Connecting the dots

[tiedyedave]

I remember someone at CMU mentioning ongoing research in type-safe genetic programming (my money is on aleffert or qedragon). I have spontaneously sprouted an interest in type-safe reflection, and will probably be needing whatever is available in type-safe GP too. Do you guys (or whomever I was speaking with) have a URL handy for current info on this stuff? Or is it still too fresh?

Comments

[sleepsong.livejournal.com]

Non-sequitor:

Yesterday in conversation the following sentence came into existence:

"I would like to VI a bowel."

We thought you would be amused.

[ikeepaleopard]

Twas me. Actually most of the literature I've found on genetic programming was from the mid nineties. It seems to have mostly dried up as far as I can tell. I'll send you some links this weekend, but I found them just wandering around citeseer looking for strongly typed genetic programming.

type safe reflection

[(Anonymous)]

See Pfenning, "LEAP: A Language With Eval and Polymorphism" for what appears to be the state of the art (despite being ~10 years old). Taha's stuff on metaprogramming is also relevant, but is only runtime code generation (a subset of full reflect/reify).

Pfenning and Nanevski's work on explicit substitutions is (IMHO) the most elegant architecture, but they haven't developed it very far within the context of reflect/reify (they have one JFP preprint covering it, however).

The Cayenne paper on type-safe interpreters is another interesting direction. My personal hunch is that anything less powerful than GTT or UTT (Generalized Type Theory and Universal Type Theory) won't cut it.

I'm trying to adapt Quine's system NF for use as a "type system" (although it is a single-sorted theory of unstratified sets rather than of Russel-Whitehead types). I'm also very tempted to use Positive Set Theory, which has a much more constructive flavor than QNF, but there isn't enough foundational work on that and I don't think I'm qualified to make up for that.

I'm very interested in this stuff, but I generally keep research discussions on archived forums (preferably usenet); blogs have this awful tendency to vanish into nowhere on a regular basis. I read comp.lang.functional about twice a week, or you can email me offline (megacz@cs.berkeley.edu).

- a