2004 Fiscal Year Final Research Report Summary
Research on Large-scale Program Corpus for Open Source Software
| Project/Area Number |
14380148
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | Aichi Prefectual University |
|---|
Principal Investigator |
YAMAMOTO Sinichiro Aichi Prefectural University, Faculty of Information Science and Technology, Associate Professor, 情報科学部, 助教授 (40240098)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OHKUBO Hirotaka Aichi Prefectural University, Faculty of Information Science and Technology, Assistant Professor, 助手 (40295580)
KASUYA Hideto Aichi Prefectural University, Faculty of Information Science and Technology, Assistant Professor, 助手 (10295579)
NAOI Tohru Gifu University, Department of Information Science, Associate Professor, 工学部, 助教授 (10207699)
YOSHIDA Atsushi Wakayama University, Center for Information Science, Lecturer, システム情報学センター, 講師 (50283495)
HACHISU Yoshinari Nanzan University, Faculty of Mathematical Science and Information Engineering, Lecturer, 数理情報学部, 講師 (30319298)
|
|---|
| Project Period (FY) |
2002 – 2004
|
| Keywords | Program Corpus / Software Reuse / Software Retrieval / Idiom / Similarity / Clustering |
|---|
| Research Abstract |
he purpose of this research is to establish technologies for building source program corpus of large scale open-source software and constructing sophisticated concordancer.
The main results is listed below:
(1) We proposed Function Call Dependency Graph (called FCDG) for representing typical combination of C program libraries. Node of FCDG is library call and edge of FCDG is data flow between them. By using FCDG, we could formulate the flow between library's return value and actual parameter in highly abstract level.
(2) We proposed the method for providing coding patterns to programmer using FCDG. We demonstrated that our method could automatically extract 238 valued coding patterns from FreeBSD source tree. The patterns consists 92 return-value checking pattern and 61 patterns which represent typical combination of libraries.
(3) On using libraries in software development, the developer looks up the usages for their libraries with the manuals and the actual examples. Some typical examples are shown in the manual, however they are only a part of the usages. On retrieving the actual examples, the results are too many to get the needed information. This research proposes the method to categorize and order the usages. Moreover, this research shows the results that were categorized and ordered usages for 152 software in FreeBSD /usr/src/usr.sbin by our retrieval system. It was enable to categorize an average of 40 usages per library function into 5 categories.
|
