At the moment, I’m playing with the release plugins in a sandbox project and it’s created lots of tags. I want to start afresh, so I’ve written a little utility to get rid of all the tags

git tag --list | xargs -I {} git tag -d {}
git fsck --unreachable 2>&1 | grep 'unreachable tag' | awk {'print $3'} | xargs git show | grep ^tag | awk {'print $2'} | xargs -I {} git push origin :refs/tags/{}

(2 years since the last post, … shocking😀 )

I love this. I also have a headache now…


A co-worker of mine recently ran across this post by Eamonn McManus, “Using the builder pattern with subclasses,” and after due consideration, settled on what McManus calls “a shorter, smellier variant.”

Shorter because:

  1. it has only one Builder class per class

Smellier because:

  1. it uses raw types
  2. the builder’s self() method requires an unchecked cast

Frankly, I don’t find the shortness here a compelling argument for the smell, but I also think there’s more shortness to be found in McManus’ design while remaining fragrant. Thus:

 class Shape { private final double opacity; public double getOpacity() { return opacity; } public static abstract class Builder<T extends Shape> { private double opacity; public Builder<T> opacity(double opacity) { this.opacity = opacity; return this; } public abstract T build(); } public static Builder<?> builder() { return new Builder<Shape>() { @Override public Shape build() { return new Shape(this); } }; } protected Shape(Builder<?> builder) {…

View original post 300 more words

lapis server 1.2.9

Posted: 18 July 2014 in Teamsite

It has been a somewhat short sprint with not a lot to talk about. The interface is getting more and more usable, which is a good thing since I’m using it to manage the sprints. The filters and comparators and really helping with this.

Server instance name

the lapis start program has been modified to take a second argument which is used to distinguish between multiple running instances. The argument is simply there to appear on the command line when checking the running processes as shown below:

$ ps -ef | grep LapisServer.jar laurent 21255 21010 5 10:42 ? 00:01:26 java -Djava.util.logging.config.file=etc/ -DinstanceName=dev -jar lib/LapisServer.jar etc/ laurent 21067 21065 5 10:42 ? 00:00:39 java -Djava.util.logging.config.file=etc/ -DinstanceName=tst -jar lib/LapisServer.jar etc/

Before this was introduced, there was no real way to find out which process was for which instance. Now I know…

Due Date & Priority

I’ve added the due date and priority to workflows and tasks to complete the move towards Getting Things Done. Along with that, I created due date filters and comparators and priority filters and comparators.

Lapis Server 1.2.8

Posted: 7 July 2014 in Teamsite

It just never stops! Lapis Server 1.2.8 is now done, but it’s a release that is part of a bigger program of work, so I’m straight on 1.2.9

I had to create a few things to make it all available as a software platform.

  • a Signup process (signup form, signup servlet, signup workflow (it’s pretty nice that after signup up to a workflow engine, the workflow engine uses a workflow to send you a nice email – it uses an email task!)
  • Some terms and conditions and a privacy policy
  • A new stylesheet
  • Changes to the login process to include “forgot username”, “forgot password” functionalities. Again, both of these use a workflow to send the emails
  • Rewrite rules to make sure you can’t see all the workflows you’re not supposed to.
  • Role based security constraints to match what will be the various packages. The tomcat realm class now reads the role of the user so the web app has the same roles as the Lapis server.
  • Web Interface improvements, such as moving the tasks link into its own menu, links to the privacy policy and terms and conditions
  • A task status filter (active, completed, rejected, all) which somehow had been missed out in the design process and is DEFINITELY required to make it user friendly. The default is “active”, so that the task link is effectively a link to your current to-do list.
  • An available workflows datasource to make it more user friendly by selecting workflows from a drop-down list instead of typing them in. The datasource only lists the files the user has access to.
  • Changes to the server to be able to bind itself to a specific network interface in multi-home network environments such as that of the redhat cloud.
  • Changes to the client command line tools to bind themselves to a specific network interface in multi-home network environments such as that of the redhat cloud.

what’s next?

as part of my 1.2.9 sprint, I am looking at finishing the integration of the redhat environment (with the shutdown and startup scripts and a whole round of testing) and send a link to a few people to do a soft launch and gather feedback before getting me a real domain name!
After that, I’ll see how to get a payment solution in place to offer upgrades and start the process of making changes towards working with an organisation’s groups.

Lapis Server 1.2.7

Posted: 29 June 2014 in Teamsite

The lapis server version 1.2.7 is now done.

JCR integration

As promised, the JCR intergration has been done.

We can define repositories centrally via an XML configuration file such as the one below.

<repositories> <repository name="" path="/jackrabbit/cq5/dev/author" type="jsr170" url="http://localhost:4502/crx/server" username="admin" password="SrAxtm8ihY+hBPYklrYYJQ=="> <action name="Go to" pattern="http://localhost:4502/crx/de/index.jsp#/crx.default/jcr%3aroot${path}" /> </repository> <!-- <repository name="" path="/jackrabbit/cq5/dev/publish" type="jsr170" url="http://localhost:4503/crx/server" username="admin" password="SrAxtm8ihY+hBPYklrYYJQ==" /> --> </repositories> Note that we can associate a number of actions for the repository. This is used by the web app to display link to your repository browser app (in my example, it's Adobe's CQ5) 
As per the previous post, the client can retrieve the details of the repository and connect to it. The node action (e.g. the custom java of your tasks) would connect to the JCR in the following way: 
 List<String> repositories = client.getRepositories(); for (String repository : repositories) { Repository jcrRepository = JcrUtils.getRepository(client.getJCRRepositoryURL(repository)); Credentials credentials = client.getJCRCryptedCredentials(repository); Session jcrSession = jcrRepository.login(credentials); /* do something with the nodes */ List<String> attached = nodeToken.getGraphExecution().getAttached(); for (String anAttached : atached){ /* e.g. /jackrabbit/cq5/dev/author/some/path/within */ String basePath = client.getJCRRepositoryPath(attached); /* e.g. /jackrabbit/cq5/dev/author */ LapisPath lapisPath = new LapisPath(basePath,attached);; /* e.g. /some/path/within */ Node node = jcrSession.getNode(lapisPath.modulatePath()); /* retrieve the node at the modulated path */ / * do something with it */ } }

Attached JCR nodes

Attached nodes are now also supported by the command line tool via a “lapis.attached={path}” argument. The new command line tool also supports attaching and detaching jcr nodes via the -attach and -detach options.

The web application allows to browse the repository and select nodes to attach.

LapisServer on RedHat OpenShift and Saas

I’ve started placing a new version of the web app on OpenShift, the Redhat cloud server. This is to be able to offer the software as a service. I still have to define my pricing plan, but the free option will enable users to instantiate a “todo” workflow. the first upgrade other options would enable users to instantiate a “assign” workflow where groups and group tasks would be enabled. The next upgrade option would see the introduction of custom workflows where email tasks would be enabled. The final upgrade option after that would be a custom instance of the server with our support.

Multi-homed environments

An interesting side effect of looking at the Redhat Openshift was that the servers have multiple IP addresses and the default RMI configuration was of course picking the one which had security restrictions. It is now possible to add a server property named “bindAddress” which forces the Lapis server to listen on a particular network interface. Likewise for the client tools, the local end of the socket can be made to pick the correct IP address when connecting to the server.

Audit trail

In addition to the workflows and the tasks showing an audit trail of when they are created, activacted, rejected and completed, all the events ( such as when a property is changed or a JCR node attached) are now stored, thus giving a complete audit trail of what happened during the life of the workflow.

What’s next?

I will be focusing on the Saas web app for signup and the todo workflow (I’m already 2/3rds through so it shouldn’t take long) and then start on the assignment workflow and find out what that means about the groups.