Model-based code generation works: But how far does it go?—on the role of the generator.

The article discusses the use of model-based code generation and its limitations. While there have been successful industrial projects that have used code generation from abstract models, the widespread adoption of code generation from explicitly defined models has not been as successful. The article identifies several challenges, including developers' ability to think in abstract form, the limitations of existing tooling, the need for flexible adaptation of the generator, and weak tool support for defining abstraction levels in a model. The article suggests that a "smart" generator could address these challenges by generating code for various functionalities, such as data transportation, user interfaces, logging, and search functionality. However, the article acknowledges that current code generators are not yet capable of such smartness and that more work is needed in this area. [Extracted from the article]