improving code readability models with textual features
TRANSCRIPT
Improving Code Readability Models with Textual FeaturesSimone Scalabrino, Mario Linares-Vásquez, Denys Poshyvanyk, Rocco Oliveto
Software maintenance accounts for
70%of the costs of a project
for (int i=1;i<=100;i++) { int a=((528>>i%15-1)&1)*4; int b=((-2128340926>>(i%15)*2)&3)*4; System.out.println("FizzBuzz".substring(a,b)+(a==b?i:""));}
What does it do?
Not really clear...
for (int i=1;i<=100;i++) { String fizzBuzz = “”; if (i % 3 == 0) fizzBuzz += “Fizz”;
if (i % 5 == 0) fizzBuzz += “Buzz”;
if (fizzBuzz.isEmpty()) fizzBuzz += i;
System.out.println(fizzBuzz);}
What about this one?
Why is it easier to understand?
Code readability
Code readability prediction
Source code
Source code
Source codeSource codeFeatures
Source code
Source codeSource codeFeatures
Training set Machine learner
Source code
Source codeSource codeFeatures
Readable Unreadable
Training set Machine learner
Code readability prediction
Code readability prediction
Code readability prediction
Code readability prediction
Structural features
Code readability prediction
Structural features
Visual features
Code readability prediction
Structural features
Visual features
Two datasets
Something is missing...
Code is text!
New features
Comments readability
Comments readability
Flesch-Kincaid index
Comments and identifiers consistency
Comments and identifiers consistency
Overlap between terms in commentsand terms in identifiers
Identifier terms in dictionary
Identifier terms in dictionary
Percentage of correct terms in identifiers
Narrow meaning identifiers
Narrow meaning identifiers
“Sum” more specific than “computation”
Number of meanings
Number of meanings
“Present” more ambiguous than “Attendee”
Textual coherence
Textual coherence
Consistency between pairs of syntactic blocks
Textual features
Source code
Source codeSource codeFeatures
Readable Unreadable
Training set Logistic regression
Case study
New datasetNew dataset
200 Java snippets
9annotators
Do textual features complement the others proposed
in the literature?
Overlap metrics
Textual Features vs Buse’s
Textual Features vs Posnett’s
Textual Features vs Dorn’s
Textual Features vs Dorn’s
Readability of 12%-21% of snippets can
be explained only using textual features
What is the accuracy of a readability model based
on structural and textual features?
79%
81%
78%
80%
74%
Dataset by Buse and Weimer
ALL-F: All featuresT-F: Textual features
D-F: Dorn’s featuresP-F: Posnett’s features
BW-F: Buse’s features
84%
79%
73%
80%
77%
Dataset by Dorn
ALL-F: All featuresT-F: Textual features
D-F: Dorn’s featuresP-F: Posnett’s features
BW-F: Buse’s features
80%
71%
66%
76%
68%
New dataset
ALL-F: All featuresT-F: Textual features
D-F: Dorn’s featuresP-F: Posnett’s features
BW-F: Buse’s features
80%
71%
66%
76%
68%
New dataset
ALL-F: All featuresT-F: Textual features
D-F: Dorn’s featuresP-F: Posnett’s features
BW-F: Buse’s features
A model which includes all features
achieves an higher accuracy on 2 datasets
In summary...
Code readability for defect prediction
Thanks.