Abstract
Context:
Using feature toggles is a technique to turn a feature either on or off in program code by checking the value of a variable in a conditional statement. This technique is increasingly used by software practitioners to support continuous integration and continuous delivery (CI/CD). However, using feature toggles may increase code complexity, create dead code, and decrease the quality of a codebase.
Objective:
The goal of this research is to aid software practitioners in structuring feature toggles in the codebase by proposing and evaluating a set of heuristics and corresponding complexity, comprehensibility, and maintainability metrics based upon an empirical study of open source repositories.
Method:
We identified 80 GitHub repositories that use feature toggles in their development cycle. We conducted a qualitative analysis using 60 of the 80 repositories to identify heuristics and metrics. Then, we conducted a survey of practitioners of 80 repositories to obtain their feedback that the proposed heuristics can be used to guide the structure of feature toggles and to reduce technical debt. We also conducted a case study of the all 80 repositories to analyze relations between heuristics and metrics.
Results:
From the qualitative analysis, we proposed 7 heuristics to guide structuring feature toggles and identified 12 metrics to support the principles embodied in the heuristics. Our survey result shows that practitioners agree that managing feature toggles is difficult, and using identified heuristics can reduce technical debt. Based on our case study, we find a relationship between the adoption of heuristics and the values of metrics.
Conclusions:
Our results support that practitioners should have self-descriptive feature toggles, use feature toggles sparingly, avoid duplicate code in using feature toggles, and ensure complete removal of a feature toggle.
Using feature toggles is a technique to turn a feature either on or off in program code by checking the value of a variable in a conditional statement. This technique is increasingly used by software practitioners to support continuous integration and continuous delivery (CI/CD). However, using feature toggles may increase code complexity, create dead code, and decrease the quality of a codebase.
Objective:
The goal of this research is to aid software practitioners in structuring feature toggles in the codebase by proposing and evaluating a set of heuristics and corresponding complexity, comprehensibility, and maintainability metrics based upon an empirical study of open source repositories.
Method:
We identified 80 GitHub repositories that use feature toggles in their development cycle. We conducted a qualitative analysis using 60 of the 80 repositories to identify heuristics and metrics. Then, we conducted a survey of practitioners of 80 repositories to obtain their feedback that the proposed heuristics can be used to guide the structure of feature toggles and to reduce technical debt. We also conducted a case study of the all 80 repositories to analyze relations between heuristics and metrics.
Results:
From the qualitative analysis, we proposed 7 heuristics to guide structuring feature toggles and identified 12 metrics to support the principles embodied in the heuristics. Our survey result shows that practitioners agree that managing feature toggles is difficult, and using identified heuristics can reduce technical debt. Based on our case study, we find a relationship between the adoption of heuristics and the values of metrics.
Conclusions:
Our results support that practitioners should have self-descriptive feature toggles, use feature toggles sparingly, avoid duplicate code in using feature toggles, and ensure complete removal of a feature toggle.
| Original language | English |
|---|---|
| Article number | 106813 |
| Pages (from-to) | 1-14 |
| Number of pages | 14 |
| Journal | Information and Software Technology |
| Volume | 145 |
| Early online date | 14 Jan 2022 |
| DOIs | |
| Publication status | Published - 1 May 2022 |
Bibliographical note
Funding Information:The first author is funded by North Carolina State University . The second author is funded by the National Security Agency (Science of Security Lablet) at North Carolina State University. We thank all the reviewers for their valuable feedback. We also thank the members of the RealSearch group.
Publisher Copyright:
© 2022 Elsevier B.V.
Keywords
- Feature toggle
- Continuous integration
- Continuous development
- Heuristic
- Metric