Contributions to Declarative Implementation of Static Program Analysis

Forskningsoutput: AvhandlingDoktorsavhandling (sammanläggning)

Abstract

Programming languages are ever evolving, with new languages being invented to solve new problems, and old languages being extended to solve old problems in new ways. With the continued evolution of programming languages, and with new and improved static program analyses, we need flexible systems for building our static analyses and compilers. Declarative programming with Reference Attribute Grammars (RAGs) is a fruitful approach for building extensible and maintainable static program analyses and compilers. With declarative programming, code is easier to write, easier to understand, and easier to extend.

This thesis presents contributions to declarative static program analysis implementation with RAGs. In particular, I have developed new language extensions for the Java programming language, as well as new static analyses and tools for Java, based on the extensible Java compiler ExtendJ. Language extensions include an implementation of Java 7, and a new programming mechanism, multiplicities. A new static analysis-based tool presented in this thesis is a regression test selection tool, which reduces testing time for Java software development.
My contributions also include new design patterns for declaratively specifying static analyses in RAGs, and significant improvements to the ExtendJ compiler. My work in ExtendJ includes developing the first version of the compiler with fully declarative static analysis.
This dissertation also includes contributions to RAGs, including new algorithms for concurrent attribute evaluation with implementation in the JastAdd metacompiler. With these new algorithms, it is possible to parallelize any RAG-implemented analysis. In particular, I parallelized static analysis in the ExtendJ compiler to achieve a twofold speedup and orders of magnitude reduction of latency.

Detaljer

Författare
Enheter & grupper
Forskningsområden

Ämnesklassifikation (UKÄ) – OBLIGATORISK

  • Datavetenskap (datalogi)
Originalspråkengelska
KvalifikationDoktor
Tilldelande institution
Handledare/Biträdande handledare
Tilldelningsdatum2019 jan 18
UtgivningsortLund
Förlag
  • Department of Computer Science, Lund University
Tryckta ISBN978-91-7753-944-5
Elektroniska ISBN978-91-7753-945-2
StatusPublished - 2018 dec 19
PublikationskategoriForskning

Nedladdningar

Ingen tillgänglig data

Relaterad forskningsoutput

Jesper Öqvist & Görel Hedin, 2017 okt 23, Proceedings of the 10th ACM SIGPLAN International Conference on Software Language Engineering. ACM, s. 151-162 12 s.

Forskningsoutput: Kapitel i bok/rapport/Conference proceedingKonferenspaper i proceeding

Jesper Öqvist & Görel Hedin, 2017 okt 18, Department of Computer Science, Lund University. 20 s. (Technical report, LU-CS-TR; vol. 2017-254, nr. Report 103)

Forskningsoutput: Bok/rapportRapport

Jesper Öqvist, Görel Hedin & Boris Magnusson, 2016 aug 29, Proceedings of the 13th International Conference on Principles and Practices of Programming on the Java Platform: Virtual Machines, Languages, and Tools. ACM, 5

Forskningsoutput: Kapitel i bok/rapport/Conference proceedingKonferenspaper i proceeding

Visa alla (5)

Related projects

Visa alla (2)