Skip to main content

Bibliography

rascal-0.40.17

This bibliography is 🚧under construction🚧; we started collecting the data in May 2023 and plan to complete it around December 2023. PDF links and Bibtex snippets will be added later and many references are still missing.

The literature about Rascal goes back all the way to 1980's when its predecessor ASF+SDF spawned. This page intends to provide a comprehensive list of papers, PhD theses and masters' theses about what is happening "under the hood", on the language level and with the applications of Rascal.

The bibliography is organized in batches of decades (1980's, 1990's, 2000's, 2010's, 2020's), and grouped by the following three categories:

  • Reports on the Applications of Meta Programming languages and systems.
  • All Meta Languages and Systems; their design, construction, and evaluation.
  • The Data Structures and Algorithms which are fundamental to the above.

Applications​

Work that applies the meta-technology and theories is scattered over several sub-disciplines of Computer Science; e.g. Programming Languages, Software Engineering, Reverse Engineering, Re-engineering, Refactoring, Domain-specific Languages, Model Driven Engineering, and System Security.

2020's​

  • 🎓 Lina María Ochoa Venegas, Break the Code? Breaking Changes and their Impact on Software Evolution, (2023) TU Eindhoven.

    In this thesis Rascal was applied to detect breaking changes in Java libraries and predict their impact in client projects. The Maracas tool reuses japicmp and the JDT M3 bridge for extracting detailed static information from source code and bytecode.

  • 🎓Tim Soethout, Banking on Domain Knowledge for Faster Transactions --- Leveraging Models to Avoid Coordination, (2022) TU Eindhoven.

    This thesis applies Rascal for Model Driven Engineering. Detailed communication protocols are generated from high-level specification. The code that is generated implements algorithms that use domain knowledge to avoid locking in distributed communication systems.

  • Mathijs T. W. Schuts, Rodin T. A. Aarssen, Paul M. Tielemans, and Jurgen J. Vinju. Large-scale semi-automated migration of legacy C/C++ test code. (2022). Software: Practise & Experience

    In this paper the engineers at Philips Healthcare used Rascal to automatically refactor a large body of C++ test code to a more modern test harnass by Google. This work is based on the clair front-end for C++ parsing from Rascal based on the Eclipse CDT.

  • Jouke Stoel, Tijs van der Storm and Jurgen J. Vinju. Modeling with Mocking, 14th IEEE Conference on Software Testing, Verification and Validation

    The Rebel2 language is implemented in Rascal; it is a contract-based event-sourcing language based on communicating state machines, with mappings to SMT solvers for simulation and verifation purposes. We use meta-programming to filter the specifications towards configurations with state spaces that are still practical. A collaboration with ING.

  • Tim Soethout, Tijs van der Storm, Jurgen J. Vinju. Contract-Based Return-Value Commutativity: Safely exploiting contract-based commutativity for faster serializable transactions. (2021). AGERE 2021.

    This is about one of the code generator backends for Rebel2 (see above). By re-defining the concept of serializability in terms of equal data-dependent post-conditions of a transaction, the communication protocol we generate can be shown to have strong isolation properties. This is a collaboration with ING.

  • Tim Soethout, Jurgen J. Vinju and Tijs van der Storm. Path-Sensitive Atomic Commit - Local Coordination Avoidance for Distributed Transactions. (2021) <Programming> Journal

    This is about an advanced communication protocol that is generated for Rebel2 (see above), the protocol uses domain knowledge specified in Rebel to avoid unnecessary blocking synchronization with a big benefit for throughput in situations with high congestion. This is a collaboration with ING.

2010's​

  • Tim Soethout, Jurgen J. Vinju and Tijs van der Storm. Static Local Coordination Avoidance for Distributed Objects. (2019) SPLASH AGERE.

    This is an optimized (statically detected) version of the above PSAC protocol, where we use Rascal to generate communication protocols that do not need any synchronization to achieve isolation guarantees (serializability).

  • Lina Ochoa, Thomas Degueule and Jurgen J. Vinju. An Empirical Evaluation of OSGi Dependencies Best Practices in the Eclipse IDE (2018) IEEE International Conference on Mining Software Repositories (MSR)

    An ecosystem-level analysis of actual dependency management practises in code and meta-data.

  • Davy Landman, Alexander Serebrenik and Jurgen J. Vinju. Challenges for Static Analysis of Java Reflection – Literature Review and Empirical Study. (2017).. IEEE/ACM International Conference on Software Engineering ICSE 2017. ACM SIGSOFT Distinguished Paper Award.

    In this paper we used Rascal to dig deep into client code of the Java reflection API. We analyze the corpus for typical usage patterns in annotated syntax trees and also patterns that can be associated with the limits of static analysis tools. A typical application of the M3 intermediate fact representation model.

  • 🎓 Davy Landman, Reverse Engineering Source Code. (2017) Universiteit van Amsterdam.

    We explore the limits of what can be learned purely from source code analysis. In this thesis Rascal is used for both shallow and deep code analysis (metrics and static analysis of reflective code). By mining large open-source software repositories, we discover that the cyclomatic complexity metric can not be shown to correlate with source lines of code, and also we discover the destructive impact of reflection on the accuracy of static analysis tools, and what to do about that.

  • Davy Landman, Alexander Serebrenik, Eric Bouwers and Jurgen J. Vinju. Empirical analysis of the relationship between CC and SLOC in a large corpus of Java methods and C functions. (2016) Journal of Software: Evolution and Process.

    Trying extremely hard to corroborate the reports on Pearson correlation between Cyclomatic Complexity and Source Lines of Code in C functions and Java methods, we have to conclude that on a large and unbiased corpus there is only low correlation. Higher correlations reported in literature can be explained by data transformation techniques (log transform) and aggregation methods (addition over code modules). SLOC DOES NOT CORRELATE LINEARLY TO CC ACCORDING TO OUR RESULTS

Meta languages and systems​

This literature is about the meta systems themselves, their design and their evaluation.

2020's​

  • 🎓Mauricio Merino Verano, Engineering Language-Parametric End-User Programming Environments for DSLs (2022) TU Eindhoven.

    Inspired by industrial use cases at Canon Producting Printing, this thesis adds "Notebooks" and "Block-based languages" to the Rascal technological space. Generating notebook interfaces for DSLs, like IDEs, and deriving usable block-based language interfaces from context-free grammars are two highlights.

  • Mauricio Verano Merino, Tom Beckmann, Tijs van der Storm, Robert Hirschfeld, and Jurgen J. Vinju. Getting Grammars into Shape for Block-based. (2021) International Conference on Software Language Engineering.

    We can generate block-based syntax editors from context-free grammars. The paper bridges between the grammarware and blockware technological spaces. A collaboration with TU Eindhoven en Canon Production Printing.

  • Jurgen J. Vinju and Tijs van der Storm. Bacatá: Notebooks for DSLs, Almost for Free. (2020) <Programming> Journal and DSLDI 2020

    We can generate Jupyter Notebook interfaces for language descriptions in Rascal, similar to the way we can generate Eclipse plugins and VScode extensions, but with a very different user experience.

2010's​

  • Jouke Stoel, Tijs van der Storm, Jurgen J. Vinju. AlleAlle: Bounded Relational Model Finding with Unbounded Data. (2019) SPLASH Onward!

    AlleAlle is an (intermediate) domain specific language for expressing constraints on relations with data. It can be seen as as generalized version of Alloy. It is implemented in Rascal and uses Z3, and serves as a high-level back-end for DSL analysis tools.

  • Rodin Aarssen, Tijs van der Storm, Jurgen J. Vinju. Concrete Syntax with Black Box Parsers. (2019) <Programming> Journal.

    We add concrete syntax for parsers not generated from context-free grammars in Rascal, but rather coupled external (open) parsers for existing programming languages. This is in collaboration with TU Eindhoven and Philips Healthcare.

  • Paul Klint, Tijs van der Storm and Jurgen J. Vinju. Rascal, 10 years later. (2019) Invited paper after the IEEE SCAM Most Influential Paper award for Rascal: A Domain Specific Language for Source Code Analysis and Manipulation

    This paper reports on 10 years of Rascal development and application in research, education and industry.

  • Paul Klint, Tijs van der Storm, and Jurgen Vinju. EASY Meta-programming with Rascal. Leveraging the Extract-Analyze-Synthesize Paradigm for Meta-programming. (2010) In Proceedings of the 3rd International Summer School on Generative and Transformational Techniques in Software Engineering (GTTSE'09), LNCS. Springer.

    This is a comprehensive overview of the first release of Rascal as it was used in education at the Master Software Engineering for the first time in 2009, and at the summer school in Braga that year. Note that the language has evolved quite a bit since then, so as a manual it is not apropriate anymore. The paper shows how the transformation and analysis paradigms have been merged into a single meta-language. You can find elements of ASF, SDF, Box, ATerms, ApiGen, RScript, and other design elements in this paper. This paper is often cited as the origin paper for Rascal. The other origin paper was in IEEE SCAM 2009 (see below).

2000's​

  • Paul Klint, Tijs van der Storm, and Jurgen J. Vinju. Rascal: A Domain Specific Language for Source Code Analysis and Manipulation. (2009) In Ninth IEEE International Working Conference on Source Code Analysis and Manipulation, SCAM 2009, Edmonton, Alberta, Canada, September 20-21, 2009, pages 168-177. IEEE Computer Society.

    This paper motivates and positions Rascal in the domain of source code analysis and manipulation by introducing it along the running example of the "infer generic type arguments" refactoring. Since refactoring requires both analysis and transformation it is an ideal vehicle for motivation and demonstration of Rascal's language-integrated meta programming primitives. This paper is often cited as an origin paper for Rascal. It won the 2019 most influential paper award from IEEE SCAM, due to the observed impact of Rascal in education, research and industry.

  • Jurgen J. Vinju and J.R. Cordy. How to make a bridge between transformation and analysis technologies? (2006) In J.R. Cordy, R. Lämmel, and A. Winter, editors, Transformation Techniques in Software Engineering, number 05161 in Dagstuhl Seminar Proceedings. Internationales Begegnungs- und Forschungszentrum (IBFI), Schloss Dagstuhl, Germany.

    This technical report explains the relation between the code transformation and code analysis domains from a perspective of tool builders. From here several main motivations for the design of Rascal were derived.

  • Philippe Charles, Robert M. Fuhrer, Stanley M. Sutton Jr., Evelyn Duesterwald, and Jurgen Vinju. Accelerating the Creation of Customized, Language-specific IDEs in Eclipse. (2009) Proceedings of the 24th Annual ACM SIGPLAN Conference on Object-Oriented Programming, Systems, Languages, and Applications, OOPSLA 2009, October 25-29, 2009, Orlando, Florida, USA., pages 191-206, 2009

    This paper by IBM research introduces Eclipse IMP (IDE meta-tooling Platform) which had been designed to offer a programmeable interface to Eclipse in terms of language properties rather than UI interactions. It resembles the ASF+SDF Meta-Environments communication protocol quite a bit. With some minor generalizations we were able to port the entire ASF+SDF Meta-Environment to Eclipse IMP within a few weeks. Later the Rascal Eclipse environment was also built directly on top of IMP. Others also had benefit from IMP, like the Eclipse-based environment for the X10 language, the Spoofax lanuage workbench and the Hege language.

1980's​

  • J. Coutaz, J. The Box, A Layout Abstraction for User Interface Toolkits. (1984) Pittsburgh, PA: Carnegie Mellon University.

    This technical report from CMU captures the domain of two dimensional layouts, including pretty printing in an elegant, functionally complete combinator language. Later this Box abstraction would form the basic layer for all pretty printing and formatting utilities in ASF+SDF Meta-Environment, and Rascal. It also ended up in other meta programming systems such as StrategoXT, SPoofax and DMS.

  • Paul Klint. An overview of the SUMMER programming language (1980) POPL '80: Proceedings of the 7th ACM SIGPLAN-SIGACT symposium on Principles of programming languagesJanuary

    Summer was a source code analysis language with a wide array of pattern matching facilities, just as Rascal has. The meta notation used in this paper was later the inspiration for much of the SDF notation. The move towards algebraic specification later during the ESPRIT GIPE projects focused the pattern matching on trees instead of strings. String matching came back in Rascal later only in 2009.

Algorithms and Data-structures​

Here we find the algorithms and data-structures that enable the meta languages to work, to be efficient and scalable or generalizable. Sometimes the tools are intrically linked to the meta-language's designs; we list those contributions here rather than above.

2020's​

  • Jurgen Vinju. Comparing Bottom-up with Top-down Parsing Architectures for the Syntax Definition Formalism from a Disambiguation standpoint. (2023) OASIcs, Volume 109, EVCS: Eelco Visser's Commemorative Symposium

    This paper motivates why Rascal moved away from scannerless GLR and to scannerless GLL, and why data-dependent context-free grammars are the next step.

2010's​

  • 🎓 Ali Afroozeh and Anastasia Izmaylova, Practical General Top-down Parsers. (2019) Universiteit van Amsterdam.

    This double thesis explores the GLL parsing algorithm as well as parsing combinators for adding context information to parsing in a principled manner with the aim of finding practical solutions to the disambiguation problem. The result is the elegant use of data-dependent context-free grammars to desugar high-level disambiguation constructor to (such as operator precedence and the offside rule).

  • 🎓 Naveneetha K. Vasudevan, A non-deterministic approach to ambiguity detection in Context Free Grammars. (2017)Department of Informatics King’s College London.

    This thesis from outside of the Rascal community provides very effective ambiguity detection with a practical and easy-to-implement approach. This knowledge is now the basis of Dr Ambiguity's ambiguity detection methods.

  • Michael Steindorfer and Jurgen J. Vinju. To-many or To-one? All-in-one! Efficient Purely Functional Multi-Maps with Type-Heterogeneous Hash-Tries. (2018) ACM International Conference on Programming Languages Design and Implementation (PLDI).

    Here we generalize the hash-trie data-structure implementation of the JVM to efficient multi-maps for fast indexed many-to-many relations, as used in relational analysis of source code with Rascal. The data-structure optimizes for the general and statistically normal one-to-one case and degrades in memory and iteration efficiency gradually as more one-to-many tuples are added.

  • 🎓 Michael Steindorfer, Efficient Immutable Collections. (2017) Universiteit van Amsterdam.

    This thesis describes the core data-structures and optimizations required for Rascal's implementation of relational calculus: relations, sets and maps. The trie map is the core vehicle, and the contributions found their way outside of the Rascal community as well.

  • 🎓 Bas Basten, Ambiguity Detection for Programming Language Grammars. (2011) Universiteit van Amsterdam.

    In this thesis ambiguity of context-free grammars is a problem, which must be detected, diagnosed and cured. Static ambiguity detection in this thesis takes the approach of trying to prove unambiguity and reporting failures with causes if such a proof can not be found. There is also a chapter on the diagnostics and curing of ambiguities once found.

  • Bas Basten and Jurgen Vinju. Parse forest diagnostics with Dr. Ambiguity. (2011) ACM International Conference on Software Language Engineering (SLE).

    This paper explains how differential analysis can isolate the cause of ambiguities; in particular we focus on the kind of ambiguities that can be solved with declarative disambiguation constructs. The algorithm uses the essence of the disambiguation constructs semantics (which characterize a difference between two comparable trees), as the vehicle for detecting the causes of ambiguity. That they can be solved then, falls out trivially.

2000's​

  • Giorgios Economopoulos, Paul Klint, and Jurgen J. Vinju. Faster scannerless GLR parsing. (2009) Compiler Construction, 18th International Conference.

    Here we merge RNGLR with SGLR to obtain a faster and simpler implementation that works more predictably. This also rationalized the implementation of the reject filter and as a side-effect we optimized the garbage collection of parallel parse stack nodes and edges.

  • Elizabeth Scott and Adrian Johnstone. Right nulled GLR parsers (2006), ACM Transactions on Programming Languages and Systems 28(4):577-618

    This paper proposes a simpler and more efficient version of Tomita's GLR than the versions of Farshi and Rekers, and so later we merged this into RNSGLR (see above).

  • Mark van den Brand, Hayco de Jong, Paul Klint, Pieter A. Olivier. Efficient annotated terms. Softw. Pract. Exp. 30(3): 259-291 (2000)

    This paper introduces the C library on annotated maximally-shared algebraic terms that would not only become the basis of ASF+SDF Meta-Environment, but also of StrategoXT, ELAN4 and mCRL2. Later the vallang library replaced it, but most of its design principles were kept in tact.

1990's​

  • 🛠 Jurgen Vinju, Optimizations of List Matching in the ASF+SDF compiler, University of Amsterdam (1999).

    Making list matching faster made almost everything in ASF+SDF faster because it is a core feature.

  • 🎓Eelco Visser, Syntax Definition for Language Prototyping, University of Amsterdam. (1997)

    This thesis re-designs the Syntax Definition Formalism (SDF) from scratch and removes the need for a scanner. The scannerless parsing algorithm SGLR contains the disambiguation filters required to make scannerless parsing possible and feasible.

  • 🛠Eelco Visser, Combinatory Algebraic Specification & Compilation of List Matching, (1993)

    Explorations of matching "modulo" theories (list matching) for more efficient term rewriting.

  • 🎓Jan Rekers, Parser Generation for Interactive Environments, University of Amsterdam. (1992)

    This thesis modifies Tomita's GLR algorithm to fix some issues but also makes it incremental and work on substrings for error recovery purposes.

1980's​

  • M. Tomita. Efficient Parsing for Natural Languages. (1985) A Fast Algorithm for Practical Systems. Kluwer Academic Publishers.

    Tomita's GLR parsing algorithm provides non-deterministic parsing efficiently; the effect is that it basically has limitless lookahead, and provides an implementation layer for modular/compositional grammars. More importantly, since most existing languages do not fit into the LR or LR categories, the GLR algorithm provided a way for ASF+SDF to scale from toy to real programming languages such as Pascal, COBOL and C. We include it here because this paper informed almost all of the succeeding work in parsing of this community.