Status

Active.

The Big Bang project aims to create a typed programming language with the feel and convenience of modern scripting languages. Projects such as DRuby and Typed Racket retrofit type systems onto existing scripting languages; unfortunately, these projects are burdened by backwards compatibility, as dynamic scripting languages are developed without regard to the static typeability of the features they include. Big Bang re-examines the design of scripting-style languages from a static typing perspective. In particular, we prioritize:

No explicit typing

One primary appeal of scripting languages is that programmers need not explicitly specify types, even at module or function boundaries.

Global type inference

The type system should not be brittle to refactorings such as method extraction.

Performance

Using static type information, we contend that it will be possible to compile Big Bang scripts to much more efficient programs than with scripting languages such as Python or Ruby.

Path sensitivity

Scripting programmers often use path-sensitive reasoning when confirming that a program is sound; the Big Bang type system does as well.

No arbitrary cutoffs

Polymorphism in Big Bang does not rely on fixed cutoffs or scoping limitations; this ensures that the programmer will not be surprised by arbitrary compiler decisions.

The Big Bang project is still in a fledgling state. For questions or comments, please contact big-bang@jhu.edu.

Publications

  • Pottayil Harisanker Menon, Zachary Palmer, Alexander Rozenshteyn, Scott Smith. Types for Flexible Objects, Asian Symposium on Programming Languages and Systems. Singapore, November 2014. (Draft, March 2014. Draft, January 2014. Draft, May 2013.)

    Abstract

    Scripting languages are popular in part due to their extremely flexible objects. Features such as dynamic extension, mixins, and first-class messages improve programmability and lead to concise code. But attempts to statically type these features have met with limited success. Here we present TinyBang, a small typed language in which flexible object operations can be encoded. We illustrate this flexibility by solving an open problem in OO literature: we give an encoding where objects can be extended after being messaged without compromising the expressiveness of subtyping. TinyBang’s subtype constraint system ensures that all types are completely inferred; there are no data declarations or type annotations. We formalize TinyBang and prove the type system is sound and decidable; all examples in the paper run in our most recent implementation.

  • Pottayil Harisanker Menon, Zachary Palmer, Alexander Rozenshteyn, Scott Smith. PatBang: Flexible type-safe pattern matching, technical report. Last updated September 2013.

    Abstract

    Patterns provide an important dimension of expressiveness to functional programming languages because they describe a concise syntax for data destruction. However, most languages treat patterns as second-class citizens: case match expressions cannot be extended, patterns cannot be selected dynamically, and patterns cannot be composed or modified by program logic. In this paper, we present a rich, full-featured pattern language, PatBang, which treats patterns as first-class data, and additionally supports highly expressive patterns including recursive and disjunctive patterns, yet is still provably type-safe. We additionally show how such an expressive pattern language enables new programming patterns, including use of patterns for expressing pytes , a form of lightweight static interface declaration. Type soundness is proven and an implementation of the type inference algorithm and interpreter is provided.

  • Pottayil Harisanker Menon, Zachary Palmer, Alexander Rozenshteyn, Scott Smith. A Practical, Typed Variant Object Model, Foundations of Object-Oriented Languages. Tucson, AZ, USA, October 2012. Slides.

    Abstract

    Traditionally, typed objects have been encoded as records; the fields and methods of an object are stored in the fields of a record and projected when needed. While the dual approach of using variants instead of records has been explored, it is more challenging to type: the output type of a variant case match must depend on the input value; this is a form of dependent typing.

    In this paper, we construct a variant-based encoding of objects which is statically typeable and which improves on the flexibility of typed object models in several dimensions: messages can be represented as simple first-class data, object extension is more generally typeable than in previous systems and, arguably, a better integration of objects and functions is obtained.

    This encoding is possible due to the features of our new core language, TinyBang, which incorporates a subtype constraint type inference system with several novel extensions. We develop a generalized notion of first-class cases – functions with an inherent pattern match that are composable – and we extend previous notions of conditional constraint types to obtain accurate typings. For added flexibility, TinyBang’s record-like structure is type-indexed, meaning data can be projected based on its type alone. Our record structure’s concatenation operator is asymmetric by default, naturally supporting object extension. Finally, we develop a refined notion of parametric polymorphism which aims to achieve a good combi- nation of flexibility and efficiency of inference.

  • Pottayil Harisanker Menon, Zachary Palmer, Alexander Rozenshteyn, Scott Smith. Big Bang: Designing a Statically-Typed Scripting Language, International Workshop on Scripts to Programs. Beijing, China, June 2012. Slides.

Dissertations