Comprehension Effort and Programming Activities: Related? Or Not Related?
Download as PPTX, PDF1 like333 views
The document investigates the correlation between program comprehension effort and various coding activities such as navigation, editing, debugging, and building. It utilizes empirical analysis methods like Halstead's metrics and statistical tests to analyze session data, categorizing them into high and low comprehension effort groups. Results suggest significant relationships, particularly that increased navigation corresponds to higher comprehension demands, leading to implications for tool design and future research directions.
Comprehension Effort and Programming Activities: Related? Or Not Related?
- 1. Comprehension Effort and Programming Activities: Related? Or Not Related? Akond Rahman aarahman@ncsu.edu https://akondrahman.github.io/
- 2. Motivation “the amount of navigation may correlate with the need for code understanding”
- 3. Motivation “the amount of navigation may correlate with the need for code understanding”
- 4. Research Question • RQ: How does program comprehension effort relate with navigation frequency, edit frequency, debugging frequency, and build event frequency?
- 5. Methodology Filtering Comprehension Effort Calculation Dividing Sessions Empirical Analysis Session Duration and Method Information Halstead’s Metrics K-Means Clustering Mann Whitney U Test and Cliffs Delta
- 6. Results: Dataset Summary Measurement High Low Session count 33 704 Build event count 239 42,880 Debugging event count 636 18,458 Edit event count 4,808 228,315 Navigation event count 2,888 184,720 Two groups of sessions: High and Low
- 7. Results: Answer to RQ Metric Median (High, Low) p-value Cliffs Delta Navigation Interval (1.11, 0.04) < 0.001 0.45 Edit Interval (0.10, 0.03) < 0.001 0.34 Debugging Interval (0.09, 0.13) 0.33 0.07 Build Interval (0.00, 0.00) 0.70 0.06 “the amount of navigation may correlate with the need for code understanding”
- 8. Implications • Tool design: advanced IDEs to provide cues • Future work: fMRI studies, programmer information needs
- 9. Summary aarahman@ncsu.edu akondrahman https://akondrahman.github.io/ Contact Motivation “the amount of navigation may correlate with the need for code understanding” Methodology Filtering Comprehension Effort Calculation Dividing Sessions Empirical Analysis Session Duration and Method Information Halstead’s Metrics K-Means Clustering Mann Whitney U Test and Cliffs Delta Results: Answer to RQ Metric Median (High, Low) p-value Cliffs Delta Navigation Interval (1.11, 0.04) < 0.001 0.45 Edit Interval (0.10, 0.03) < 0.001 0.34 Debugging Interval (0.09, 0.13) 0.33 0.07 Build Interval (0.00, 0.00) 0.70 0.06 “the amount of navigation may correlate with the need for code understanding”
Editor's Notes
- #3: Navigating may happen when a programmer regularly stops coding to understand and read code Amount of navigation and short inactivities are correlated
- #6: Session duration at least 10 minutes Has method data Halstead’s metrics: parameters, volume, difficulty … normalization K-means clustering + Silhouette Width
- #8: When programmers are involved in high comprehension they tend to navigate less When programmers are involved in high comprehension they tend to write less code
- #9: - May be a programmer needs some extra information while comprehending a program , the IDE can provide that extra information - what do programmers want they are engaged in high comprehension effort , fmri studies