1.
Refactoring Debt: Myth or Reality?
An Exploratory Study on the Relationship
Between Technical Debt and Refactoring
Anthony Peruma, Eman A. AlOmar, Christian D. Newman,
Mohamed W. Mkaouer & Ali Ouni
The 19th International Conference on Mining Software Repositories
Mining Challenge Track

2.
Intentionally documenting improvements
to the source code
Self-Admitted Technical Debt
Non-optimal code shipped to production
Technical Debt

3.
GOAL
IMPACT
RESEARCH
QUESTIONS
CONTRIBUTION
Explore the repayment of technical debt
through refactoring operations
Provide developers and tool vendors with
insight to improve code quality
Extent of the removal of technical debt through refactoring
Types of debt addressed through refactoring
Dataset and discussion of the impact of
refactoring and debt repayment

4.
Source Dataset
SmartSHARK
MongoDB Release 2.1
(full version)
Extracted Data
▪ SATD Removed Commits
▪ Refactoring Commits
▪ Refactoring Operations
▪ Hunks (Code Diffs)
Quantitative
Analysis
▪ Descriptive
Statistics
▪ Odds Ratio
Qualitative
Analysis
▪ Manual
Analysis of
Code
Custom Python Scripts
Experiment Design
Data Overview
▪ # of projects: 77
▪ # of project with SATD removed: 76
▪ # of commits with SATD removed: 13,259
▪ # of refactoring commits with SATD removed: 7,341

5.
Research Questions
RQ1 – To what extent do developers refactor
code when removing technical debt?
Greater likelihood of debt repayment through refactoring activities:
• 76 of 77 systems have debt removal via refactoring
• 76 systems have an Odds Ratio > 1
• Statistically significant difference in the number of refactoring
operations applied to files
Developers apply a variety of refactoring operations:
• Extract Attribute is frequently applied multiple times followed by
Change Variable Type

6.
Research Questions
RQ2 – What motivates a developer to refactor
their source code to remove technical debt?
Technical debt categories resolved via refactoring:
▪ Error Handling
▪ Code & Structural Improvements
▪ Feature Updates

7.
Code & Structural Improvements
Clean-up Activities
Design Improvements
Includes removing temporary code or
renaming identifiers
Prior studies show that code clean-up is
an activity associated with refactoring
Includes moving code (such as moving
methods, method extraction, etc.) and
data type changes
Frequent bigrams & (stemmed) terms:
“get rid”, “mov”, “remov” & “chang”
Rename attribute
refactoring operation
Change data type
refactoring operation

8.
Feature Updates
Refactoring operations developers
perform to incorporate feature changes
into the system
Frequent terms:
▪ “implement” – “FIXME: our
implementation is flawed….”
▪ “add” – “TODO: Add method to
extract...”
Extract method
refactoring operation

9.
Error Handling
Developers knowingly write code prone
to errors or utilize generic (or auto-
generated) error handling code
(try/catch blocks)
Common bigrams: “catch block”, “error
handling”, “exception handling”, and
“throw exception”
▪ “TODO: We could use a better
strategy for error handling”
▪ “TODO: Fix exception handling”.

10.
Potential Code Quality Tools:
• Support for robust error handling – automatic detection of
shortcomings in the code; customization of auto-generated try-
catch blocks
• Improving the accuracy of refactoring recommendations –
considering the occurrence of SATD in the code
• Aligns with existing studies showing a co-occurrence between debt
repayment and refactoring actions
• Opens the door for potential future work
A more formal and exhaustive set of causes for repayment categories
Conclusion & Takeaways

11.
Thank You!
Anthony Peruma
h t t p s ://pe ru ma. me