Month: November 2014

Grouping Java skills by categories

I would like to group skill items from typical Java developer resumes – aka CVs – in a somewhat standardized way. See this jsfiddle for some work-in-progress.

I came up with the primary and secondary categories listed below, each with some sample items.

Your feedback is highly appreciated. Please comment on this blog post.

  • Java features and APIs:
    • Configurable JVM feature (Garbage collection, Remote debugging, JMX, etc)
    • Java language feature (Generics, Threads, Annotations, Enums, Lambdas, …)
    • Java SE API (Collections, Date/Time, NIO, java.util.concurrent, …)
    • Java EE API (EJB, JMS, JPA, JSF, JSP, …)
    • Java SE UI (Swing, JavaFX, WebStart, Applets, …)
    • 3rd party Open Source web frameworks (Struts, Wicket, Spring MVC, …)
    • Other 3rd party Open Source framework (Spring, Hibernate, …)
    • Closed-source vendor technology (Java based but no JSR)
  • Java Tools/Servers:
    • Application Servers (Websphere, Weblogic, JBoss, Glassfish, Tomcat, Jetty, …),
    • Developer Tools (IntelliJ, Eclipse, Netbeans, …)
    • Build automation tools (Maven, Ant, Jenkins, Teamcity, Bamboo, …)
    • Profiling and test tools (JVisualVM, SoapUI, JMeter, …)
  • Methodologies (software engineering approaches and practices):
    • Project level: Agile, Waterfall, RUP, …
    • Engineering level: Dependency Injection (DI), Object-Oriented Development (OOP), Continuous Integration (CI), …
  • Relevant non-Java technologies:
    • Web technologies (HTML, CSS, JQuery, JavaScript, SVG, …)
    • Scripting languages (bash, Perl, ksh, Groovy, ruby, etc)
    • Operating systems (Linux, MacOS, Solaris, …)
    • Database servers (Oracle, MS SQL server, Sybase, …)
    • Team tools (Confluence, JIRA, Crucible, …)
    • Other programming languages (C, C++, Cobol, Fortran, Scala, …)

Human readable timestamp for filenames

I use this bash alias (should work in Linux, Cygwin, probably MacOS, maybe other Unixes):

alias timestamp="date --rfc-3339=ns | tr ' '  '_'"

Then use it like this for example to archive a file:

mv somefile somefile_$(timestamp)

You should see something like

`somefile' -> `somefile_2014-11-13_11:45:46.980175800-04:00'

If you don’t like colons in file names, change tr ' ' to tr ' :'.

JQuery: Button to cycle through select options

I wanted an html drop-down list to automatically appear as a simple button that changes color on every click as it cycles through the option values of the (hidden) select element.

The code sample is in this jsfiddle. It uses JQuery.

For each relevant select element, it styles the associated label so it looks like a button, hides the select element and handles clicks on the label button by cycling through the select options.

The display text of each option is used as a tooltip (‘title’ attribute) of the button which also serves as a css selector to color the button according to currently selected option.

The beauty of this is that the selected values of the hidden drop-downs can be submitted by regular form submit and that the form is fully functional if JavaScript is turned off.

The button styling requires CSS3, so IE8 is not suitable. I have tested this successfully in Firefox 33 and IE11.

Design and code visualization

As a team lead at work and for my own good, I want to improve how I visualize system architecture, software design and code structure.

There seem to be at least two angles to this:

  • Show ideas
  • Show (software) reality

For me, ideas often develop top-down, starting from overview level, and require putting something I have in mind on paper or whiteboard, often as a mix of bullet points, sketchy diagram fragments, boxes, ad-hoc dividers and increasingly chaotic arrows.

In the past I have used scans of my paper scribblings or photos of whiteboards to capture the results. But of course, those “snapshots” are not very workable nor professional.

So for idea visualization, I need a flexible yet intuitive mindmapping and/or diagram tool. It should be easy to install or require no installation at all, stylistically appealing but clear, and of course be free/libre open source (i.e. no Visio, please).

I looked at tools like Dia, some online services like LucidChart, and other allegedly free alternatives, and have settled on Pencil Project for now. It is very basic, but seems to provide all I need for now. The cross-platform availability is excellent, it can even be installed as a Firefox extension.

On the other hand, depicting the reality of existing systems and code bases usually requires some sort of generation process that analyzes the real structure, e.g. a network layout, the content of a Java package, an Apache Camel route, etc. and turns it into a renderable file.

For Java source code, I found PlantUML quite interesting, with its javadoc integration and quick dependency diagram support.

I might follow up with more detailed posts about how I use the tools mentioned above. One aspect I have not looked into yet, is how I can embed the visualizations within AsciiDoc.

Java Value Types and Generics specialization

It is exciting to see that Brian Goetz and others are working towards Generics over Primitive Types (List<int> et al) and Value Types (structured immutable data types with value semantics and without object overhead) for Java 10.

The motivation for these initiatives comes from the somewhat clunky divide between objects and primitives in Java.

Objects are the only choice in Java when it comes to collections content and generic type instantiation in general. There is no such thing as List<int> in Java right now.

Objects are also the only option for structured data in Java, including simple aggregation. There are no structs, no tuples, etc. And because the primitive types in Java cannot be “primitively” extended, combined or range restricted, even things like complex numbers, unsigned ints, big decimals and the like all require classes and objects to represent the values.

Arrays are the only way to aggregate a bunch of primitives (of the same type) in Java. But Arrays themselves are objects, with reference identity not value identity, and cannot be used as a substitutes for tuples.

Object purists might argue that Java should just get rid of primitives altogether and strictly follow the “everything is an object” paradigm.

But objects bear management overheads in terms of memory and processing time, bring complexity around identity versus equality, require reference checking for garbage collection purposes, cannot live on the stack, lead to array data in non-contiguous memory, etc.

As the Java architects point out, there currently is an unsolvable conflict between object identity and value semantics in Java: Even if objects a1 and a2 of immutable type A have completely identical state, the JVM cannot treat them as identical “values”, because synchronization relies on monitor object identity, and A1 and A2 would represent separate independent locks. Conflating them into one “value object” would break that distinction.

All this suggests that Java needs to be extended to support structured types with value semantics, conceptually somewhere between primitives and objects. Hopefully Java 10 will come with such types and hopefully smart engineers like Brian Goetz will make sure the resulting type system won’t feel like an incoherent Frankenstein monster.