On the Investigation of Exception Pull Request Characteristics: Exploring the Apache Ecosystem

Robustness is critical for ensuring that software functions correctly under adverse conditions. Exception-handling mechanisms in programming languages enable developers to structure the code that deals with these adverse conditions. However, implementing the exception-related code may present significant challenges to developers. In this context, this paper reports a study in which we investigated the attention given by developers to the exception-related code across Java projects of the Apache ecosystem. We analyzed the exception-related pull requests (exception-PRs), which were detected with a validated heuristic, to characterize and classify the aspects of these contributions. We produced a comprehensive dataset of 986 exception-PRs detected with the proposed heuristic. A tally of 280 PRs was manually analyzed to characterize and classify the aspects of these contributions to the exception-related code. Our results reveal several trends. First, no statistically significant differences are observed in complexity metrics for exception-PRs versus non-exception-PRs. Similarly, no significant differences are observed in developers' behavior metrics, indicating consistent engagement regardless of whether the pull request addressed exceptional code or not. Most exception-PRs focus on system improvements rather than bug fixes, underscoring proactive efforts to enhance software robustness. Moreover, the most frequently addressed aspects of exceptional code in these exception-PRs are: (i) the external representation of the adverse situations to end-users (more than 40% of the PRs), and (ii) the implementation of effective error handling actions (nearly 35% of the PRs) to promote the program recoverability. Interestingly, a significant proportion of exception-PRs addresses simultaneously various of these (and other) aspects.

RQ1: How does the dynamics of exception-PRs and non-exception-PRs differ?

Artifacts

• Full list of exception-PRs identified by the heuristic in the Druid.
• Full list of exception-PRs identified by the heuristic in the Pulsar.
• Full list of exception-PRs identified by the heuristic in the Flink.

RQ2: What are the aspects addressed by exception-PRs?

Artifacts

• Full list of exception-PRs and answers from manual validation.

Full tables

Exception aspect occurrence	%	#	Bug fix	Improvement	New feature	Test	Other
Error representability - External	42.52%	108	19.44%	77.77%	1.85%	0.92%	0.00%
Error recoverability - Effective error handling	36.22%	92	48.91%	46.73%	2.17%	1.08%	1.08%
Error detectability	23.62%	60	40.00%	56.67%	0.00%	3.33%	0.00%
Error representability - Internal	20.47%	52	36.53%	63.46%	0.00%	0.00%	0.00%
Error recoverability - Program continuity	20.47%	52	55.76%	42.30%	0.00%	1.92%	0.00%
Other	19.69%	50	68.00%	26.00%	0.00%	6.00%	0.00%
Error propagartion	16.54%	42	26.19%	69.05%	2.38%	2.38%	0.00%
Error recoverability - Cleanability	7.48%	19	36.84%	63.15%	0.00%	0.00%	0.00%

Table VI: Exception aspects and the aim of the contribution in exception-PRs

Exception aspect occurrence (combinations)	%	#	Bug fix	Improvement	New feature	Test	Other
Error recoverability, Error representability	12.99%	33	39.39%	57.58%	3.03%	0.00%	0.00%
Error detectability, Error recoverability, Error representability	7.09%	18	44.44%	55.56%	0.00%	0.00%	0.00%
Error propagartion, Error recoverability, Error representability	5.51%	14	42.86%	50.00%	7.14%	0.00%	0.00%
Error detectability, Error representability	4.33%	11	18.18%	81.82%	0.00%	0.00%	0.00%
Error detectability, Error recoverability	3.94%	10	70.00%	30.00%	0.00%	0.00%	0.00%
Error propagartion, Error representability	3.94%	10	10.00%	90.00%	0.00%	0.00%	0.00%
Error detectability, Error propagartion, Error recoverability	1.57%	4	25.00%	75.00%	0.00%	0.00%	0.00%
Error propagartion, Error recoverability	1.18%	3	33.33%	66.67%	0.00%	0.00%	0.00%
Error detectability, Error propagartion, Error representability	0.79%	2	0.00%	100.00%	0.00%	0.00%	0.00%
Error recoverability, Other	0.79%	2	50.00%	50.00%	0.00%	0.00%	0.00%
Error propagartion, Other	0.39%	1	0.00%	100.00%	0.00%	0.00%	0.00%
Error detectability, Error recoverability, Other	0.39%	1	100.00%	0.00%	0.00%	0.00%	0.00%
Error propagartion, Error representability, Other	0.39%	1	0.00%	100.00%	0.00%	0.00%	0.00%
Error representability, Other	0.39%	1	0.00%	100.00%	0.00%	0.00%	0.00%

Table VII: Combination of exception aspects and the aim of contribution in exception-PRs.