Welcome a powerful templating engine that is available for every DSL in a general purpose language.
Let’s consider a general purpose language (e.g. Scala, Java, Kotlin) and a library (or a DSL API library) which helps to define objects for some domain.
We use the library in a program to yield a domain objects. This means we are allowed to mix the libraly calls with other general language calls features. This forms a templates.
You may recall a
.jsp condition or loop tags as an example.
For the DSL case the general purpose language turns into a powerfull templage engine. Unlike string templage engines, this approach allows a semantic aware templating as all calls goes to the DSL API library.
Semantic aware templates can be used to enrich The DSL Way approach too.
Let’s consider examples.
A DSL example
I will be using IntelliJ IDEA as an IDE
and Kotlin as \(Target Language \) below. Suppose
we have a DSL API library for logger configuration implemented in Kotlin. An
log4j function yield a logger configuration
(e.g. for Log4j).
Say we have the following code to setup loggers.
The goal is to configure a number of loggers in the exactly same way. Thanks to Kotlin language features, one is allowed to use a loop, e.g.
This is a straitforward example of a template engine. There is nothing specific to be done to have a template engine at all. A mix of languge features and DSL API library calls forms the template engine.
There are no loops support in the logger configuration itself, but thats to DSL API we are allowed to loop over several categories to generate all definitions.
The following part is now a template:
One may go further and decide to extract the collection of categoring into a function. So we turn the logger configuration code into the following
Here we assume the
listAllRootPackages() to return a list of
categories. There is no longer necessary to have this
function to return a constant list. Instead, it can be implemented
as we like it to, e.g. it may scan an application package to collect
all possible root categories. It may use some resources as the input.
Overall, this is the way to turn a static (and declarative) logger configuration to a flexible thing. It is now psedo-declarative. Meaning there is another program, that yields a declarative configuration script during a build phase. On that phase all templates are getting substituted.
A next step is to extract the actual category setup code (a template)
At that point we have a shared library, where a category is declared in the right way and reused. Next we use a function aka template all other the place.
The examples above show how a general purpose language can be turned into a powerfull templating engine for any DSL APIs.
It turns out that a general purpose language features turning it to a powerfull template engine for a given DSL. It’s up to a developer to decide which features to use. The only requirement is to have a properly designed DSL API, so that such transformations were possible.
The example below illustrates a side-effect or a benefit of using The DSL Way to extend an IDE without writing any IDE specific code. It shows how powerfull an \(Original Launguage\) can be form the \(Target Language\) perspective.
You may take a look to the post for more formal DSL description for a logger configurations.comments powered by Disqus