How to Make Your Advertisement Look Like Spam

I had experienced ZeroTurnaround‘s marketing and sales before, and I thought it was a bit agressive, although their products seem really nice and cool, so I had forgiven them at the time… However, looking at the numerous links to their product JRebel on CodeRanch just looks like quite offending spam to me. Consider an arbitrary question:

The question is about a silly NullPointerException and its answer is straightforward. Now the JRebel link at the bottom:

I agree. Here's the link
I agree. Here’s the link…

“I agree. Here’s the link” – Whatever. Completely off-topic, no? It looks as though CodeRanch had a spam filtering problem which was exploited by a decent company selling decent products. Why would they do that? Well, I understand that being able to put your link on one of the Top 20 Java Websites can be good for search engine optimisation. But making it look like spam will probably start pissing off potential customers, no?

The crazy thing is, it really isn’t spam, it’s a regular advertisement on CodeRanch. Scroll down on this question here:

I’m really curious if this advertisement works out for ZeroTurnaround! I’m also curious about your opinion, and whether you also think this is spammy.

10 Things in SQL Server Which Don’t Work as Expected

So far, I have been blogging about curious RDBMS caveats mostly related to Oracle and MySQL databases. Some examples:

But there are also other databases, which have 1-2 things that do not work as expected. For example, SQL Server. Here’s an interesting blog post, showing 10 things in SQL Server which don’t work as expected:

Nice Tutorial for Creating an External DSL with Xtext

When blogging about DSLs, I’m mostly blogging about internal domain-specific languages, because jOOQ is a good example for internal DSLs. But external domain-specific languages are an interesting topic as well. As an external DSL does not depend on any host language (e.g. Java), it can be much richer in syntax and expressivity. On the other hand, you need tools to parse, compile and maybe even interpret / run them.

Here’s a very interesting blog post showing how to create an external DSL with Xtext, Eclipse’s DSL tool chain. It creates a simple DSL to describe video game control flows and shows how Eclipse Xtext immediately provides an editor with all sorts of IDE features for that custom DSL:

The Great Benevolent Dictator Getting Some Beating

I know how it feels to be a benevolent dictator. I am one myself on my jOOQ User Group. Supporting Free Open Source Software isn’t always easy. Sometimes you get demanding users, sometimes you get grumpy bugfixers. And the occasional troll. You try to be nice and sometimes, you make mistakes, too. But hey, even if it’s free and Open Source, that’s not an excuse for being rude towards your users (= customers!). Except, of course, for The One Great Benevolent Dictator. Everyone should know by now who I’m talking about. But not everyone accepts his rudeness. As seen on the Linux Kernel List:


Internal DSLs on the Fast Lane

I’ve read this interesting article about internal DSLs in Java, a short summary of Martin Fowler’s book on DSLs in general. I’ve been blogging about external and internal DSLs quite a lot myself, naturally, as jOOQ is the largest and most advanced free and Open Source implementation of an internal DSL in the Java ecosystem. Unlike some other DSLs that are currently being developed, jOOQ uses a BNF as a basis for its API. This guarantees that not only simple method chaining, but also grammar-like contexts can be formalised in an API.

How to construct such an API for your own DSL and grammar manually was explained in this popular blog post here:

SQL Templating with jOOQ or MyBatis

Many people compare jOOQ with MyBatis. Both are seen as popular alternatives to Java’s persistence standard JPA, as both are much more SQL-focused than JPA itself. When comparing the two tools, the first obvious difference is this:

  • jOOQ is an internal domain-specific language modelling SQL through a Java fluent API.
  • MyBatis is an XML-based SQL templating and mapping engine where dynamic SQL can be created through an XML-DSL.

MyBatis’ current success is mostly based on it having provided a viable alternative to JPA in a time when JPA was still a controversial standard, and when JPA had to prove that it is better than JDO, which solves very similar problems. This provided alternative is something that many SQL-oriented users like to have:

  • The separation of Java and SQL code, extracting SQL code into external files. This allows DBA to patch SQL strings in productive environments, adding hints, and other tuning.
  • The automatic mapping of tabular result set data to objects. This too is achieved in the same XML DSL as the dynamic SQL specification.

Implementing SQL templating with jOOQ

These things can be achieved with jOOQ as well. But unlike MyBatis, jOOQ’s SQL templating (as will be introduced in jOOQ 3.2) will not make use of a proprietary templating language. You should be able to make your own choice of language, providing jOOQ with a very simple adapter to it. This will allow for using:

Let’s have a look at a Velocity template example. This example adds a dynamic list of ID parameters to the WHERE clause:

  t_author a
  t_book b ON = b.author_id
  1 = 0
#foreach ($param in $p)
  OR = ?

The above template can be passed to the following jOOQ Template implementation, which uses arbitrary input objects to produce a concrete jOOQ QueryPart. A QueryPart is an object that can render SQL and bind variables:

class VelocityTemplate
implements org.jooq.Template {

  private final String file;

  public VelocityTemplate(String file) {
    this.file = file;

  public QueryPart transform(Object... input) {

    // Velocity code
    // -----------------------------------------
    URL url = this.getClass().getResource(
    File file = url.getFile();

    VelocityEngine ve = new VelocityEngine();
    ve.setProperty(RESOURCE_LOADER, "file");
      new File(file ).getAbsolutePath());

    VelocityContext context = new VelocityContext();
    context.put("p", input);

    StringWriter writer = new StringWriter();
    ve.getTemplate(file, "UTF-8")
      .merge(context, writer);

    // jOOQ Code
    // -----------------------------------------
    return DSL.queryPart(writer.toString(), input);

Very simple glue code. As you’re in full control of your template engine implementation adapter, you can also add caching, object pools, etc to your adapter.

The above template can then be used easily throughout the jOOQ API, wherever jOOQ allows for plain SQL. For example as a top-level query:

Template tpl = new VelocityTemplate(

   .resultQuery(tpl, 1, 2, 3)

Or as a nested select, embedded in jOOQ’s typesafe DSL:

   .from(new TableSourceTemplate("my-table.vm"))

Of course, you can also take advantage of jOOQ’s record mapping features, which allows you to implement your own custom table to object mapping algorithms. This may often be a better choice than relying on any hard-wired XML configuration, such as the one from MyBatis:

List<MyType> result =
   .from(new TableSourceTemplate("my-table.vm"))
   .fetch(new RecordMapper<Record, MyType>() {
      public MyType map(Record record) {
        // Custom mapping logic here

Or with Java 8:

List<MyType> result =
   .from(new TableSourceTemplate("my-table.vm"))
   .fetch((Record) -> new MyType().init(record));

The possibilities are great

SQL templating is a powerful tool when you prefer simple, string-based SQL that can be tweaked every now and then with a little loop or if statement, to inject some dynamic SQL clause. There are a couple of SQL engines that try to solve this problem in one way or another:

Among the above, all tools ship with a simple, proprietary template language, but jOOQ is the only one encouraging you to use your template engine of choice and thus offering arbitrary templating extensibility in the future.

Usability vs. Reusability

This blog post I’ve found from 2009 has a nice way of looking at the problem of comparing the ease of use with the ability to reuse. It claims that usability and reusability is always a tradeoff between building

  • a heavyweight, coarse-grained software component with few dependencies (very usable)
  • a lightweight, fine-grained software component with complex dependencies (very reusable)

The following picture nicely depicts this relationship:

Image taken from:

Indeed, coarse-grained components are hard to reuse because they try to solve too many problems in the context of the coarse overall point of view. But I’m not sure if they’re necessarily easier to “use”. Being coarse, and thus complex, they may have solved the problem in the wrong way. And since they’re so complex, they cannot be changed easily to fit a slightly different problem domain. Since time can change the original problem setup (and it will in any project), heavyweight, coarse-grained components often cannot even be used nor reused for their “original” purpose. I.e. a coarse solution that is developed in a project over several years cannot be finished, because it cannot be changed after those four years.

And fine-grained components aren’t necessarily hard to use. It is possible to create components with very little dependencies, such that they do not introduce a lot of complexity. jOOQ is one example of such a component, which has no dependencies itself, apart from the JDK. But jOOQ is a library and not a business module, i.e. it implements horizontal reusability, not vertical reusability.

So, let’s hope the original post was not entirely correct and there is a good, middle way! See for yourself: