An open compiler front end framework for
building Java language extensions

mailing lists

Polyglot is a highly extensible compiler front end for the Java programming language. It is implemented as a Java class framework using design patterns to promote extensibility. Using Polyglot, language extensions can be implemented without duplicating code from the framework itself. Polyglot has been used to implement domain-specific languages, to explore language design ideas, to simplify Java for pedagogical purposes, and for various code transformations such as optimization and fault injection. Polyglot has been used for both major and minor language extensions; our experience suggests that the cost of implementing an extension scales well with the degree to which it modifies Java.

Polyglot compiles and runs on (at least) Linux, Solaris, Windows, and Mac OS X. Ant and the JFlex scanner generator are required to build it. On Windows, you will also need Cygwin.

Polyglot supports Java 1.4, Java 5, and Java 7, including features such as generics and annotations. Key features of Java 8 are also supported: notably, lambdas. Support for Java 5, 7, and 8 is provided as Polyglot extensions, but other Polyglot extensions can build on any version of the language, and can generate any version of Java as output.

Polyglot includes PPG, an extensible LALR parser generator based on the CUP LALR parser generator for Java (extended with improved debugging support). The Accrue analysis framework (from Harvard) extends Polyglot with interprocedural, object-sensitive program analysis.

Some Polyglot-based projects

  • Jif – information flow and program transformations for security
  • JMatch – abstract iterable pattern matching for Java
  • Jx/J& –  nested inheritance for extensible, composable frameworks.
  • J0 – a Java for novice programmers
  • Soot – a Java optimization framework (McGill)
  • X10 –  the concurrent programming language (IBM)
  • Jedd – a BDD-based relational extension of Java (McGill)
  • abc – an AspectJ compiler (Oxford, McGill)
  • JPred – practical predicate dispatch (UCLA)
  • LazyJ  – seamless lazy evaluation in Java (UCLA)
  • XJ –  integrating XML processing into Java (IBM)
  • J-LO – the Java Logical Observer, for runtime verification (Aachen)
  • DJ – adding type-safe higher-order code mobility to Java (Imperial)
  • HydroJ – object-oriented pattern matching for evolvable distributed systems (U Washington)
  • AtomJava – a novel approach to atomic blocks (U Washington)
  • Classages – interaction-based programming (Johns Hopkins)
  • jCilk – a multi-threaded programming language (MIT)
  • DeepJava – multi-level programming in Java (Victoria University, TU Darmstadt)
  • SessionJ – session-based distributed programming in Java (Imperial)
  • IMP –  Eclipse IDE Meta-tooling Platform (Eclipse, IBM TJ Watson Hawthorne)
  • JCoBox –  generalizing active objects to concurrent cooperative tasks (TU Kaiserslautern)


Download Polyglot

[Change log]

Polyglot 2.x

[Show all] [Change log]

Polyglot 1.x

[Show all]
  • – 21 Feb 2008
  • Previous versions:
    • – 4 Aug 2006
    • – 1 May 2006 (fixed broken javacup tarball)
    • – 6 Jul 2005
    • – 24 Mar 2005
    • – 30 Jul 2004
    • – April 2004


Mailing lists

Email address:
(optional) Your name:
(required) Set Your Password:
Verify Your Password:
    Subscribe to polyglot-users-l@cornell.edu to be notified of new releases of Polyglot, to participate in discussions about Polyglot and to ask and answer questions about Polyglot. The developers do monitor the list: bug reports are welcome; bug fixes are even more welcome. To subscribe, send mail to polyglot-users-l-request@cornell.edu with a subject line of join, or use the form on the right.

A partial family tree of Polyglot extensions

Polyglot developers

The Polyglot project is currently supervised by:

The current Polyglot developers can be contacted at polyglot-developers-L@cornell.edu. Contributions can also be made via the GitHub repo.

Many other people have contributed to Polyglot, including but not limited to:

  • Owen Arden
  • Michael Brukman
  • Michael Clarkson
  • Matt Harren
  • Chinawat Isradisaikul
  • Rob Jellinek
  • Aleksey Kliger
  • Daniel Lee
  • Samarth Lad
  • Jed Liu
  • Todd Millstein
  • Xin Qi
  • Naveen Sastry
  • Dan Spoonhower
  • Milan Stanojevic
  • Steve Zdancewic
  • Lantian Zheng
  • Xin Zheng
  • Sam Zhou

Development of Polyglot has been supported by several funding sources, including DARPA Contract F30602-99-1-0533, monitored by USAF Rome Laboratory, ONR Grant N00014-01-1-0968, NSF Grants CNS-0208642, CNS-0430161, and CCF-0133302, an Alfred P. Sloan Research Fellowship, and an Intel Research Ph.D. Fellowship.