Sunday, June 20, 2010

Refining Syntactic Sugar

Another great dissertation just came out, Ryan Culpepper's Refining Syntactic Sugar.
Over the past two decades, Scheme macros have evolved into a powerful API for the compiler front-end. Like Lisp macros, their predecessors, Scheme macros expand source programs into a small core language; unlike Lisp systems, Scheme macro expanders preserve lexical scoping, and advanced Scheme macro systems handle other important properties such as source location. Using such macros, Scheme programmers now routinely develop the ultimate abstraction: embedded domain-specific programming languages.

Unfortunately, a typical Scheme programming environment provides little support for macro development. The tools for understanding macro expansion are poor, which makes it difficult for experienced programmers to debug their macros and for novices to study the behavior of macros. At the same time, the language for specifying macros is limited in expressive power, and it fails to validate syntactic correctness of macro uses.

This dissertation presents tools for macro development that specifically address these two needs. The first is a stepping debugger specialized to the pragmatics of hygienic macros. The second is a system for writing macros and specifying syntax that automatically validates macro uses and reports syntax errors.
It describes two contributions: a debugger that's tightly integrated with macroexpansion, as well as a new expander called syntax-parse.

syntax-parse seems like it will become a standard facility for implementing macros. It lets you describe the shapes of the s-expression inputs of macros in a parser-like language. You can define new "syntax classes", that contain both information about the shape of a form, as well as additional, procedural checks.

One example of a syntax class would be the variables bound by a LET: they have to be identifier/expression pairs, and the identifiers must be unique. syntax-parse lets you describe these constraints succinctly, and facilitates useful error reporting (i.e. you don't get an error in terms of the expanded output of a macro, which would require understanding of a macro's implementation). This obviates the need for explicitly programmed checks.

syntax-parse seems like a big step forward for macro writing. The macro debugger is also impressive. Congratulations, Ryan!

No comments: