Knowledge-Based Environment Dependency Inference for Python Programs

Besides third-party packages, the Python interpreter and system libraries are also critical dependencies of a Python program. In our empirical study, 34% programs are only compatible with specific Python interpreter versions, and 24% programs require specific system libraries. However, existing tech...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2022 IEEE/ACM 44th International Conference on Software Engineering (ICSE) S. 1245 - 1256
Hauptverfasser: Ye, Hongjie, Chen, Wei, Dou, Wensheng, Wu, Guoquan, Wei, Jun
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: ACM 01.05.2022
Schlagworte:
Data mining
environment dependency inference
Feature extraction
Knowledge based systems
knowledge graph
Libraries
Open source software
Python
Runtime
version constraint
ISSN:1558-1225
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Besides third-party packages, the Python interpreter and system libraries are also critical dependencies of a Python program. In our empirical study, 34% programs are only compatible with specific Python interpreter versions, and 24% programs require specific system libraries. However, existing techniques mainly focus on inferring third-party package dependencies. Therefore, they can lack other necessary dependencies and violate version constraints, thus resulting in program build failures and runtime errors. This paper proposes a knowledge-based technique named PyEGo, which can automatically infer dependencies of third-party packages, the Python interpreter, and system libraries at compatible versions for Python programs. We first construct the dependency knowl-edge graph PyKG, which can portray the relations and constraints among third-party packages, the Python interpreter, and system libraries. Then, by querying PyKG with extracted program features, PyEGo constructs a program-related sub-graph with dependency candidates of the three types. It finally outputs the latest compatible dependency versions by solving constraints in the sub-graph. We evaluate PyEGo on 2,891 single-file Python gists, 100 open-source Python projects and 4,836 jupyter notebooks. The experimental re-sults show that PyEGo achieves better accuracy, 0.2x to 3.5x higher than the state-of-the-art approaches.
ISSN:1558-1225
DOI:10.1145/3510003.3510127