An IR-based Approach Combining With Code Dependencies For Updating Requirements

When maintaining and evolving software,a requirements specification which is consis-tent with the code implementations provides valuable knowledge for supporting several maintenance and evolution tasks.However,updating the requirements is still a manual task that is expensive and time consuming.Therefore,maintainers usually change the code directly and leave the requirements unchanged.This leads to requirements rapidly becoming outdated and useless,then weakening the maintainability of software system.From the perspective of code changes,existing works using an IR-based approach to tracing code changes descriptions to requirements for detecting the impacted parts of the requirements specification in order to facilitate the efforts for maintainers to update requirements.However,when constructing code changes descriptions,existing works mainly focus on the changed parts of code but ignore the context code of code changes which have close code dependencies with the changed ones.These context code may also provide valuable information about functional descriptions to supply textual information for code changes descriptions.Besides,facing with the detected requirements,maintainers may still need more knowledge which is related with the code changes to finish updating the requirements,the existing works have not cover this part.Meanwhile,the existing techniques can hardly be used to solve real-world problems of software activities because currently there is lack of tools for supporting requirements update in field.For the above problems,we finished several works as following:1.We proposed an IR-based approach combining with code dependencies for up-dating requirements.The approach contains two parts:automatically detect out-dated requirements and recommend knowledge about the update.In the detecting part,we bring the contexts which are tight with the changed code in structure to supply code changes descriptions by analyzing the closeness of code dependen-cies.In the update recommending part,we combine the code dependencies with textual information to measure the weight of code elements,then recommend the hotspot of these code elements which are related to requirements update for maintainers.2.We evaluated the above requirements updating approach on three different case studies.Meanwhile,we explore the technical source of closeness analyzing of code dependencies and we evaluated its effectiveness on traceability generation problem.3.We also developed an assistant tool for updating outdated requirements and inte-grating the above approach into it.