Actions

Summer Of Code 2009

From Gambit wiki

Revision as of 05:56, 15 March 2009 by Pflanze (talk | contribs) (grammatical fixes)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Please check this page regularly as we will be adding new ideas as they come up.

Note that this page is editable. If you have suggestions for projects please add them to the second section ("project suggestions"). The first section contains the projects that are officially proposed by the organization. You can make corrections to the text but please don't make substantial changes.

The Gambit wish list may also be a good source of inspiration for projects. However the wish list contains projects which require a very good knowledge of the Gambit system's structure and internal operation, so they are not advisable for students with little or no experience with Gambit.

Official project proposals

Documentation framework

Gambit's documentation is mainly in the form of a traditional user manual. To increase its usefulness it needs to be easier to access while developing code. Documentation needs to be integrated with the tools including: from the text editor (pop-up documentation when a call to a built-in function is typed, with information on parameters), from the REPL, from the command-line. Moreover the documentation should be automatically checked for consistency on each release (for instance by checking that the code examples actually produce the result indicated).

Difficulty: easy

Required skills: Scheme, Gambit, and possibly LaTeX, HTML

Gambit packages

There is an urgent need for easily reusable packages for miscellaneous features including data-structures, message digests, encryption, pattern-matching, regular expressions, etc. Portable Scheme code already exists for many of these (e.g. SRFI's, Snow packages, Chicken eggs, etc). The project consists in porting these codes to the Gambit system and to write new packages when appropriate. We expect the student to choose the set of packages to port/write based on their personal interests. Due to the high number of packages that need to be worked on, this project can accomodate up to 3 students. If more than one student participates on this project, the mentor will help to coordinate the work of the students.

Difficulty: easy to medium

Required skills: Scheme, Gambit, algorithms

SLIME interface

SLIME (http://common-lisp.net/project/slime/) is a full-featured Emacs mode for multithreaded Common Lisp development. A port of it exists for Scheme48 ("SLIME48"). The project consists in adapting SLIME to Gambit to allow multithreaded remote debugging from Emacs. The current gambit.el emacs-lisp package could be used as a starting point.

Difficulty: medium

Required skills: emacs-lisp, Scheme, Gambit, multithreading

Web framework

Thanks to Gambit's high-level networking functions, non-blocking I/O, and efficient thread system, a simple high-performance web server can be written in a few minutes. With Gambit's serializable closures and continuations it is possible to implement cleanly a fancy web framework with advanced features (client-side computation, task migration, load balancing, persistent interactions, undoable transactions). The project consists in constructing the framework to support these advanced features, and to demonstrate a few of them).

Difficulty: hard

Required skills: HTML, HTTP, Scheme, Gambit

Native Threading

Gambit (like most Scheme systems) implements its own lightweight threading system which is not aware of native, operating system threads. The so-called Gambit Virtual Machine (GVM) can only be bound to one native thread. This makes it difficult to utilize true parallel processing, which will be an important step in adapting to future hardware. This also causes C code to block the entire GVM for the duration of the C code's execution.

One solution is to use Termite and spawn multiple Gambit processes. Although it is a valid solution, it is unacceptable for any real-time system such as video games. This project aims to extend Gambit in some fashion to utilize system threads. Various solutions and assumptions will be explored: spawn multiple Gambit systems across threads in one process and have a special shared memory buffer? Extend Gambit to enforce pure functional programming?

Difficulty: hard

Required skills: Gambit, threading, memory allocation, garbage collection

GVM virtual-machine interpreter

The Gambit system is based on a virtual machine (the GVM). It is the intermediate representation for the compiler. The size of the generated code is an issue for very large programs (such as the Jedi IDE) and for embedded systems. An interesting approach, to be explored by this project, would be to design a compact bytecode representation for the GVM and implement a bytecode interpreter.

Difficulty: hard

Required skills: Scheme, Gambit, compilation

Module System Infrastructure

The Gambit system adhere's to the philosophy of providing basic, robust and efficient infrastructure on top of which users can implement higher-level abstractions. This project consists in designing and implementing a module system infrastructure for Gambit with which more sophisticated module systems can be implemented. In particular it should be possible to use the result of this project to implement the R6RS module system.

Difficulty: hard

Required skills: Gambit, functional programming, metaprogramming, OO programming

Automating FFI Bindings

Gambit's foreign-function-interface (FFI) can be tedious to use for interfacing to C libraries with many functions (such as Xlib, OpenGL). There exist some tools to parse header files and extract information relevant to the creation of FFI bindings. For example, SWIG (http://www.swig.org/) is an interface generator that supports several high-level programming languages, including a few Schemes (Chicken, MzScheme, Guile). There is also gcc-xml which is being used in the Ctypes Python library. The goal would be to get one of these tools working with Gambit. The Gambit FFI subsystem will also have to be extended minimally to support features such as "const" types.

Difficulty: medium

Required skills: Gambit, C/C++

Project suggestions

Add to this section any projects you want to suggest to the organizers. Acceptable projects will be moved to the above section.

Flash bytecode / ActionScript back-end

This project will explore ways to compile Scheme code into either Bytecode targeted for the Flash Player, or alternatively ActionScript, and write a prototype back-end for the Gambit compiler.

Difficulty: hard

Required skills: Scheme, Gambit, compilation

LLVM back-end

A working prototype generating LLVM code (intermediate representation) has been written by two students as part of a term project. Going from this prototype, a back-end covering all of Scheme and Gambit's extensions can be created. This will allow Gambit to take advantage of the LLVM infrastructure, and also allow to compile code without requiring a complete C compilation environment to be installed on the user's machine.

Difficulty: hard

Required skills: Scheme, Gambit, compilation