| Project/Area Number |
60550265
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計算機工学
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
TORII Koji Osaka University, Professor, 基礎工学部, 教授 (10172222)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HIGASHINO Teruo Osaka University, Research Associate, 情報処理教育センター, 助手 (80173144)
KUDO Hideo Osaka University, Research Associate, 情報処理教育センター, 助手 (60133146)
SUGIYAMA Yuji Osaka University, Assistant Professor, 基礎工学部, 講師 (50116050)
FUJII Mamoru Osaka University, Associate Professor, 基礎工学部, 助教授 (00029464)
TANIGUCHI Kenichi Osaka University, Professor, 基礎工学部, 教授 (00029513)
|
|---|
| Project Period (FY) |
1985 – 1986
|
| Project Status |
Completed (Fiscal Year 1986)
|
| Budget Amount *help |
¥2,000,000 (Direct Cost: ¥2,000,000)
Fiscal Year 1986: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 1985: ¥1,200,000 (Direct Cost: ¥1,200,000)
|
|---|
| Keywords | Measurement / Bug Numbers / Programmer's Capability / ソフトウェア設計法 / プログラム作成能力 |
|---|
| Research Abstract |
This research project has developed tools for collecting data and data analysis tools. Using one of these tools, we collected data in our University class for compiler construction experiment. Based on these data, several software predictive models on residual bug numbers are compared. This experiment showed that Yamada-Osaki model predicts with minimum errors comparing to other models. Then we applied Yamada-Osaki model differently from their original fashion, so that we could get more close prediction to the really existed error numbers. The main reason why our modified application succeeded in good approximation value is that we can divide the compiler into two separate modules as a syntax analysis module and a code generation module. So the prediction has made separately in two modules and then superpose two results with some time lag, since the code generation module may be tested after the syntax analysis module has been tested. We wish this approach should be applied widely to e
… More
valuate it.
As a second result we propose a measure of a "programmer's (debugging) capability". Based on the reasonable assumptions that the programmer's capability can be measured by the number of erroneous codes in the source programs and also the consumed time in removal of the erroneous codes if any. Let C be a programmer's capability. Then the reciprocal value of the sum of the life-span of each error, according to the above assumption, can be defined as follows:
1/C = <SIGMA> ( <T_e> /F(p)) where F(p) is some monotonically increasing function depending on the complexity of the problem. Here <T_e> is obtained by post-mortem analysis of program texts by hands. In order to get C without exhaustive efforts of text tracing, we introduce an approximate measure C' which is on-line computable by a function of duration of texts modification. C' and C has a good correlation more than 0.9. And also we developed a tool which evaluates the approximate programmer's capability C' during programming activity. Less
|
