My Workshop

This page is pretty much entirely of historical interest at this point. Oh, well.

If you like something that you see here, please feel free to take it and make it yours. But still free. Thanks.

Half-Baked

• HWEB, a new tool for what I would call "semi-literate" programming. Essentially comes in two parts: a fancy-text parser that understands some special (but fairly intuitive) markup, and a math parser that understands TeX-like markup. The current version is a mock-up that concentrates on getting the markup right, without doing anything about the "literate" side of things. Status: alpha, stalled. Snapshot here. Rationale here. Description here.
• The classic Crowther & Woods adventure game done entirely in HTML/CGI. Status: stalled. Snapshot here.
• A utility program for manipulating C code. Given a definition for one or more C struct types, it will produce C code to read and write objects of that type in a platform-independent manner, including serialization of large data structures when necessary. Status: stalled. Snapshot here.
• A program or suite of programs to make easier the entry and typesetting of Chinese text using the standard Unix tools available over a telnet connection. Should take a file of marked-up pinyin text, and produce the corresponding ideographs in some flavor of Unicode (see unitrans). Status: alpha, too naive for general use. Snapshot here.
• A hand-coded, clean token counter for C. This is in connection with Andreas' Practical Language Comparison. Please take a look, if you like to hack with C and think I might have missed a corner case. Status: beta. Snapshot here.

Pipe Dreams

• A Funge interpreter to end all Funge interpreters. I like the Befunge language, and if I had a really spectacular Funge implementation, I might actually try to learn Funge-98. Requires a lot of thought in order to straighten out the "fingerprint" and "IP mode" semantics, though. And I haven't really examined other Funge interpreters to see how they represent Funge-space; I've tried splay trees and found them reasonably fast, but not blazingly so. Tight loops in Funge code are hard, as far as I can tell. Maybe I'll get around to it this fall.
• A smarter indent than the GNU implementation. The GNU indent tool suffers from dictionary-itis: it has too many arcane options to control its behavior, and it still can't handle anything close to my own coding style. What would be nice is a "smart indent" that could read and absorb a corpus of code, and then indent any other code to conform with its style. Code can probably be processed as a combination of a token stream and a parse tree. The parse tree dictates much of the whitespace placement, while the individual tokens are what is actually output. Parallel structure between sequences of tokens ought to be exploited, and the internal format needs a way to describe connections between sections of code (dependent clauses, parallel structure, alignment in the original file). This part seems highly non-trivial.
• A writeup, with sample implementations, on the subject of horror in roguelike games, with special attention to Lovecraftian horror. The horror "feel" depends on suspense, which is a form of literary irony; how does the designer create a sense of literary irony in a game in which the audience controls the movement of the characters? Horror focuses on "everyman" characters, while classical roguelikes deal with characters who end the game as invincible supermen; how to reconcile these ideas? How in general to create atmosphere in a game whose interface consists of a top-down ASCII map and whose NPCs (human or monster) are usually described only with a single symbol and a two- or three-word "species name"?
• A MD5-based Web file-finder. I sometimes find myself in the position of having a program, image, or file of which I no longer remember the source. I can't use Google to track down the original location, because I might not have any keywords. But what I do have is the original binary data itself. So if a web spider were to remember, say, the MD5 hash of every page and file it ever saw, I could plug the MD5 hash of my file into the spider and it would be able to tell me that files with the same hash were available in such-and-such locations. This would take orders of magnitude less storage (and effort) than a full text search engine, and would provide a service sadly lacking from Google and all other major engines. (Bitzi provides the same service, but without the useful part: the automatic web spidering and the notification of where on the Web the given file could be found. It also provides metadata on the files, which would be redundant in this case.)