendert.

JSON 2 HTML Renderer gibts wie Sand am Meer. Unser Ehrgeiz war aber, dass wir das Problem mittels DOJO lösen (naja auch einige Lizenzfragen sind dann einfach geklärt). Leider ist es mit DOJO so, wie mit manchem genialen Projekt. Irgendwie hat der Dokumentator frei gehabt. Erschwerend kommt dazu, dass im Gebrauch mit XPages noch ein paar Dinge zu beachten sind.

Also zuerst das Kleingedruckte (oder wie krieg ich so ein Design hin, wie es Rechts zu sehen ist).

  1. Wir haben das HTML Template als PAGE erfasst und dort den Contentype auf HTML gesetzt.
  2. Das HTML Template darf nur 1 Root Element haben, sonst hat DOJO ein Problem
  3. Das neue DOJO Modul muss mittels
    <xp:dojoModulePath url=”biz.webgate.mywebgate.person”
    eingebunden werden.

Tja und dann ist alles ziemlich simpel. Wir erstellen eine Javascript Client -Library, die wir biz.webgate.mywebgate.person.js nenne und schreiben dort folgenden Code rein:

Mittels dojo.provide definieren wir, was dieses File liefert, dojo.require definiert was wir alles brauchen, hier ist dojox.dtl._Templated die Rendering Enginge, die auf django basiert. Im dojo.declare geschieht die Magie. Mittel der cache Anweisung lesen wir das HTML File ein und peron:null erweitert unser Widget um ein neues Objekt.

In der Funktion openPersonTab() fügen wir dann die Person ein. Dabei greifen wir auf ein Spezielles TAB Objekt zu, dass uns das dynamische generieren der Tabs erlaubt. Die Variable xhrArgs definiert den Zugriff auf unser REST Service und lädt von dort das Personen Objekt. Der effektive Aufruf geschicht dann mit dojo.xhrGet(…), aber im Objekt haben wir auch die Verarbeitung des Resultats definiert. load wird dann aufgerufen, wenn wir erfolgreich die Daten abgegriffen haben. Innerhalb dieser Funktion initalisieren wir ein neues Widget und geben die data (das ist das Person Objekt vom Rest Service) direkt an das Widget weiter. dojox.dtl._Template redert dann das Template. Und durch placeAt() hängen wir das Renderresulat an das richtige Ort.

Blicken wir noch in das HTML Template:

Der Zugriff erfolgt über die {{person.____ }} Anweisungen. Wie man aus dem Code erkennen kann sind sowohl IF abfragen, wie auch for Schleifen möglich. Der Sprachumfang von DJANGO wird hier detailiert beschrieben.

Der REST Service liefert übrigens folgenden Code zurück:

Dieses Konstrukt lässt es nun zu, dass wir ohne Rendering Aufgabe für den Server die verschiedensten Personen öffnen:

Der vollständige Code wird im Juni auf openNTF veröffentlicht und ist unter der Apache Lizenz verfügbar.


Warum FAST Prototyping nicht RAD (Rapid Application Development) ist… und warum wir lieber SCRUM einsetzen

Innerhalb der WebGate Consulting AG, meinem Arbeitgeber, sind wir es gewohnt, wichtige Themen kontrovers zu diskutieren. Dabei werden natürlich auch Argumente von Spezialisten ins Feld geführt. Eine solche Diskussion hat mich zu folgendem Video geführt:

Natürlich bin ich nicht mit allem Einverstanden, was Herr Goslin innerhalb dieses Videos sagt, aber vieles davon hat mich zum Nachdenken angeregt. Vorallem die Aussage, dass ein Prototyp weggeschmissen wird, und danach wird nochmals neu begonnen. Gemäss James Gossling ist das Ziel eines Prototyps zu sehen, ob etwas so funktionieren kann, während Entwicklung mit Architektur und Engineering zu tun hat.

Ich glaube, dass wir gerade in der Notes / Domino Welt bezüglich FAST Prototyping und RAD einen grossen Fehler gemacht hatten. Betrachtet man RAD nämlich unter der Lupe, so sind im RAD sehr viele Ähnlichkeiten zu SCRUM zu entdecken. Und dabei geht es NICHT darum einen Prototypen zu bauen, sonderen Elemente um Element wird lauffähige Software entwickelt. Auch wenn dem Endkunden in kurzen Zyklen jeweils ein lauffähiges Element seiner Software gezeigt wird, sollte bei RAD vom Endprodukt her entwickelt werden.

Nun wird aber schon in der Beschreibung von RAD das Wort Prototype verwendet, was nicht sehr förderlich ist. Das Problem beim Prototype ist, dass “Abkürzungen” gemacht werden. So werden zum Beispiel Schichten vermischt oder gar übersprungen. Error Handling wird nicht implementiert und noch vieles weiteres wird nicht korrekt umgesetzt. Das gleiche Problem ergibt sich auch den vielen CodeSnippets, die im Internet herumschwirren. Die Snippets zeigen eine isolierte Funktionsweise eines Elementes auf. Auch hier werden einfach starke Vereinfachungen gemacht.

Wer jetzt bereits bei SCRUM nachgeschaut hat, fragt sich sicher, was denn anderst ist als bei RAD. Nun für mich ist eines der signifikaten Elemente, dass der Fokus auf den Stories des Users liegt. In kurzen Sätzen (3 Zeiler) definiert der Kunde, was er will. Und er gewichtet, was er haben will. Der Architekt der Software hingegen ermittelt die Abhängigkeiten der Stories und kann so aufgrund der Kundengewichtung ermitteln, welche Stories wie miteinander zusammen hängen. Das iterationsbasierte Vorgehen liefert dem Kunden lauffähige Software mit abgeschlossenen Userstories. Und abgeschossen heisst eben: ABGESCHLOSSEN. Kein Prototype, sondern es ist für die Userstorie alles umgesetzt und es wurde (da auch hier der Scope auf dem Endzustand ist) eine saubere Architektur und korrekte Entwicklungsprinzipien umgesetzt.

Mit dem Fokus auf die Userstory wird automatisch auch der Fokus auf den Kundennutzen gelegt. Ich glaube dies hilft auch in der Offertphase. Die Frage ist dann nämlich ganz simpel, ist die Umsetzung der Userstories die Offerte wert? Ist mein Nutzen so gross, dass ich sehr gerne das Geld ausgebe?


Tags: , , , ,