Chu-Pan Wong; Yingfei Xiong; Hongyu Zhang; Dan Hao; Lu Zhang; Hong Mei
Year of Publication
2014
Summary
BRTracer introduces segmentation and stack-trace feature on top of BugLocator for bug localization.
Contributions of The Paper
Key Contribution-
BRTracer divides source code files into small series of segments and match the bug report with each segment for an initial similarity score. How to divide a source code into segments is very important (have to determine the value of K (num of segments) and L (maximum number of words in each segments). Then calculate a length score for each source code file with a new parameterized length function- Beta which serves as a magnifier to adjust how much favor to give to larger files.
Bug reports sometimes contain stack-trace information, which may provide direct clues to the buggy files. Therefore, a separate stack-trace information score calculated based on the stack frames order is supposed to be helpful for improving the performance of bug localization. Using some regular expresson and ranking along with an equation, boostScore is calculated.
These two heuristics (segmentation and stack trace) are used with BugLocator to calculate the final score.
Future work: Automatically optimizing the parameter values (Beta/ L/K), consider code structure when it comes to segmentation of source code, language specific syntactic and semantic analysis for stack trace analysis
Publisher
ICSME
Link to The Paper
https://ieeexplore.ieee.org/abstract/document/6976084
Name of The Authors
Chu-Pan Wong; Yingfei Xiong; Hongyu Zhang; Dan Hao; Lu Zhang; Hong Mei
Year of Publication
2014
Summary
BRTracer introduces segmentation and stack-trace feature on top of BugLocator for bug localization.
Contributions of The Paper
Key Contribution- BRTracer divides source code files into small series of segments and match the bug report with each segment for an initial similarity score. How to divide a source code into segments is very important (have to determine the value of K (num of segments) and L (maximum number of words in each segments). Then calculate a length score for each source code file with a new parameterized length function- Beta which serves as a magnifier to adjust how much favor to give to larger files. Bug reports sometimes contain stack-trace information, which may provide direct clues to the buggy files. Therefore, a separate stack-trace information score calculated based on the stack frames order is supposed to be helpful for improving the performance of bug localization. Using some regular expresson and ranking along with an equation, boostScore is calculated. These two heuristics (segmentation and stack trace) are used with BugLocator to calculate the final score. Future work: Automatically optimizing the parameter values (Beta/ L/K), consider code structure when it comes to segmentation of source code, language specific syntactic and semantic analysis for stack trace analysis
Comments
No response