Advanced Programming Language Technology for Reflective, Dynamic, Adaptive Software

The Problem: Software produced with current technology is typically opaque and brittle. That is, applications adapt poorly to changes in environment (OS, hardware, data profile) or intended use, scale poorly, and are difficult to debug, maintain, and enhance. As the scale and complexity of applications attempted by the software development community as a whole and by the artificial intelligence research community in particular grow, software needs to be developed with the ability to adapt to a complex, dynamic environment. Motivation: We are developing an infrastructure to support the addition of advanced language features to the toolset available to programmers. We are focusing on: • Reflection, dynamism, and metaprogramming: Reflection refers to the ability of an application to introspect, at runtime, on its internal state, both data and control. Examples include examining the class hierarchy at runtime, or using runtime profile information to find performance bottlenecks. Based on analysis of data gathered using runtime reflection, an application may want to dynamically adapt its behavior or internal structure – this is what we refer to as dynamism. Reflection and dynamism enable a form of programming called metaprogramming: writing programs that reason about and modify the behavior of themselves and/or other programs. • Syntactic extension: Based on ideas from the Lisp and Dylan programming languages, we are designing a powerful procedural macro facility for Java