Exploring and Improving Code Completion for Test Code

Published: April 15, 2024

Code completion is an important feature in Integrated Development Environments (IDEs). These years, researchers have been making efforts for intelligent code completion. However, existing work on either only considered production code, or did not distinguish between and test code. It unclear how effective models are completion, nor whether we can further improve it. In this work, focus the of We first find through experiments that suboptimal Then analyze specific characteristics observe has inter- intra-project similarities, a strong relationship with its focal class other classes depending (i.e., focal-related code). By incorporating from projects to fine-tune models, leverage inter-project similarity tokens introducing local component constructing as well project references, enhance Experiments show each characteristic exploit bring substantial improvement our integrated framework outperforms baseline frameworks. Compared base model, token-level optimal model relatively improves all-token identifier accuracy by 7.68% 19.96%, respectively; line-level it edit-distance exact-match metrics 8.89% 22.82%, respectively. Moreover, perform error analysis point out potential directions future work.

Language: Английский

---

TestDossier: A Dataset of Tested Values Automatically Extracted from Test Execution

Published: April 15, 2024

Language: Английский

---

Method-Level Test-to-Code Traceability Link Construction by Semantic Correlation Learning

IEEE Transactions on Software Engineering, Journal Year: 2024, Volume and Issue: 50(10), P. 2656 - 2676

Published: Aug. 27, 2024

Language: Английский

---

AgoneTest: Automated creation and assessment of Unit tests leveraging Large Language Models

Published: Oct. 18, 2024

Language: Английский

---

Combining Coverage and Expert Features with Semantic Representation for Coincidental Correctness Detection

Published: Oct. 18, 2024

Language: Английский

---