Incremental MTL vs. GPLs: Class into Relational Database Schema Transformation Case

Model transformation languages (MTLs) are domain specific languages tailored to express model-to-model transformation programs. They typically offer hihger-level syntactic constructs, such as rules, and specific features, such as automatic traceability support, than general-purpose languages (GPLs). Moreover, some MTLS allow for multiple execution modes, such as incremental or bidirectional, based on a single specification. Many MTLs have been proposed over the past decades but GPLs are still widely used to write model transformations in practice. Previous work has identified some reasons for this, in the context of batch execution mode. Our hypothesis is that the situation is different for other execution modes.

Therefore, this transformation tool contest calls for incremental solutions implemented using various MTLs and GPLs, with the purpose of building a data set consisting of labeled solutions specified in diverse languages. The overall objective is to leverage this data set to better undertand whether GPLs are up to incremental tasks or whether MTLs are significantly more appropriate.

Contents of this repository

• [case description](case description/ttc23_GPL_vs_MTL.pdf): TTC 2023 case description
• labeling: examples of how to provide labels to an MTL (ATOL) and GPL (Java) solution
• metamodels: source and target metamodel as well as the Change metamodel used for incremental updates of the input model
• metrics: R-scripts to extract plots from the labels collected with the LabelCounting tool, using a Docker setup
• models: 13 cases of input source models, change models and expected models (all in xmi-format)
• solutions: solutions to the case implemented in various transformation languages, each in one directory
• utils: tools that benchmark the solutions, and automatically count the labels provided in the solutions
  • the SolutionDriver which executes and benchmarks a given transformation solution ,
  • the Comparator which evaluates the outcome of batch and incremental transformations
  • the LabelCounter which scans a transformation solution and counts how many words are used for a transformation aspect
  • ecore2class: utility Ecore2Class transformation which can turn any Ecore model into an instance of the expected source model

Using Github maven repository

We provide two maven packages, the fuzzer / model comparator and the solution driver with metamodels, they are hosted on Github. They are published publicly but Github requires users to authenticate to access their maven repository.

For a Gradle example you can read the repository declaration from ATOL build.gradle file. To generate your credentials you can read Github documentation for maven or Gradle.

Checking the solutions

Register the Github authentication information properly, as explained in the previous section (e.g., via the GITHUB_ACTOR and GITHUB_TOKEN environment variables, or in ~/.gradle/gradle.properties). From the utils/ sub-directory, launch check_solution.sh. For instance:

	$ cd utils
	$ ./check_solution.sh

Requirements

• ATL:
  • java 11+
• ATOL:
  • build: java 17 & <21
    • fails to build with java 21 because xtend compiler is not compatible yet
  • run: java 17+
• bxtend:
  • java 17+
• cheptre:
  • java 17+
  • docker
    • user must be able to start/stop containers
• csharp, csharp_cp, nmf:
  • dotnet 7
• java:
  • java 14+
• Refinery:
  • java 21+
• yamtl-java:
  • java 17+