Full text
| |
| Author(s): |
Elder de Oliveira Rodrigues Júnior
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Computação |
| Defense date: | 2020-08-13 |
| Examining board members: |
Leonardo Montecchi;
Breno Bernard Nicolau de França;
Daniel Lucrédio;
Ricardo Terra
|
| Advisor: | Leonardo Montecchi |
| Abstract | |
Coding conventions are a means to improve the reliability of software systems. They can be established for many reasons, ranging from improving the readability of code to avoiding the introduction of security flaws. However, coding conventions often come in the form of textual documents in natural language, which makes them hard to be managed and to enforced. Following model-driven engineering principles, in this dissertation we propose an approach and language for specifying coding conventions using structured models. We call this language Coding Conventions Specification Language (CCSL). We also propose a model transformation to concretely generate checkers to find violations of the rules specified with our language. To evaluate the proposal, we performed two experiments. The first experiment aims to evaluate the Coding Conventions Specification Language metamodel, while the other aims to check the capability of the derived checkers to find violations of the specified rule in Java codes. The obtained results are promising and suggest that the proposed approach is feasible. However, they also highlight that many challenges still need to be overcome. In the first experiment, we analyzed a total of 216 individual rules from two large sets of existing coding conventions. Overall, it was possible to represent 63% of the considered coding rules using our language. In the second experiment, we selected 53 rules from those implemented in the PMD tool to compare the results between our tool and the PMD tool in three real projects. In general, we achieve equal or better results of the PMD tool in more than half of the selected rules (79%), while only 6% of the rules could not be specified using our language. There are also cases where PMD performed better than our approach (9%) as well as cases where the results were different for each of the tools (6%). We conclude by discussing directions for future works (AU) | |
| FAPESP's process: | 18/11129-8 - A metamodel to support the formalization of coding conventions |
| Grantee: | Elder de Oliveira Rodrigues Júnior |
| Support Opportunities: | Scholarships in Brazil - Master |