Over the course of the last year or so I have begun working on a family of Scheme language designs as the basis for other research in which I am currently interested and, in the grand Lisp tradition, I have started off with meta-circular experimental systems. While driving to my day-job this morning I had a thought about the macro-timing problem that may have been forgotten in our long-standing efforts to have a "nice" static description of the language semantics. Of course, an actual working interpreter provides a very high-level (or possibly low-level depending on your point of view) formal description of the semantics using the implementation platform as providing a set of semantic primitives. So while I expect that most of the people interested in Scheme standardization will be familiar with the structure of a meta-circular interpreter, I think it is worthwhile to provide a brief outline.
A meta-circular interpreter (MCI) is typically defined by two mutually recursive functions: APPLY and EVAL. Broadly speaking, APPLY has the job of marshalling arguments and invoking abstractions, while EVAL has the job of finding values for primitive syntactic forms. If the reader is alert, you will note that there are probably *four* separate functions involved (binding management, syntactic reduction, argument marshalling, abstraction invocation), but the natural synergies between them have largely dictated the binary breakdown of the MCI into APPLY and EVAL. Which is all well and good, but what does any of this have to do with macros? Early Lisps treated macros as syntactic abstractions and therefore incorporated their reduction into the APPLY function. This led us to FEXPRs, and ultimately (I haven't done a deep study of it yet) to John Shutt's vau-calculus as implemented in the Kernel programming language. However, it seems that macros could also be viewed as syntactic *extensions* and as such be more properly considered as part of EVAL's functional remit. From this view it seems that we ultimately derive hygienic macros - as their reduction can be accomplished entirely in terms of textual manipulation with only minimal (actually zero I think) reference to anything outside of the syntactic reduction machinery in the MCI. I think it can be safely assumed that hygienic macros - in particular SYNTAX-RULES have found a certain "sweet spot" in terms of macro semantics. The problems come in when we try to break hygiene. Then we start having to cross, not only EVAL's internal distinction between storage management and syntactic reduction, but the stronger border between EVAL and APPLY. Now it is interesting to note that hygienic macros float the syntactic extension functionality outside of the entire EVAL/APPLY loop and into a conceptually separate pass. This makes environmental manipulations completely impossible to the macro processor. But my question (and challenge) is there a semantically sound way to model macros that returns them to their place inside EVAL? I am thinking in terms of an operator that can extend the reduction rules of the operational semantics f the language - perhaps that's daft in the way that I frame it, but it is certainly how the code *could* be (and certainly was) structured in some MCIs. Something to think about, especially w/rt Thing1 and first-class environments &cet... david rush -- GPG Public key at http://cyber-rush.org/drr/gpg-public-key.txt _______________________________________________ r6rs-discuss mailing list r6rs-discuss@lists.r6rs.org http://lists.r6rs.org/cgi-bin/mailman/listinfo/r6rs-discuss
