软件迁徙中的应用编程接口翻译方法和缺陷修复研究

Software migration is a common activity in software evolution. In software migration, a source software application is translated into a target software application of the same functionalities. Software migration typically takes much effort, which is tedious and could be automated. However, due to various limitation, software migration in industry is mostly manual and error-proning. During software migration, a key step is to translate source API elements into target API elements that have the same functionalities. Modern API libraries provide thousands of API elements, but due to the huge effort to write API mapping manually, existing tools typically support only a small portion of API elements. To address the problem, researchers have proposed various approaches. Their work produces some positive results, but there are still many barriers to overcome. To improve the state of the art, we plan to explore a novel approach that combines advanced techniques to mine API mapping from multiple data sources. Our novel approach should be able to mine more comprehensive API mapping relations. In addition, we plan to explore a novel approach that detects and repairs defects that were introduced during software migration. In summary, our research could be further improve the state of the art, increasing the quality and reducing the effort of software migration.

软件迁徙是软件演化中一种常见的活动。在软件迁徙中，一个软件项目被修改或者翻译为功能一致的另一个软件项目。软件迁徙需要较大的代价，其中存在大量可以自动化的重复劳动。但受制于各种缺陷，软件迁徙的自动化程度很低，并容易引入缺陷。其中，一个关键的步骤是如何将原来软件系统中的应用编程接口正确地翻译为新软件系统里的应用编程接口。现代的软件库提供了成千上万的应用编程接口，受制于编写的巨大代价，现有的工具一般只支持小部分应用编程接口。为此，研究者提出了各种技术自动地获取该映射关系，并取得了积极的成果。申请人在该方向也有较强的论文发表。为了进一步深入探索该研究方向，本申请书计划提出一种汇合各种处理技术，能综合利用各种数据源的应用编程接口翻译方法，以及一种缺陷的检测和修复方法，该方法应能自动检测和修复软件迁徙中引入的各种缺陷。总之，通过本项目书的研究，我们期望能够提高软件迁徙的质量，并减少其代价。

在项目支持下，从2015年以来我们对软件迁徙中进行了深入研究，具体涉及实证分析、应用接口处理和缺陷修复几个方面。截至2019年12月，我们已经发表论文二十余篇，包括9篇被CCF评级为A类的论文和8篇被评级为B类的论文。本项目研究涉及应用迁徙和缺陷修复两个方面。基于我们的研究成果，在应用迁徙方面，我们开发了名为TestMig的迁徙工具，可以将Android的测试用例翻译为iOS上的测试用力；在缺陷修复方面，我们开发了SOFIX和MIMO等缺陷修复工具。上述工具在我们的验证中展示了较好的效果，一些已经开源。在执行过程中，我们还发现了一些新的研究问题。例如，我们发现现有的工具不能很好的支持分析修复历史。我们正在申请基金以期能够更深入的进行研究。