The Universe of Discourse

Wed, 13 May 2015

I did a residency at the Recurse Center last month. I made a profile page on their web site, which asked me to list some projects I was interested in working on while there. Nobody took me up on any of the projects, but I'm still interested. So if you think any of these projects sounds interesting, drop me a note and maybe we can get something together.

They are listed roughly in order of their nearness to completion, with the most developed ideas first and the vaporware at the bottom. I am generally language-agnostic, except I refuse to work in C++.

Or if you don't want to work with me, feel free to swipe any of these ideas yourself. Share and enjoy.

Linogram

Linogram is a constraint-based diagram-drawing language that I think will be better than prior languages (like pic, Metapost, or, god forbid, raw postscript or SVG) and very different from WYSIWYG drawing programs like Inkscape or Omnigraffle. I described it in detail in chapter 9 of Higher-Order Perl and it's missing only one or two important features that I can't quite figure out how to do. It also needs an SVG output module, which I think should be pretty simple.

Most of the code for this already exists, in Perl.

Orthogonal polygons

Each angle of an orthogonal polygon is either 90° or 270°. All 4-sided orthogonal polygons are rectangles. All 6-sided orthogonal polygons are similar-looking letter Ls. There are essentially only four different kinds of 8-sided orthogonal polygons. There are 8 kinds of 10-sided orthogonal polygons:

There are 29 kinds of 12-sided orthogonal polygons. I want to efficiently count the number of orthogonal polygons with N sides, and have the computer draw exemplars of each type.

I have a nice method for systematically generating descriptions of all simple orthogonal polygons, and although it doesn't scale to polygons with many sides I think I have an idea to fix that, making use of group-theoretic (mathematical) techniques. (These would not be hard for anyone to learn quickly; my ten-year-old daughter picked them right up. Teaching the computer would be somewhat trickier.) For making the pictures, I only have half the ideas I need, and I haven't done the programming yet.

The little code I have is written in Perl, but it would be no trouble to switch to a different language.

[ Addendum 20150607: the orthogonal polygon sequence is now in OEIS! ]

Simple Android app

I want to learn to build Android apps for my Android phone. I think a good first project would be a utility where you put in a sequence of letters, say FBS, and it displays all the words that contain those letters in order. (For FBS the list contains "afterburners", "chlorofluorocarbons", "fables", "fabricates", …, "surfboards".) I play this game often with my kid (the letters are supplied by license plates we pass) and we want a way to cheat when we are stumped.

My biggest problem with Android development in the past has been getting the immense Android SDK set up.

The project would need to be done in Java, because that is what Android uses.

gi

Git is great, but its user interface is awful. The command set is obscure and non-orthogonal. Error messages are confusing. gi is a thinnish layer that tries to present a more intuitive and uniform command set, with better error messages and clearer advice, without removing any of git's power.

There's no code written yet, and we could do it in any language. Perl or Python would be good choices. The programming is probably easy; the hard part of this project is (a) design and (b) user testing.

Twingler

Twingler takes an example of an input data structure and and output data structure, and writes code in your favorite language for transforming the input into the output. Or maybe it takes some sort of simplified description of what is wanted and writes the code from that. The description would be declarative, not procedural. I'm really not at all sure what it should do or how it should work, but I have a lot of notes, and if we could make it happen a lot of people would love it.

No code is written; we could do this in your favorite language. Haskell maybe?

Bonus: Whatever your favorite language is, I bet it needs something like this.

I want a simple library that can render simple pixel graphics and detect and respond to mouse events. I want people to be able to learn to use it in ten minutes. It should be as easy as programming graphics on an Apple II and easier than a Commodore 64. It should not be a gigantic object-oriented windowing system with widgets and all that stuff. It should be possible to whip up a simple doodling program in Crapspad in 15 minutes.

I hope to get Perl bindings for this, because I want to use it from Perl programs, but we could design it to have a language-independent interface without too much trouble.

Git GUI

There are about 17 GUIs for Git and they all suck in exactly the same way: they essentially provide a menu for running all the same Git commands that you would run at the command line, obscuring what is going on without actually making Git any easier to use. Let's fix this.

For example, why can't you click on a branch and drag it elsewhere to rebase it, or shift-drag it to create a new branch and rebase that? Why can't you drag diff hunks from one commit to another?

I'm not saying this stuff would be easy, but it should be possible. Although I'm not convinced I really want to put ion the amount of effort that would be required. Maybe we could just submit new features to someone else's already-written Git GUI? Or if they don't like our features, fork their project?

I have no code yet, and I don't even know what would be good to use.