I was terrible in English. I couldn't stand the subject. It seemed to me ridiculous to worry about whether you spelled something wrong or not, because English spelling is just a human convention---it has nothing to do with anything real, anything from nature.

Richard P. Feynman

This repository is all of the original source material for my book Thinking with Types: Type-Level Programming in Haskell. If you're curious about what goes into writing a book, it might be a good place to peruse.

Building this thing is particularly hard; I had to write three separate build tools, and patch a few upstream libraries. You're free to try to figure it out, but I'd suggest just buying a copy instead!

Don't make me regret open-sourcing this.

My overarching organizational principle for this book was to make it as hard as possible to fuck up. That meant that code samples should automatically be tested, GHCi sessions should be automated, solutions and exercises should be co-located, and that there is always a clearly defined source of truth for all material.

The result was a joy to write, but remarkably terrible to deal with after the fact. Paying a marginal compile-time cost of 1s per code example is fine on a chapter-by-chapter basis, but my god does it add up when building the entire project.

Doing it in LaTeX was good for the short-term, but turned into an eventual liability. LaTeX is sweet for quickly producing good-looking pdf documents, but it's sort of the worst of all worlds. It's sort of a content-language, and sort of a real programming language, and doesn't force you into either paradigm. As a result, there was lots of weird fiddling in order to get something to look right---without knowing how it really works or without any discipline.

For writing a thesis or a report, this is fine, but the problem is an eternal one: it's not denotational. LaTeX emphasizes how to do it rather than what to do. The difference bites you in the ass when you want to produce an ebook, for example. You can't use LaTeX to produce the ebook, but you also can't not use LaTeX, because you've automated necessary things in its shitty programming environment.

Also, the tooling breaks all the time, seemingly without any sort of discernible reason.

If I were to do this project again, knowing what I know now, I would write the entire book as a series of Haskell modules. I'd use quasiquoters to write inline prose and build meaningful abstractions in a principled, well-understood language. In essence, I'd write a book DSL, and then write interpretations of that into my eventual desired formats.

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.