48 - Do software engineering practices impact on identifiers and comments? (ACM SAC 2020) скачать в хорошем качестве

48 - Do software engineering practices impact on identifiers and comments? (ACM SAC 2020)

Identifiers and comments are among the most important sources of information in software. Indeed, an obvious way to understand a piece of code is to read the available comments. When these are not present, program comprehension depends strongly on informative identifier names. In this paper, we investigate the relationship between the use of solid software engineering (SE) practices - such as unit testing and continuous integration - and features of identifiers and comments. To gather code developed with and without the use of well-established SE practices, we adopted a previously developed classifier that predicts whether a repository on GitHub contains an engineered software project. We collected approximately 6,000,000 identifiers - class, interface, field, method, and variable names - as well as 6,000,000 comments from Java code classified as either engineered or non-engineered. We find that identifiers and comments differ significantly between engineered and non-engineered code. Among other findings we discovered that, on average, identifiers are around 10% longer and comments 90% shorter in engineered vs. non-engineered code; method names with the word test - evidence of the use of automated tests - are much more prevalent in engineered code; and comments that warn developers about matters or code to be handled with care - so-called notice comments - are twice as common in engineered code. To the best of our knowledge this is the study that analyzes the largest number of identifiers and comments to date. Pre-print: https://otaviolemos.github.io/papers/...